The Learning Technology Managers

GUIDE TO xAPI
Getting Started Case Studies Professional Insight
 LEARNING RECORD STORE
Make Learning Measurable

THE MOST INSTALLED LRS IN THE WORLD

Organisations use Learning Locker to connect systems together, to prove the impact of
training and to make more informed decisions on future learning design.

Now its your turn.

Find out more at HT2Labs.com/learning-locker
 CONTENTS

2
INTRODUCTION

Ben Betts, CEO of HT2 Labs, introduces the guide with a short piece on the
role of data in learning technologies.

4 PART 1: INTRODUCING THE EXPERIENCE API

In part 1 we get acquainted with the Experience API (xAPI), where it came
from and how it relates to existing technologies. Well explore how we can
get systems sending data in the xAPI format and how organisations use
xAPI data to improve their businesses.

14 PART 2: INTRODUCING THE LRS

Part 2 of our guide will introduce the concept of the Learning Record Store,
or LRS. Well investigate key questions like how your LRS might relate to an
LMS, where you might store your data and how youll secure it. Well also
introduce our LRS, Learning Locker, its key features and pricing.

20
PART 3: DEVELOPING XAPI STATEMENTS

Well get a bit more technical in part 3 as we look at the anatomy of an

xAPI Statement in more detail. Well go over each of the key elements that
make a Statement and give examples. Finally well give you three tips to
make interacting with an LRS just a little bit easier.

PART 4: EARLY ADOPTERS OF THE XAPI

Our final section is dedicated to case studies of early adopters who

32
are using the xAPI today. Youll read about a range of corporate and
education uses, including relating training activity to sales performance,
increasing learner success and creating a gamification system.

Credits Copyright
Authored by: Edited by: This work is licensed under version 4.0 of the Creative
Ben Betts & Ryan Smith Andie Coupland Commons with the restrictions CC-BY-SA. You can use,
study, apply, redistribute and create derivative works based
Layout & Design: Produced by: on this document as long as you give proper attribution
Ollie Newton HT2Labs www.ht2labs.com to the author, retain the license notice and release any
derivative works under an equally permissible license. See
creativecommons.org for license wording and more info.

HT2Labs The Learning Technology Managers Guide to xAPI 1

 IF ONLY EVERYTHING WORKED TOGETHER

In an age of where my phone knows when Im leaving the office, checks the traffic
home and tells me how long it will take, it doesnt seem like much to ask that
our learning systems work together. They dont seem to be the most complex
pieces of software in existence, so why is it that sometimes they behave like it?

2 ht2labs.com
 There are a range of problems When systems cannot share data
afflicting the learning systems What are the top three it becomes impossible for them to
weve gathered over the years. work together. The LMS doesnt know
From organisations with multiple data issues faced by when the simulation is complete. The
LMSs, to those also deploying talent App cant track back to the social
management systems, knowledge Learning Technology learning platform, and the social
bases, social networks, social learning platform doesnt do SCORM
learning portals, mobile apps and Managers? so we cant track that either. Content
simulations. The list of potential from 3rd parties? Forget about it...
software that the L&D department
could use is growing by the day. But, Getting control of learning data
for many people, its a real problem should be the top of the to do list for
just trying to get the existing systems 1 every L&D department.
talking to each other. Never mind
introducing something new and Incomplete data. It feels like you must Without the control that a single,
potentially impactful; the old stuff be tracking everything but somehow key accessible source of record can give,
just doesnt really work. data points get lost along the way. You you have no base from which to start
find yourselves with all sorts of expensive making things work. Time is of the
For learning technology managers compliance issues. essence. You need a basis in historical
tasked with realising the L&D data to start benchmarking current
departments ambitions, the pain performance against past results.
is real. Whilst leadership wants to Every day you are not standardising
get behind 70/20/10 principles and the collection of learning data is
start implementing next generation 2 another day you are letting the future
learning technologies, the mere lifeblood of the L&D department slip
thought of migrating to a new LMS Incorrect data. Potentially even worse, the away.
is enough to make most learning data you have might not be timely or even
technology managers want to cry. accurate. With no real standard way of This guide will introduce you to the
collecting data in the organisation, its up concept of the Experience API, or
But with an increasing marketplace to your vendors to have done a good job in xAPI as it is more commonly referred
for 3rd party tools that the L&D standardising and storing data. to; a specification for collecting,
department wants to integrate with storing and querying learning data
the business, learning technology that can help you establish a new
managers are learning that they need learning technology architecture for
to adopt a more service orientated your business. Its aimed at learning
architecture, plugging lots of small, technology managers working within
best-of-breed systems together, 3 L&D departments (but is equally
to offer up a playbook of tools and relevant to those working in other
technologies that HR colleagues can Siloed data. Youve got no way to get data areas). Well take it easy to begin with
use to improve learning outcomes. from one system to another, without calling and slowly ramp up the complexity
up IT. Adding in another system just gives to give you the complete guide to
Establishing a service orientated rise to more headaches with the first two getting started with xAPI.
architecture often starts with problems. And migrating to a new LMS?
establishing a method of getting Forget about it!
data from one system to another;
data reconciliation is a big, expensive
business. With no single source of
truth in the organisation, we find
ourselves trapped.

DR BEN BETTS - CEO, HT2 LABS

Ben leads the passionate team at HT2 Labs and is a globally-

recognised thought-leader in Learning Technology with more
than 15 years industry experience.

HT2Labs Introduction 3
 PART 1: INTRODUCING THE EXPERIENCE API

THE EXPERIENCE API OR XAPI FOR SHORT

is a specification document created by a consortium of learning experts led by

the Advanced Distributed Learning Initiative (ADL), which is a US Government
Programme. Whilst the document was being formed its prototype name
was project Tin Can, but these days it is known by its official name, xAPI.

xAPI is a document that, if out the answers, you dont have to give
you follow it, will give you complete access to a database.
a method for how you can
communicate learning One problem with APIs is that they can be
activity data between one quite specific; I can come up with the way
system and another. It is a I want my system to answer your questions
derivative of the term API, This is why we use and youll have to customise your questions
or Application Programming to match the information Im willing to give.
Interface. Before the advent the Experience API Simply having an API isnt enough; it must
of APIs, getting two systems be really well documented or even better,
to talk with each other (xAPI) over and subscribe to use some wider standard that
required a large amount we can all learn about, saving ourselves the
of code development and above any other API: hassle of using a new API each time we
integration time. APIs have want to make a connection.
transformed this relationship, its really relevant
meaning that basic The xAPI is an example of the latter;
integrations between different to what we do. a group of people coming together
software packages are now to decide on the language we will
relatively trivial. collectively use to build our interfaces
- a standard operating procedure.
Youll use APIs everyday, whether Because xAPI has been built by
or not you notice them. At its learning people, it uses language
most basic you can consider an that allows us to talk about things
API a bit like a Question and Answer we do in learning - like querying
stream: One system will ask the for assessment results or
other system a question, and get a understanding who was a
response in a format it can interpret. learners tutor.
So, one system might ask another, can
you show me a list of all the registered
users on your system? and the other
system would respond with a formatted list
of registered users. By using this Q&A style
format we can limit the access one system
has to another - your system will only answer
the questions for which it is authorised to give

4 ht2labs.com
 THE RELATIONSHIP BETWEEN SCORM AND XAPI

We do already have systems in Were also increasingly cognisant

place for tracking learning, the of the impact of informal learning
most common being SCORM (the experiences; stuff that happens
Shareable Content Object Reference outside of the sphere of L&Ds control.
Model) - which allows you to package How can we account for this?
up content to deploy on a Learning
Management System (LMS) and then SCORM has become inadequate
track who uses it. for our current and future needs.
Learning doesnt always happen on
There are similarities between SCORM the LMS. Enter xAPI.
and xAPI. The same government
department created both. In places xAPI gives us a looser framework that Expert View: Specification
they overlap with what they do and allows us to communicate between
how they do it. You could even say any two systems, not just between Writer, Aaron Silvers
xAPI is an evolution of the SCORM learning content and the LMS.
principles, but they are mostly quite SCORM began as a specification in 1997. Its
different. You wouldnt necessarily In creating the xAPI, its authors 20 years old, and still SCORM Version 1.2 is
replace SCORM with xAPI and vice couldnt readily put any IT system dominant as the means by which eLearning
versa. out of scope - we could be just as exists -- and when we think about working
easily learning on a phone, or a TV, or with learning technology systems, it is still all
SCORM allows you to specify a in real-life. This is where xAPI thrives, about courseware and SCORM.
package of content and the sequence allowing us to understand what
in which it must be navigated, and occurred on one system, in another SCORM Version 1.2 is so popular and so
other things too. With xAPI you cannot system, in a human-readable way. adopted that people for the most part
do this. never even bother with SCORM 2004. This is
In the short-term we still have a need regardless of the resources and years that
The API setup around SCORM is based for SCORM. Right now it isnt easy to went into improving SCORM from 1.2 to 2004.
on the principle that you have a package content up using some of
package of learning content and you our most popular authoring tools Even by 2009, SCORM Version 1.2 was so
want to load it onto an LMS for users and get it working with your LMS. The entrenched that no vendor really wanted
to launch and be tracked using. xAPI community is working on this to touch SCORM at all. If anyone was going
through initiatives such as CMI-5. to address mobile, social media, games,
But learning has evolved and were simulations or any other use cases, it would
no longer all about logging in to the But it is to the future, where we no need to be outside of SCORM.
LMS to launch learning content. Now longer consider packaging learning
we learn from micro-content, located content up, that most of our attention
in a wide variety of places. We use is focussed.
mobile phones, simulations, games
and more.

HT2Labs Part 1: Introducing The Experience API 5

 xAPI

6 ht2labs.com
 HOW DOES THE XAPI WORK? WHAT IS THE XAPI FORMAT?

In an xAPI world we talk about Activity Providers (APs), Learning If youve read much about the xAPI previously, chances
Record Stores (LRSs) and Activity Consumers (ACs). are youve come across the idea of an xAPI Statement,
that is, the manner in which we transfer data between
Activity Providers create data in the xAPI format and submit it the Provider, the LRS and the Consumer. That statement
to the LRS. APs are systems and applications on which learning takes the form of what we call a triple: An Actor (John
activities and events take place. Learning content, learning Smith), a Verb (completed) and an Object (a Learning
portals, apps and more can all behave as Activity Providers. Activity).
In an xAPI ecosystem we expect many Activity Providers to be
submitting data to the LRS at the same time.

Learning Record Stores are databases that verify that the input
matches the xAPI specification, storing all valid data for retrieval
by Activity Consumers, or by administrative users who wish to
access the raw xAPI data for analysis.

Activity Consumers are similar systems to Activity Providers (an

AP could in fact be an AC as well), in that they are typically
systems and applications that modify the user experience
based on xAPI data. This might be an LMS that checks off a
completed learning activity, because that activity appears in the
LRS. Or it could be something more complex; a leaderboard, a
badge issuing system or learning content.

For example, if you had a piece of eLearning content that was set
to track via xAPI we would consider that the Activity Provider; its
the source of the learning activity data. This activity data would
then be sent to the LRS in the form of an xAPI statement. The LRS
would accept the data (providing it is valid) and store it. Later on,
an administrator may visit the LRS to check on a learners progress, To extend the meaning of the statement, we add context
which is stored here. This data might also be consumed by another about John Smith, about the Verb, the Object and about
piece of eLearning content, making it both an Activity Provider and the wider circumstances in which the activity took place.
Consumer. When acting as a Consumer, the eLearning content This context might include details like where John Smith
could advance the user onto a point deeper into the learning was in the world when the activity occurred, what course
activity, bypassing content that the learner probably already the learning activity was part of, and how long John
knows, if the relevant previous experience is found in the LRS. spent on the activity, for example.

ELearning is just one example of an activity provider. Learning Using this simple format, you can understand whats
Management Systems (LMSs), mobile apps, portals, smart devices been done, anywhere - be it on an LMS, application,
and other software packages, like a CRM, could all be sources of mobile device, even on a point of sale device. In fact
xAPI data. Almost all of these sources could also be consumers pretty much any device that can be connected to the
under the right circumstances. We also see analytics programs, internet can be translated into this format.
certification and badging platforms and other Learning Record
Stores (LRSs) as likely consumers of xAPI data. Technically speaking, xAPI statements are parsed using
JSON - JavaScript Object Notation. JSON looks a bit
like XML but it is generally simpler to follow and more
lightweight - it takes fewer characters to write what you
mean. Whilst it might look a little odd at first, JSON is
quite readable as a format, as you can see on the next
page.

HT2Labs Part 1: Introducing The Experience API 7

 Take a look at the text block below. Thats how use the Void technique to render the previously
we say John Smith completed Activity 1 using stored statement redundant (more about
JSON. Notice that even at this stage weve got Voiding later). But you wont find an LRS with the
quite a bit more detail than the headline triple.: ability to directly edit or even delete a statement
as its not a feature of xAPI.
We know that John Smith is a person that
can be identified as by his unique ID 123 at Theres a lot more to statements than that,
http://www.example.com/users/, as well and well go into more detail about additional
as the name John Smith (which is kinda context later on, but statements arent the only
helpful as we might imagine a few John thing we can talk about using xAPI. In fact the
Smiths kicking about). xAPI specification actually gives us four different
We know that when we say completed we API endpoints to which we can submit data and
actually mean the definition of completed get data back. These are the Statement, State,
found at http://adlnet.gov/expapi/verbs/ Agent Profile and Activity Profile APIs.
completed.
And we know that John Smith worked on The Statement API allows you to create
statements and retrieve one or many
statements that you previously created, as
seen above.
The State API component of the xAPI allows
you to associate the relationship between
{ a user and an activity in a given state such
actor: { as a user progressing through a course,
objectType: Agent, e.g. User X is on level 4 in Tetris. The xAPI
name: John Smith, specification states that generally, this is a
account: { scratch area for Activity Providers that do
name: 123, not have their own internal storage, or need
homePage: http://www.example.com/users/ to persist state across devices.
} The Agent Profile is the component of
}, the xAPI that allows you to store some
verb: { additional data against a profile, such as
id: http://adlnet.gov/expapi/verbs/completed, whether that profile is part of a specific
display: { group that has access to additional
en-GB: completed settings or information.
} The Activity Profile API is a bit like the Agent
}, Profile API and allows you to store some
object: { additional information about an activity
objectType: Activity, in a profile, that can later be queried. It
id: http://www.example.com/activities/1 can be used in fairly simple things like
} leaderboards, or more complex things like
} language maps.

a learning activity that can be found at

http://www.example.com/activities/1 -
after all, there could be millions of learning
activities that define themselves broadly
as number 1, but there can only be one
learning activity at that web address at a
given time.
To avoid issues of data being tampered with,
or otherwise altered after it has been sent, the
Statement API is immutable - that is, you cant
edit or delete a statement after it has been
sent. If youve genuinely inserted a statement by
mistake (or with the wrong detail) then you can

8 ht2labs.com
 INTEROPERABILITY AND XAPI

xAPI gives us a structure by which it, we in the xAPI community refer to

we can communicate - a syntactic the idea of controlled vocabulary,
set of rules that specifies how were recipes and profiles to establish
going to talk about data. these interoperable standards.

We think about this in terms of the A controlled vocabulary is a Creating an interoperable

triple - the actor / verb / object - published set of xAPI terms (typically
and then the context. Some of these the verb) that we might use to approach to documenting
terms are very specific. For example, describe a particular action. For
you can only identify an actor in example, if we want to describe the social learning experiences
very specific ways; you must have way in which a user interacts with
an IRI (Internationalised Resource a video, then we can extrapolate
Identifier): to identify a verb. those terms to be relevant to any
occurrence of someone using a HT2 Labs first dealings with xAPI came as a result
But this doesnt mean were talking video online - play, pause, finish; of building our social learning platform, Curatr.
about the same things, just that the actions are the same.
were using the same structure. Using Curatr, learners were tasked with viewing
Indeed, sometimes we can use the The exact same set of terms can content, making comments and then adding back
structure in a conformant manner be used for every video player, their own contributions to an online learning portal.
but end up tracking two different meaning that data from one video Whilst we could track all this inside the system,
data points in different ways. Activity Provider is identical to data we couldnt readily export the data to any other
from another video Activity Provider. platform. This is when we discovered the xAPI and
In order for us to actually use xAPI it first set about adopting the specification as a
is not enough to adopt the structure, Communities of xAPI adopters have means of reporting on what learners had achieved
we also have to think a little bit more got together to publish their sets of in a social learning platform.
about the semantics of what were Controlled Vocabulary which you
going to put in the spaces that the can copy. This means that we all Since deciding to go with xAPI, one of the struggles
structure provides. generally define playing a video weve had to deal with is creating the right Recipe
with the definition held at http:// for tracking social learning interactions using xAPI.
If we dont do this we run into some adlnet.gov/expapi/verbs/played. For example, consider how we track a learner
big problems. liking another users comment. We could construct
a statement that says User A liked User B, with
For example, the verb run can the comment itself in the context. That allows us
mean many things - you can run to easily return how many times User B has made
a workshop, or you can run a comments that have been liked.
business, or a marathon. We have
to think about how we will use these But, semantically, its not true. User A didnt like
semantics consistently, not just in User B, they liked a particular comment. In which
our department, but throughout our case we would say User A liked User Bs comment.
business and even throughout our Thats more accurate, semantically, but now its
industry. harder to count how many likes User B has got,
because were more focussed on tracking how
To truly realise interoperability, we many likes User B got for that particular comment,
must be talking the same language which is unique, not every comment they
across organisational boundaries. contributed.

Fortunately the writers of the xAPI This is just one example of how adopting the
specification already realised this. specification has been somewhat harder than we
To help us define not just how we first thought it would be.
shape data, but what we say with

HT2Labs Part 1: Introducing The Experience API 9

A Recipe takes the concept of
Controlled Vocabulary further, WHAT OTHER OPTIONS DO I
giving you a complete set of
rules to encode a common HAVE, THAT ARENT XAPI?
activity. This goes further than
the vocab and into the structure,
objects, properties and the Weve come this far in life xAPI. Where you only need
required values associated with without xAPI so surely there to connect two systems
each part of an xAPI statement. must be alternatives to together in a fairly trivial
For example, a recipe for having collecting all this data? manner it is entirely
watched a video might encode appropriate to use a
the length of the video, such There are a few competing custom API. These will all
that we can always tell how specifications that you might be somewhat proprietary
long someone spent watching come across and there are - whilst you can follow the
the video in order to trigger the different techniques you documentation to connect
watched statement. can use to transfer data, but the two systems they will
xAPI is generally in a class very rarely work out-of-the-
Finally, we can aggregate recipes of its own. The only other box together, youll have
together to form what are known specification on the scene to customise elements to
as Profiles. Profiles detail a is IMS Caliper specification. get them talking. And when
complete set of statements for a This accomplishes a lot of you send data over an
particular context. For example, the same things xAPI does API between two systems
our video player profile could and even looks similar. you are in a fire and
specify that we make at least However, it is limited by forget environment. Once
two statements from a video what are known as its Metric the message is sent, it is
player; one at the start and one Profiles - these are a form received but never stored.
at the end. And so, just as you of recipe that are baked into That makes it hard for a 3rd
might have a recipe for video the specification. You can party to become involved in
playback, you might have one only use Caliper with the few the relationship. Using an LRS
for summative assessment, or for Metric Profiles that are to be with the xAPI is an easier way
social learning, or for tracking a found in the specification to connect multiple systems
mentoring session. itself - covering learning together using an already
activities like taking tests, standard approach.
The community has come watching learning activities
together to make profiles for and so on.
common contexts such as
SCORM content and CMI-5 (a IMS has taken a top down
potential SCORM successor). approach to dictating what
Where no relevant community you can talk about using
profile is present already, its Caliper, whereas xAPI is more
suggested that you publish crowd sourced, a bottom-up
your own and try to explain the approach. That isnt to say
rationale behind your decisions Caliper isnt useful; it is. But
for others to follow. its more likely to become a
set of recipes for xAPI than it
is to super cede xAPI. In fact
IMS and ADL, the two parent
organisations involved, are
actually working together to
make this happen.
A companion specification
that just describes xAPI Then there are other ways
vocab is available online: of talking between systems.
https://adl.gitbooks.io/ We already talked about
APIs generally speaking
and this is a potentially
viable alternative to using

10 ht2labs.com
 WHERE CAN I GET XAPI DATA?

There are a number of ways to start collecting xAPI data. You can integrate software applications that already come prepared to
output xAPI, you can customise the output of systems with APIs to translate their language into xAPI, or you can find ways to convert
existing databases or spreadsheet data into xAPI.

Existing Applications APIs and WebHooks Databases and CSV

There are more than 150 The majority of newer software Finally, if no Web Service is available
applications available in the world tools, especially SaaS-based but you can still get access to raw
that already export data in the xAPI offerings, come with some form of data, all is not lost.
format. These include Learning Web Service or API. This allows you
Management Systems (like Moodle, to setup scripts that will go and It will take a bit of custom
Blackboard or Totara), Content retrieve data from the application development, but it is possible to
Management Systems (Wordpress for use in your own. Its entirely convert data from a spreadsheet
and Drupal for example), Authoring possible to convert these APIs into or SQL database into xAPI format,
Tools (like Storyline, Captivate and xAPI format, given that the right given a bit of time, the right recipe
Lectora), Social Learning systems data is available. Sometimes you and access to a developer.
(like Curatr), Mobile apps (like need to set up scripts or triggers to
Tappestry or Learning Lockers access the API but more modern Typically developers create scripts
Acorn app), Performance Support applications implement WebHooks. that can Extract, Transform and
tools (like Red Panda, the Trek These bypass the need to query Load data from these sources into
Learning experience system, or xAPI the API, instead they push data out an LRS. In fact theres a whole
Apps), Assessment platforms (like directly to listening endpoints when industry dedicated to the ETL
TAO testing or Question mark) and an event occurs. Popular tools like process.
many more. Slack, SalesForce and Dropbox
all implement WebHooks, as do Typically scripts dont run in real-
It is becoming increasingly common integration technologies like Zapier. time; they might be used initially as
to see xAPI enablement as a key a means of loading data, or might
feature in RFPs put out by large run as a timed event (often referred
organisations. to as a Cronjob) every week or
overnight.
Zapier is worth a note on its own. This tool
and others like it (consider IFTTT) allow you
to transform data from one API into another
- like xAPI - without having big technical
knowledge. They make money based on how
much data you put through the services, but
as a route to a proof of concept they are well
worth investigating.

HT2Labs
Part 1: Introducing The Experience
API 11
 Typically, when you have to custom integrations you
connect multiple systems need to make, it also gives
together, you end up doing you a single source of
a fair amount of custom record for all this activity -
work to get them talking to the Learning Record Store.
each other.
In time this will give you
Sales folks for software more flexibility and choice
companies will tell you in the systems you procure.
that their technology Right now it can be a
will integrate seamlessly very demanding task to
because it has an API. migrate between Learning
But having just any API is Management Systems.
not enough - unless it is Much of this complexity is
HOW IS XAPI USED? a standard API, you will caused by learning records
still have to write custom being kept inside the
integration methods to get LMS, in some proprietary
There are many use cases for xAPI but it is most common systems talking. format. By adopting a
to see the specification used in connecting systems Learning Record Store
together, performing data analysis and facilitating There are many relevant you take control of your
performance support. standards to be learning data and keep it
considered. For example, separate from any system
Single Sign On has a you might want to replace
whole raft of standard in future.
approaches such as SAML
CONNECTING SYSTEMS or oAuth. But for sharing And because Learning
learning activity data, Record Stores are
there is really only one interoperable with each
For many organisations, getting to a single source of record to viable API that has been other, you can switch out
account for the learning activity that has occurred across all standardised - xAPI. your LRS provider with a
systems is a dream. minimum of fuss. This
When your systems use is what makes an LRS so
Its not uncommon to see multiple Learning Management xAPI they communicate different from an LMS -
Systems or a host of 3rd party tools be in use across a company. with each other via the there is a very specific set
As soon as you have more than one LMS or more than one place Learning Record Store. All of rules that account for
in which learners learn, keeping track becomes a headache. Activity Sources go in, all what an LRS must do. As
And given that the current trend is towards learners learning Activity Consumers come long as your LRS adheres
anywhere and everywhere, that headache is about to turn into out. Not only does this cut to these rules, you are safe.
a migraine. down on the number of

12 ht2labs.com
ANALYSIS AND BENCHMARKING PERFORMANCE SUPPORT

Presuming youve got some there arent hundreds of Whilst you might not get many Holy Grail answers
level of connectivity between successful case studies from from analysis you will probably get a whole heap
systems, the next thing youll xAPI that show how training of pointers on what works better, and again, xAPI
want to do is analyse how impacts performance, it is not can help here. If youve connected your systems
learning takes place across because xAPI doesnt work. together and done the analysis to understand that
these various systems. Using xAPI is simple. Its just the way when a person does X, they may need Y, you can
this data you can draw some we communicate. But actually trigger things to happen automatically.
conclusions about which finding a relationship between
forms of training actually two variables in a controlled, Imagine sending a customised email with revision
impact performance and repeatable manner is darn advice to a person failing an assessment 3 times in
which learning behaviours are near impossible. a row. Or sending the latest product video out to a
demonstrated by your highest sales person the day before they have a meeting
performers. Just because its hard, doesnt on the new product. Or texting new installation
mean you dont have to try. advice out to your on-site engineer right as they
To do that, you need to track Imagine using that as an turn up on-site. Commercial sites like IFTTT and
both the training that is excuse to the board in this Zapier have become huge working on just these
viewed and consumed, and age of analytics. Yeah, well I principles. Now you can do it within the confines of
the performance data. This is was going to prove the impact your own organisation.
why xAPI should be of interest of my training, but it turns out
to people even outside of its too hard. Thats a sure fire The idea of performing just in time interventions
the L&D department: the L&D way to get your budget cut whereby the system identifies that you need help
team will need access to this right when you need it has been prevalent in the
performance data if they Finally, its worth a note about learning industry for a number of years. As we
are ever going to be able to the value of data today for increasingly hear about the role of AI and Machine
test their hypotheses and to analysis tomorrow. Even if Learning in technology, you should know that it will
design better learning. you do not have the ability be through the standardised use of data that these
to analyse data that you techniques are actually brought to life. Its always
Analysis in this style isnt are collecting today, you dull behind the sexiness :)
straightforward and its not may develop that capability
immediately rewarding. tomorrow. Having a base in As a most basic example of Machine Learning,
You are stepping towards historical data will give your consider a performance support system that not
the world of scientific analysis more depth and give only sends out automated interventions but learns
experiments; attempting to the opportunity to benchmark when and how those interventions make a difference
understand the relationship current performance against to outcomes. If the system sends out help emails
between two variables. And past performance. Thats and the learner goes on to pass the assessment,
most scientific studies find why wed recommend it reinforces that the email is a good intervention.
absolutely no relationship organisations get started with If the emails have no effect on the later outcome,
whatsoever! xAPI sooner, rather than later. then the system learns that this approach has little
It will take time to build up a actual benefit and can try another approach. Right
Science is rigorous. Which meaningful amount of data now it might seem like Machine Learning is way
also makes it ponderous. for use in later analysis. away. But xAPI is shaping our ability to exploit these
If you are wondering why techniques in the near future.

HT2Labs Part 1: Introducing The Experience API 13

 PART 2: INTRODUCING THE LRS

A LEARNING RECORD STORE (LRS)

is a database that stores and retrieves data produced in the xAPI format. In an
ecosystem where training portals, mobiles, LMSs and other technologies are
all submitting statements on a moment-by-moment basis, the LRS is a key
piece of infrastructure that requires high availability and high reliability.

Its important to note that Building an LRS is far from

the term LRS is quite WHEN DO I NEED AN LRS? AM I trivial. There are more
specific. Where Learning than 1,500 tests for an LRS
Management System GETTING RID OF MY LMS? to pass flawlessly before
(LMS) can be used to refer it can be considered a
to a range of vastly different conformant LRS. Being
products, LRS refers to a You wont need an LRS until youve got at least one system producing conformant is important
distinct piece of software xAPI statements. But as soon as you start producing xAPI statements, because one of the
that adheres strictly to the youll need an LRS to store them. The good news is you dont need to absolute key functions of
xAPI specification for an get this right first time. LRSs are designed to be interoperable. You can the Learning Record Store
LRS. There is essentially a run more than one at once and share data, or readily transfer data is to check that data is
recipe card for building out from one to another, presuming that the data in the LRS is valid. valid according to the xAPI
an LRS; if it omits part Theres only really one fatal mistake you can make in getting started specification. And there are
of the spec or tweaks it and thats in the quality of the data you collect; garbage in, garbage a lot of nuances in doing
in some non-standard out. this: If your LRS does not
way, its not an LRS. do this to the letter, then
There are a number of fundamental differences between an LMS and you may be storing data
an LRS that mean you wont be getting rid of your LMS anytime soon. that you cannot use again.
An LMS will manage access to content, giving you the ability to upload
packages of learning content, assigning them to learners and then
tracking their progress. An LRS will do none of that - most learners
will never even be aware that the LRS is operating in the background,
let alone log into it. Your LMS will also probably deal with launching
eLearning content packages. This is not part of the xAPI specification
and doesnt form part of any standard LRS.

14 ht2labs.com
 SHOULD MY LRS BE

PART OF MY LMS?

No one wants to buy a new piece of

infrastructure; its a difficult decision that
will have time and costs associated with
it. Its attractive to think that an existing
system can be upgraded to include the
new functionalityfor instance, adding
an LRS component to your LMS. However,
your LMS is more likely to be useful as
an Activity Provider (in fact, its vital that
it does this so that you can track how
learners use the LMS as well as content).
But you will increasingly need to think
about a standalone LRS as you think about
three key issues: Connectivity, Analytics
and Scale.

Connectivity Analytics Scale

You dont need to be interacting on the LRS Although analytics are not a part of LRSs will need to store and process
to insert data. In fact, most users will never the xAPI standard per se, they are the vast amounts of data. Where we used
even see the LRS. They wont have accounts; reason you might want an LRS. The part to get a handful of data points from
they wont login. Most LMS setups require of the system that stores data is what a learner to the LMS, weve seen xAPI
users to access content and resources allows it to conform to the specification; activities generate 1,000,000 statements
through the LMS. But this is a huge barrier the ability to interpret this data is what in a day following use by a company
to measuring performance we dont makes an LRS a useful piece of software. at scale. This is an order of magnitude
login to the LMS to do our day-to-day work larger than the LMS generally deals with
and this is where the xAPI can really excel. An LMS is fundamentally not built around and dictates a different technology
data analysis; it is built around courses stack. Typically this means a NoSQL
An external LRS will come complete with and content. Its not enough to have your data store - something no mainstream
features like oAuth, a standard that allows LMS store xAPI data. Thats just the start. LMS currently available is built upon.
external applications to connect to data You are going to need some really powerful
within a system and to insert new data analytics to start answering performance
from external systems. Many LMSs are also questions. And much of that data might
setup to charge based on the number of not be the sort of results content driven
users submitting data to the system. With LMSs are accustomed to dealing in.
xAPI there is no limit to the number of users
on which data can be collected; you may
find yourself in for a big surprise when
your next LMS invoice comes around!

HT2Labs Part 2: Introducing The LRS 15

 WHAT SHOULD I BE

CONCERNED ABOUT WHEN

PROCURING AN LRS?

When we first start working with a new client we run through How much data will we
a checklist of questions to help us get the context for an
LRS installation. What follows here is a few highlights create?
from that checklist that will give you some key questions
youll want to think about when first procuring an LRS.
Its really easy to imagine that little bits
of JSON containing xAPI statements wont
add up to much data. And, in the singular,
youd be right. Most xAPI statements are
around 2KB in storage size (though we
have seen statements ten times that
Will the LRS be size!). But if you are making a million xAPI
statements a day this is going to add up
hosted in the to GBs of data in a matter of moments.

cloud or deployed Whilst cloud-based storage is generally

very cheap, the technology required
on-premise? to run an LRS at scale does not run well
on very basic equipment. For example,
One of the first decisions youll the SaaS version of Learning Locker
need to make is whether actually runs on 12 virtual machines at a
you want to host your own minimum. There is a trade-off between
LRS, or use an online service. the quantity of data you collect and
the level of detail you require in your
An online service will be quicker xAPI statements. Storing more than
to setup, probably cheaper you absolutely need can be wasteful.
short-term (unless your
labourer cost is zero), and will You should be able to estimate your
be tried and tested. However, storage requirements by understanding
you will need to be comfortable your Activity Providers recipes and
with data storage/ownership multiplying the output by the likely scale
responsibilities (is all this data of your audience. Getting this number
OK in the cloud or do you need accurate can be important in SaaS
on-premise?), and youll need circumstances, where you are most often
to be comfortable with the charged on the basis of data stored.
medium-long term costs of
continually paying for a service. Learning Locker is priced on a per GB of
stored data, so you pay $100 (75) for
each GB up to 10GB, then the price per GB
decreases. This model allows you to start
with a hosted LRS and a modest budget
whilst you see how much you use. Then
you can always switch to an on-premise
deployment if cost is becoming a factor.

16 ht2labs.com
 How is your data

secured?

Data security remains a hot

topic and your Learning Record
Store is no exception to the
trend. You should be interested
in how your xAPI data is secured
whilst in-transit to the LRS and
how it is secured at-rest in the
database. There is literally no
excuse to NOT use SSL whilst
sending xAPI data to the LRS.
And most cloud providers
(including Learning Locker) will
offer you encryption-at-rest.

Are there

any location

restrictions?

When implementing xAPI you How much data can be lost in

should become hyper-aware
of your organisations data worst case scenario?
protection privacy policies and
the geographic differences
that might occur. For example, The LRS quickly becomes data could you afford
does your organisation require a key part of your to lose and hope to
data to be stored in a particular infrastructure. If you are recover normal operating
geographic location? Or, storing all organisational practice? For many
perhaps more likely, are there learning data on it then you circumstances putting in
particular areas of the world cant really afford for it to some fallover mechanism
you are required to avoid go down, or worse, to lose and also doing off-site
storing your data-at-rest? data. The killer question for daily backups is enough.
redundancy and backup But in high-risk or testing
is always how much data environments, even
can you afford to lose? that might not be. How
And dont forget can you get that 24hr
Of course the preferred number down to 1 or 2
archiving... answer is none but that hrs without breaking the
tends to be unrealistic in bank? How much money
Because xAPI is an immutable the face of cost/benefit will you have to spend
specification (you cant edit or analysis. In the worst- to reduce this number?
delete stored statements) your case scenario, how much
data set is just going to grow.
To keep costs down and keep
servers running efficiently, youll
need to develop a process
for archiving old data. If you
dont the LRS is going to get
pretty big in years 2, 3, 4

HT2Labs Part 2: Introducing The LRS 17

 LEARN ABOUT LEARNING LOCKER

Learning Locker is the Learning Record Our hosted option currently offers a more
Store developed by HT2 Labs. Available advanced toolset than our Open Source
as both Open Source (for on-premise download. This includes the following
installation) and Enterprise (for Hosted, features, available exclusively to Learning
Software-as-a-Service), Learning Locker Locker Enterprise customers:
is one of the most popular Learning Record
Stores available on the marketplace. Customisable Dashboards
Create new dashboards at the click of a
Learning Locker comes ready to accept button. Use our drag and drop interface to
data in the xAPI format. With your data position the visuals you need, where you
flowing, Learning Locker allows you to need them. Export it all at the click of a
create customisable dashboards using a button
WYSIWYG interface. Advanced API options
are available to sort and share learning In-built Charting Library
data back to Activity Consumers, including Weve pre-built graphs and queries for you
the ability to connect data to Business to use with your xAPI data. Leaderboards,
Intelligence tools. frequency charts, correlation graphs and
more are all possible out-of-the-box.

Learner Journeys
Build up learning journeys from xAPI
statements and visualise completion on
Open Source checklist your dashboards. See who has completed
which journey, how long it takes to
Are you ready for Learning Locker Open Source complete journeys and more.
or should you go Enterprise? if you can answer
yes to these questions youre ready for our Open Trigger Events and Web Hooks
Source option: Use our pattern matching to trigger events
in external systems in real-time without
1. Do you provision Linux servers, set up the need to query the LRS.
load balancers and schedule backups?
2. Do you install and configure server User Matching
software using tools like apt-get or yum? Use our algorithm to automatically
3. Do you use git commands to manage reconcile users with multiple IDs across
your source code? the system.
4. Do you run unit tests to ensure the
integrity of your software? Ready to start? Check out the pricing
page opposite and then visit us online to
Ready to go? Visit: get your Learning Locker account:
https://www.ht2labs.com/learning-locker/
open-source-lrs/ https://www.ht2labs.com/learning-
locker

18 ht2labs.com
 ENTERPRISE PRICING PLANS

First 10GB Next 10GB*

Price per GB $100 / 75 per GB $75 / 55 per GB

*Larger plans available

How much space do I need? The Small Print...

Typically most Activity Providers output Our minimum monthly fee is $100 / 75 per
statements that are around 2KB in size. Where month. You can choose which currency (GBP or
this is the case, you will be able to store around USD) you would like to be invoiced.
500,000 statements per GB. Thats the equivalent
of 1,000 employees generating around 500 xAPI Plans are payable 1 month in arrears with data
statements each over the course of a month. usage rounded up to the nearest GB.

If you are tracking something like learning content For example (in GBP), if you used 500mb in month
(maybe from Articulate Storyline, for example), 1 you would be charged 75.
you will probably generate only relatively modest
amounts of data (less than 1GB). If, in month 3, you used 2.75GB you would be
charged 225.
However, if youre tracking a work-based system
or a simulation, your usage is likely to be greater. If then, in month 10 you used 11GB you would be
charged 805 - calculated as (10*75) + (1*55).
Get in touch with HT2 Labs and we can generate
an accurate estimate of your space requirements
based on your use case.

VISIT: HT2LABS.COM/

LEARNING-LOCKER

HT2Labs Part 2: Introducing The LRS 19

 HOW DO I WRITE AN XAPI

STATEMENT?

Beyond some very simple proofs of concept you

are going to want your systems to write your
xAPI data. Doing it by hand isnt a feasible, nor
desirable approach. That said, you will still want
an understanding of the data being created and
you might even need to step in and customise it on
occasion. This part of the guide will give you a brief
introduction to the Statement API and the variables
held within each part of a typical xAPI statement,
starting with the Actor, then the Verb, then the
Object and then, everything else!

PART 3: DEVELOPING
XAPI STATEMENTS

THIS NEXT SECTION WILL COVER

the creation of a basic xAPI statement by going into some

detail about what each element of a typical xAPI statement
may contain. It gets a bit technical in places but it is worth
the time trying to understand the structure behind xAPI as it
will mean data analysis, and actually using your data, will
become a lot easier down the line.

UNDERSTANDING THE

ACTOR ELEMENT OF AN

XAPI STATEMENT

So, first things first - an actor can be one of two things:

an Agent or a Group. An Agent represents a singular
person and a Group represents well, a group of
people. Youre probably going to use an Agent for the
most part. An Agent can be identified in 4 different
ways: mbox, mbox_sha1sum, openid, or an account.
The most robust tends to be using an account.

20 ht2labs.com
 Agent Actors

Agents can be identified in 4 different ways

The mbox is the users email address:

{
objectType: Agent,
name: John Smith,
mbox: mailto:[email protected]
}

mbox_sha1sum is an encrypted version of the email address:

{
objectType: Agent,
name: John Smith,
mbox_sha1sum: 4445904ac65039ef7a91506207f19162ac4dea73
}

The openid is a URL:

{
objectType: Agent,
name: John Smith,
openid: http://www.example.com/johnsmith
}

The Account type is a unique identifier, plus a Homepage

URL

{
objectType: Agent,
name: John Smith,
account: {
name: 123,
homePage: http://www.example.com/users/
}
}

HT2Labs Part 3: Developing xAPI Statements 21

 Group Actors

Group Actors can be either identified or anonymous

As with Agent Actors, a Group can also be identified by the

above 4 elements - in which case it is an Identified Group.

{
objectType: Group,
name: HT2,
account: {
name: 123, Both group types can have members - if the group is not
homePage: http://www.example.com/users/ an Identified Group, it is an Anonymous Group and must
} have at least 1 member.
member: []
} {
objectType: Group,
name: HT2,
member: [{
objectType: Agent,
name: John Smith,
account: {
name: 123,
homePage: http://www.example.com/users/
}
}]
}

An Example Actor Statement Summary

If youre wondering what a basic statement looks like, here it is:

{
actor: {
objectType: Agent,
name: John Smith,
account: {
name: 123,
homePage: http://www.example.com/users/
}
},
verb: {
id: http://adlnet.gov/expapi/verbs/completed,
display: {
en-GB: completed
}
},
object: {
objectType: Activity,
id: http://www.example.com/activities/1
}
}

22 ht2labs.com
 Managing Multiple IDs

One significant challenge to be overcome when

connecting multiple systems together is reconciling
user ID. It is increasingly common to see Single What is an IRI? Or a URI?
Sign-On (SSO) systems in place that make each
user within your organisation easy to identify, no At various points in this document you will see us describe an
matter which system they are logged on. But, where identifier as a URL, or an IRI, or maybe even a URI.
SSO isnt so prevalent, or where each user isnt A URL you probably know. Thats an Internet address. But what
issued a unique login, or where you are integrating about URI and IRI?
systems outside the organisations jurisdiction, then
you may find yourself storing data for the same URL - Uniform Resource Location. This is the link at which a
user under two completely different IDs. Sometimes resource resides on a network. This is what you are most likely to
this means having an mbox and an account for be referring too when using a link on the world wide web: http://
the same person. (And sometimes even multiple www.ht2labs.com is a URL for example.
mboxes.)
URI - Uniform Resource Identifier. A unique identifier for a
Learning Locker can automatically reconcile resource that can be on a network or somewhere else. A URL is
multiple IDs into the same persona for simple actually a type of URI, where the resource in question is identified
querying. This feature is a potentially very important by its network access. An ISBN number is a URI designed to name
requirement if multiple IDs are likely to be a reality a book in a unique manner and at worldwide level.
in your organisation.
IRI - Internationalized Resource Identifier - is a generalization of
When a new statement is submitted into Learning the URI. While URI supports only ASCII encoding, IRI fully supports
Locker, the system starts a worker in the background international characters
to try and figure out if weve seen this actor before.
Sometimes thats really straightforward, like when So why bother with these elements? Its all about being able to
the same account ID is used. reference something uniquely. Like the way only one person in
the world can have your mobile phone number, only one location
If thats not the case Learning Locker will check on the in the world can be described by a URL. That makes whatever
other metadata available from the new statement resides at the end of that address unique.
(further personally identifiable information like
name or a grouping) and try to match it against For example, we use a HomePage as well as an Account ID when
existing users. You can configure Learning Locker using the Account actor type. How come? Well, whilst your user
to be more or less rigorous in this process using a ID (123456) might be unique in your organisation, there could
scoring mechanism. easily be another 123456 in another company.

If the scoring mechanism finds a match, Learning But if we add your unique ID to a homepage URL that represents
Locker will merge the two actors into the same where you come from (E.G. http://www.ht2labs.com), then we
persona. This process doesnt modify any xAPI data, can refer to you as user 123456@http://www.ht2labs.com. And
it simply adds a reference within the system. If the theres only one of those!
scoring mechanism indicates that the new user is
similar, but not the same as an existing user, youll xAPI makes use of IRIs (which are nearly always URLs in practice)
be prompted to make a manual reconciliation. for the identification of verbs, activities and more. Whilst it is best
Then any remaining new users that cant be found practice that a Verb IRI resolves to some explanation of the verb
will automatically be created as a new user within meaning, it isnt actually required to work to be spec conformant.
Learning Locker.

Problem solved!

HT2Labs Part 3: Developing xAPI Statements 23

 UNDERSTANDING THE

VERB ELEMENT OF AN

XAPI STATEMENT
An Example Actor Verb Summary

Verbs get a lot of attention within the xAPI

world. Whilst they are the headline of what {
someone did, they dont carry a lot of detail actor: {
as to the context in which they did it. So whilst objectType: Agent,
Verbs are important, they arent very complex. name: John Smith,
account: {
A verb can contain two properties - an ID, which name: 123,
uniquely identifies the verb and its meaning, and homePage: http://www.example.com/users/
the Display - which tells the LRS how to display }
it in a human-readable format (and can be },
mapped into various languages for translation). verb: {
id: http://adlnet.gov/expapi/verbs/completed,
The ID qualifies the meaning of the verb being display: {
used (e.g. do we mean John fired a gun, en-GB: completed
or that he fired an employee). We do this }
by using a unique IRI to describe the Verb },
. }
Its important to note that an IRI does not need
to resolve to be unique (i.e. it doesnt need
to exist as a live Web address). However, it is
best to use a resolving link that displays more
detail as to the Verb definition where possible.

There are no official verbs in xAPI, you can

use whatever you want. But you dont need to
reinvent the wheel when choosing which ones to There is one exception to the lack of official verbs - thats
use in your statements - in fact, you are much the verb Voided. This is a special, reserved Verb that has
better off using existing verbs where possible. a particular meaning. When the LRS receives a Voided
statement it will look to cancel out the Object referenced.
A number of resources come in handy to check In this way we can remove erroneous statements from
against Verbs that other people have already being returned by the LRS. Any statement that has been
used; see the list at the bottom of this page. voided will behave as if it was never created. This isnt the
same as deletion; the original statement still exists, it just
If a verb that you wish to use isnt currently never gets found when searching.
listed in the registry you use we recommend
that you ask for it to be created. You can make
a new IRI to define your Verbs meaning, just try
not to re-invent what others have done before.

ADLs Controlled Vocabulary (http://xapi.vocab.pub/datasets/adl/)

Tin Can Registry (https://registry.tincanapi.com/)

xAPI Quarterlys xAPI Index (http://xapiquarterly.com/the-xapi-index/)

24 ht2labs.com
 UNDERSTANDING THE

OBJECT ELEMENT OF AN

XAPI STATEMENT

An Example Object where Type is Activity

In the xAPI world an object can be four things:

Activity (e.g. something that you do/did)

{ Agent/group (e.g. an individual, or group of
actor: { people)
objectType: Agent, Sub-statement (e.g. a nesting of
name: John Smith, statements)
account: { Statement Ref (e.g. a reference to another
name: 123, statement, normally used to void an
homePage: http://www.example.com/users/ existing statement, or refer to a statement
} that is yet to exist).
},
verb: { A simple object statement looks like the one
id: http://adlnet.gov/expapi/verbs/completed, opposite; an Activity that has an ID which is
display: { an IRI. In this case were saying John Smith
en-GB: completed completed http://www.example.com/
} activities/1.
},
object: { If you set your Object as an Activity you have
objectType: Activity, to declare that it has an ID (and that it is an
id: http://www.example.com/activities/1 Activity). You can extend the meaning further
} using Activity Definitions (see next page for
} an example). These definitions make your
Statement more useful in later querying. In
particular Name and Type can come in useful,
as these allow you to group like activities
together.

Your Object doesnt have to be an Activity.

It is perfectly acceptable for your statement
to be about people instead of activities. For
example, if John Smith followed Ben (in the
Twitter sense), you could set the Object type to
Agent instead of Activity and identify Ben in
the Object space.

Finally, your Object can also be about other

xAPI statements. For example, when you Void
a previously submitted Statement you use
the Statement Ref to refer to the previous
Statement that you wish to Void (using the
unique Statement ID that was created when
the xAPI Statement was first recorded in the
LRS).

HT2Labs Part 3: Developing xAPI Statements 25

Activity Definitions
An Example Object

An Object requires an ID (which is an IRI),

an object type and, optionally, further
definition. {
actor: {
If youre looking to do more intensive objectType: Agent,
querying on your LRS data, you should set name: John Smith,
the definition property. account: {
name: 123,
This further definition can have a number homePage: http://www.example.com/users/
of properties: }
},
Name - describes the activity in verb: {
short, readable format. id: http://adlnet.gov/expapi/verbs/completed,
Description describes the activity in display: {
more detail; again human readable. en-GB: completed
Type (IRI) - suggests what sort of }
Activity this Object was, e.g. a video, },
an assessment, a simulation, etc. object: {
More Info (IRL) - a link to further objectType: Activity,
information about the activity, which id: http://www.example.com/activities/1,
is useful if the description is long. definition: {
Extensions - A series of further name: {
extension that can be used to better en-GB: completed
identify the Activity or where it took },
place, for example longitude and description: {
latitude coordinates of where an en-GB: completed
activity physically took place in the },
world. type: http://adlnet.gov/expapi/activities/question,
Interaction Properties - Tend to be moreInfo: http://www.example.com/activities/1/more,
used for describing SCORM-like data, extensions: {
such as question response patterns. http://www.example.com/ext: true
},
The key thing to be aware of when interactionType: choice,
programming your Object statements correctResponsePattern: example
is that although only the ID is required, }
you really do need to go into detail when }
setting your definitions. Without this level }
of detail you wont be able to filter your
statements back out of your LRS data.

This process we often refer to as designing

for data. For example, if you want to know
who has watched the most videos it is
far easier to lookup where Activity Type
= Video than it is to look up every single
Object ID that happens to be a video.
And if you wanted to know the physical
location of your learners you would need
to be storing the Long/Lat co-ordinates
at the time of the Activity. This data is
immutable and cant be changed after-
the-fact, so you need to get it right.

26 ht2labs.com
 UNDERSTANDING THE

RESULT ELEMENT OF AN

XAPI STATEMENT

Now that weve covered the Actor > Verb > Object
An Example Result triplet, its time to turn towards other attributes
we can add as context. One of the most sort
after pieces of context is the Result - a score and
A complete statement set up correctly with a result will look something like associated information that might come as a
this: result of taking a quiz or assessment.

{ Result statements can include a number of

actor: { optional details:
objectType: Agent,
name: John Smith, Score
account: { scaled (a percentage between -1 and 1)
name: 123, raw (the actual score recorded)
homePage: http://www.example.com/users/ min (minimum score achievable)
} max (the maximum score achievable).
},
verb: { Success - Pass or Fail.
id: http://adlnet.gov/expapi/verbs/completed,
display: { Completion - Yes or No.
en-GB: completed
} Response - e.g. the actual comment/answer
}, given in response to a question.
object: {
objectType: Activity, Duration - how long it took to complete the
id: http://www.example.com/activities/1 activity.
},
result: { Extensions - e.g. more detail about the result.
score: {
scaled: 1, Using xAPI you can track the result of each
min: 0, individual question within an assessment or the
max: 100, assessment as a whole. For example, we
raw: 100 could submit each question answer as an xAPI
}, Statement and then submit
success: true, a parent Statement that
completion: true, references earlier question responses. In this
response: Example string, manner we can record the overall result as well
duration: P1DT12H, as a question-by-question analysis.
extensions: {
http://www.example.com/ext: true Later analysis is really quite straightforward
} using the Result context and xAPI. Now we can
} quickly evaluate the average score for an
} assessment, which questions are easiest (or
hardest) and how long people spend answering
each question.

HT2Labs Part 3: Developing xAPI Statements 27

 An Example Context

A complete statement with context may look something like this:

{
actor: {
objectType: Agent,
name: John Smith,
account: {
name: 123,
homePage: http://www.example.com/users/
}
},
verb: {
id: http://adlnet.gov/expapi/verbs/completed,
display: {
en-GB: completed
}
},
object: {
objectType: Activity,
id: http://www.example.com/activities/1,
}
context: {
registration: 957f56b7-1d34-4b01-9408-3ffeb2053b28,
instructor: {
objectType: Agent,
name: Joe Bloggs
account: {
name: 321,
As you can tell, if you include all of homePage: http://www.example.com/users/
this data in your statement then }
its going to get pretty long - but if },
youre looking at really getting the team: {
most from your data, then there are objectType: Group,
a couple of additional elements you
name: HT2,
account: {
might want to consider...
name: 123,
homePage: http://www.example.com/groups/
}
},
contextActivities: {
parent: [{
objectType: Activity,
id: http://www.example.com/activities/parent
}],
grouping: [{
objectType: Activity,
id: http://www.example.com/activities/grouping
}],
category: [{
objectType: Activity,
id: http://www.example.com/activities/category
}],
other: [{
objectType: Activity,
id: http://www.example.com/activities/other
}]
},
revision: 1.0.0,
platform: Example,
language: en-GB,
statement: {
objectType: StatementRef,
id: 957f56b7-1d34-4b01-9408-3ffeb2053b28
},
extensions: {
http://www.example.com/ext: true
}
28 ht2labs.com }
}
 ADDITIONAL PROPERTIES IN YOUR XAPI

STATEMENTS

Statement ID Version

Each xAPI statement stored is given a Unique The version of the xAPI that the statement
identifier for the statement - the UUID. conforms to, which is set to the latest version
by the LRS if you dont set it.
The LRS will set it for you unless you explicitly
set it yourself - which some might choose 1.0.3
to do so that they can use the ID elsewhere,
without querying the LRS for it. Attachments

Youll get an error if you create an ID that These might represent certificates or badges
already exists - so you must be aware of what that have been sent with the statement to
else exists in the LRS. The UUID is a long string: be stored in the LRS. Only the attachments
headers are stored within the statement,
8caba439-5455-485d-8c99-b3680ef8e46f such as the content type and length of the
attachment, rather than the data itself.

{
Timestamps usageType: http://www.example.com,
display: {},
There are two ISO-formatted timestamp description: {},
properties in a statement. They look like this: contentType: application/json,
length: 100,
2016-05-27T13:06Z sha2: 9cfe7faff7054298ca87557e15a1
0262de8d3eee77827417fbdfea1c41b9ec2
The timestamp property which 3,
represents the time that the experience fileUrl: http://www.example.com/
took place; files/123
}
The stored property which represents
the time that the LRS stored the
statement.

The stored property is set by the LRS, but you

can set the timestamp property. You can
use this feature to track offline activities that
sync with the LRS at a later time. The time at
which the event actually took place would be
recorded with the timestamp property; the
time it finally hit the LRS would carry the stored
property.

If you dont set the timestamp property, the

LRS will set it to the same timestamp as the
stored property.

HT2Labs Part 3: Developing xAPI Statements 29

 A Complete xAPI Statement

So what does this all look like when put together in one
statement?

{
actor: {
objectType: Agent,
name: John Smith,
account: {
name: 123,
homePage: http://www.example.com/users/
}
},
verb: {
id: http://adlnet.gov/expapi/verbs/completed,
display: {
en-GB: completed
}
},
object: {
objectType: Activity,
id: http://www.example.com/activities/1
},
id: 8caba439-5455-485d-8c99-b3680ef8e46f,
timestamp: 2016-05-27T13:06Z,
stored: 2016-05-27T13:06Z,
authority: {
objectType: Agent,
name: John Smith,
account: {
name: 123,
homePage: http://www.example.com/users/
}
},
attachments: [{
usageType: http://www.example.com,
display: {},
description: {},
contentType: application/json,
length: 100,
sha2: 9cfe7faff7054298ca87557e15a10262de8d3eee77827417fbdfea1c41b9ec23,
fileUrl: http://www.example.com/files/123
}]
}

30 ht2labs.com
 SENDING XAPI DATA

We know that xAPI is parsed using JSON and sent over a Web
Service. But how do you make this process bulletproof? How
can you be sure your xAPI data arrived at the LRS? Heres three
tips to help you out:

1 Use standard code libraries to send

activity data.
The easiest way to be conformant with
the specification is to use standard code

2
libraries to generate your Statements. You
can find these on Github as linked to by
the ADL - there is one for each of the major Setup a queuing system to send xAPI data.
web programming languages. By using For most production systems it is not good enough
these libraries to create xAPI data in the to simply fire and forget xAPI statements; they
first place you know it will be valid. should be queued and tracked to make sure they
actually get to the LRS.

For example, if the LRS wasnt available for any

reason, would the Activity Provider cache/store
the statement to be resent at a later time? A
conformant LRS will notify an Activity Provider
when a statement has been successfully stored
(returning the statement ID created as a result of
new data being stored). This makes it relatively
easy for a queuing system to retain an xAPI
statement until it is certain the LRS has received,
and stored, the data.

Batch your xAPI statements.

If your Activity Provider is sending a lot of
statements you may want to investigate

3
chunking up requests to place less load on
your LRS. Its generally easier for a server to
process 1 request with 1000 statements, than
it is to process 1000 requests with 1 statement
each. Both methods are perfectly valid; the
LRS can differentiate between the observed
time and the stored time of statements.

HT2Labs Part 3: Developing xAPI Statements 31

 PART 4: EARLY ADOPTERS OF THE XAPI

MORE THAN 2,500 ORGANISATIONS HAVE

DOWNLOADED AND INSTALLED THE OPEN

SOURCE VERSION OF LEARNING LOCKER.

With more than 150 applications directly created to work with xAPI there is
no doubt that the adoption of the specification is becoming mainstream
throughout businesses, education and government organisations. In this final
section we will highlight a few of the use cases weve been lucky enough to work
with over the previous two years. We arent able to name the organisations
using xAPI and Learning Locker in every case but rest assured, you know their
names!

USING XAPI TO ENABLE GAMIFICATION

THROUGHOUT THE LEARNING PROCESS

We worked with a major energy provider to implement a new gamification As data flowed into the LRS these challenges
system. The aim of the project was to incentivise sales personnel to engage were automatically completed based on user
with further training and then benchmark their sales performance back interaction. Successful completion of a challenge
against their participation in training. would lead to digital badge being issued to the
user.
The audience was competitive by nature and responded well to game-like
challenges. Product knowledge was spread amongst multiple systems (LMS, To show results back to users, a series of widgets
Intranet, eLearning content, Apps and more), meaning the gamification were constructed including leaderboards and
engine would need to take account of user engagement from all of the performance graphs. These widgets were
available systems. integrated with the existing CMS such that
administrators could insert gamification elements
HT2 Labs implemented Learning Locker to become the central point of to highlight performance on any given page.
data warehousing between the systems and worked with the client and its Finally, a series of dashboards were built for
partners to make Activity Providers export xAPI data based on usage. HT2 the customer to inspect the performance of its
Labs then devised a gamification engine that would enable the customer to distributors, including the ability to benchmark
create challenges based on paths of content to be completed. geographic areas against each other.

32 ht2labs.com
 Company Intranet
eLearning Authoring tools Learning Locker (Enterprise Edition),
Learning Management System running on AWS

Activity Provider Learning Record

Store

Solution

LMS
Company Intranet

On Page Widget

New OAuth Access

User Token
xAPI
eLearning Content
Data

Dashboards

Learning Record Store

Solution Activity
Consumer

Custom gamification engine

The customer highlighted a need to foster greater engagement Custom dashboards
amongst distributors with available product knowledge. Research Company Intranet
had shown that those distributors whom engaged more with learning
materials went on to drive a higher level of overall sales. But how could
they drive engagement?

HT2Labs Part 4: Early Adopters of the xAPI 33

 Research suggests that the use of Learning Analytics
by tutors and administrators in Higher Education
can improve learning outcomes for students. This
includes benefits drawn from Student Success tools
that accurately predict whether or not a student is at
JISC: OPEN LEARNING risk of failure, as well as deeper analysis of learning
content and programmes. But the implementation
ANALYTICS ARCHITECTURE cost of these tools has been a real turn-off for
universities looking to engage; there are significant
barriers to entry in time and cost. Jisc created a
In early 2014, Jisc set about creating a set of online tools 2m project with the aim of providing a basic set of
for UK universities and colleges to better understand learning analytics tools, capable of drawing data from
their learner data and make more informed decisions a range of sources, at no initial cost to the university.
so that they might increase student satisfaction,
retention, and attainment. Called the Open Learning Jiscs solution was to be predicated on standardising the
Analytics Framework, Jisc would make the new way in which universities collect activity data about their
tools available to any UK university, free of charge. students. Jisc selected the xAPI as the only viable option
currently available in education to fulfil this remit. As a
result of their decision, Jisc went on to procure a suitable
Learning Record Store: Learning Locker was selected due
to its Open Source nature and proven ability to scale.

Working as part of a consortium of vendors, HT2 Labs has

implemented Learning Locker for Jisc and advanced its
functionality significantly to include additional security
features and to expand the record store into a Data
Warehouse, capable of storing student data as well as xAPI.

Additionally, HT2 Labs worked with partners to create best-

in-class integrations with Learning Management Systems
like Moodle and Blackboard. Thanks to this work, LMSs
that represent more than 90% of the UK install base for
Higher Education now directly export their data to xAPI.

Predictive tools like Unicons Early Alert System and Tribals

Student Success Planner are now capable of taking their data
directly from Learning Locker. Jisc have succeeded in creating
an Open Learning Analytics service, to which universities
can be quickly subscribed and start sending in data without
the need for any coding or complex integration time.

34 ht2labs.com
 Moodle
Student Info Systems
Learning Locker (Enterprise
Blackboard
Edition)

Learning Record
Store

Activity Provider
Activity
Consumer

Unicon Early Alert System

Tribal Student Success
Civitas Learning
Student Mobile App
Tableau

Solution

To assess the quality of

their learning offering when
compared to the national
average for like programmes.

HT2Labs Part 4: Early Adopters of the xAPI 35

 Curatr
SAP

Activity Provider

VILLEROY & BOCH: PROVING

RETURN ON INVESTMENT WITH

XAPI DATA

Winners of the Best Social Learning project at the Learning

& Performance Institute Awards, 2016, V&B were able to
use the data collected by xAPI to prove the link between
training and performance throughout their retail stores.

Learning Record
Store

Learning Locker
(Open Source Edition)

Custom dashboards

Activity
SOCIAL AND COLLABORATIVE Consumer
THE
LEARNING
LEARNING AWARD
AWARDS
2016 GOLD WINNER

36 ht2labs.com
 Solution

Curatr is capable of exporting all

learning activity data directly to
an LRS using the xAPI. Here data is
stored globally for V&B in a format
that can then be used in more
complex analysis and reporting. By
blending this data from Learning
Locker with insights as to sales
performance taken from SAP (the
global ERP system for V&B), the
company was able to demonstrate
clear links between the uptake of
training and subsequent regional
performance increases. The
following figures demonstrate just
part of the Return on Investment
seen:

Average Transaction Value is

V&B had a strategic requirement to PDFs, presentations or even other web- +32% since rolling out More
develop a new digital learning capability to pages. By linking these assets together to Excellence in Selling (which
reduce time to competence and increase form levels, V&B encourages learners to has a focus on upselling) to UK
sales effectiveness of their global retail view content and contribute to discussions and USA
workforce (circa 12,000 employees). With in order to unlock content at the next level Since rolling out the basic
no internal capacity and no budget to - a gamified approach to the learning Excellence in Selling in Iberia
outsource content development, V&B experience. Comments are welcome on in May 2015, sales are +15%
needed to tap into the existing knowledge nearly every item of content and discussion In Japan, Porcelain Product
held by their remaining workforce - by points are raised to get the conversation Knowledge was rolled out in
crowd sourcing content and best practices started. The better learners contribute to June 2015. Sales in this product
from the people on the sales floor. the conversation, the more points they earn. category are +19% since that
time
V&B selected HT2Labs Social Learning The implementation went very well
Platform Curatr as the software to deliver but whilst Curatrs internal metrics
on these requirements. Using Curatr, the suggested engagement was high,
L&D team assembled courses quickly using V&B still hadnt made the leap to
any sort of digital asset as content - video, actually proving a link to performance.

Solution

HT2Labs Part 4: Early Adopters of the xAPI 37

 Activity Provider

Custom Learning Portal

CV Builder
3rd party applications

City & Guilds TechBac is Skills Wheel; a personalised

an entirely new professional dashboard of progress which
CITY & GUILDS TECHBAC programme designed to give Learners can export to their
14 to 19 year olds an exciting own tailored CV, customising
FROM EDUCATION TO alternative path towards an the information they present
apprenticeship, higher education back to potential employers.
EMPLOYMENT USING THE or employment. A key feature of
the TechBac is the Skills Zone, Students progress is monitored
XAPI a unique portal that brings through the Performance
together workplace skills training Optimiser, an interactive
with a mentoring programme, dashboard. This system
This project won Best in Show at DevLearn business challenges, takes data directly from the
2015 for its ingenious use of xAPI to connect a and an online CV Builder. LRS to provide a real-time
series of learning systems together seamlessly. benchmark for each cohort
The different systems are linked undertaking the TechBac.
together using the Experience It allows tutors to highlight
API (xAPI). Using this approach, areas of risk to the Learners
City & Guilds are able to store progress and allows them to
learners progress in a Learning intervene at the right time.
Record Store (LRS), unifying all of
the systems into a single record Tutors can recognise the
of progress and achievement. Learners competence in each
of the seven skill areas by
City & Guilds selected Learning awarding a Mozilla Open Badge.
Locker to power their solution, The Mozilla standard makes
allowing TechBac to scale accreditation evidence-based
upwards of 50,000 learners and portable. Learners can
with peaks of intensive usage export their Open Badges to
throughout the year. Learning their CV and the wider web as
Lockers advanced API is used proof of their achievements.
to process and display data to
both learners and their tutors, This portability is achieved
as well as reporting back usage using the xAPI standard; the
BEST OF SHOW VENDOR to City & Guilds for analysis. first solution to combine both
2015 xAPI and Open Badges into
Learners visualise their a production environment.
work on the City & Guilds

38 ht2labs.com
 Learning Locker
(Enterprise Edition)

Custom dashboards
Mozilla Open Badges

Learning Record
Store

Activity
Consumer

Solution

HT2Labs Part 4: Early Adopters of the xAPI 39

 VISUALISATION TOOLS AND
ANALYTICS FOR ONLINE

LANGUAGE LEARNING (VITAL)

A collaboration between 3 universities across Europe and

HT2Labs, VITAL is a European Funded initiative founded
with the aim of allowing students and teachers to better
benchmark language learning across the continent.

40 ht2labs.com
 Moodle
Blackboard
Custom Learning Portal

Learning Locker
(Open Source Edition)

Activity Provider

Learning Record
Store

Solution

VITAL aims to identify how students of higher education across Europe

learn languages online by taking a bottom-up approach learning analytics
approach: Mapping and feeding back to the students and their lecturers the
existing learning patterns in 4 different types of online language learning.

Easy-to-use dashboards for non-specialist users are a key part of this project
to ensure both students and their tutors:

For students so they can not only understand their own data about how
they learn online, but also to be able to compare their profiles to those of
their peers

For educators to get dynamic and real-time overviews of how their

students are progressing, who might be at risk of dropping out or of failing
for the course, and which parts of the courses cause difficulties/require
more feedback

Activity
Consumer

Solution

Custom dashboards

Despite the fact that many HE institutions have

embraced eLearning and other learning technology
solutions, the Vital consortium has found that there
is still no way for either the learning institution or the
students themselves to form a clear picture of their
online learning habits.

HT2Labs Part 4: Early Adopters of the xAPI 41

 Activity Provider

Curatr

INTERCONTINENTAL Solution

HOTELS GROUP: IHG wanted to try and measure submissions at the end of week
the actual impact on behaviour assignment to assess the quality
MEASURING THE VALUE of running a MOOC. Without of the reflective conversations
evidence as to the quality of the occurring on a 3-point scale:
OF SOCIAL LEARNING learning intervention it would
be hard to assess the actual At level 1, learners were
WITH XAPI impact the MOOC had on the simply expressing what
organisation, beyond a range of they had watched, read
feel good engagement metrics. or understood during the
InterContinental Hotels Group (IHG) IHG felt the MOOC had been weeks learning activities
commissioned a new Massive Open Online successful, but they lacked really At level 2 they were
Course (MOOC) to help its first level leaders to hard evidence that it had made a expressing a desire to
have better quality feedback conversations with difference to the organisation. change; that they might
their reports. The course would run over a 5-week try to apply some of their
period and encourage learners to contribute back Whilst the MOOC was going new insights at some future
to the online classroom, as well as consuming on, HT2 Labs was collecting point
learning content. The aim was to actually change evidence across systems using And at level 3, they were
the way in which leaders gave and received Learning Locker and the xAPI. reporting back on how they
feedback in the workplace, on a global scale. With this source of data HT2 tried to do something new,
Labs data scientists were able reflecting on how it went
to do something really clever: and on how they might do
They were able to analyse things next time as they try
conversations retrospectively to form a new habit of better
to see how many people were conversations.
actually doing something
The results of this analysis are
different in the workplace by the
shown on the following graph:
end of the 5-week experience.
The results were ground-
breaking.

Using an approach to measuring

the quality of online conversation
Award Winners in a learning environment
known as Cognitive Presence,
HT2 Labs were able to map the

42 ht2labs.com
 Learning Locker
(Enterprise Edition)

Custom dashboards

Learning Record
Store

Activity
Consumer

Within 4 weeks of the social experience starting, IHG could show that 50% of participants
had seen how they could change and 12% of participants had actually tried a new
behaviour. A remarkable insight into the impact a social learning experience has had on
individuals, and collectively, the organisation.

Solution

HT2Labs Part 4: Early Adopters of the xAPI 43

 SUMMARY

xAPIs rise to prominence in its first 4 years of existence has never really been
seen before in Learning Technology. It took SCORM a decade or more to catch-on
and many other specifications have tried (and most often failed) to capture the
imagination of the Learning Technology industry. Whilst xAPI might remain firmly
at the start of the hype cycle for some, those of us in the know recognise the sea-
change that is upon us. Learning Technology and training in general has always been
under pressure to prove its value to the organisation. We have no excuses not to
measure everything that we do and xAPI gives us a method to start doing just that.

xAPI wont help you to prove the value of something that has no value, quite the
opposite in fact. But it will help you showcase the difference training can make
to individuals and your organisation as a whole. For most of us, proving such
things will be a tricky road. But immediate value can be had from connecting
systems together - creating that single source of record that will form your basis
for analysis in years to come. And we cannot emphasise enough the importance
of starting today; the quality of future analysis will be predicated on the quality of
the data you are collecting today. Standardising your approach to this systematic
collection of data should be top of your agenda as a Learning Technology Manager.
And with this guide at your side, you should be in a good place to get started.

For more resources, including our free MOOC - An Introduction to the xAPI - which has
helped thousands of Learning Technology Managers like yourself get to grips with xAPI
for the first time, check out the resources section of our website, HT2labs.com/resources.

HT2 Labs is dedicated to helping organisations like yours improve lives through learning technology. Were experts in xAPI
and in learning analytics. We developed the worlds most installed Learning Record Store, Learning Locker, and weve been
the vendor at the heart of some of the most complex xAPI initiatives in the world. We are here to help, no matter how small
or large your problem.

To talk with an advisor about how to get started with xAPI, or to kick up a free trial of Learning Locker, please do get in touch
today by emailing [email protected]

44 ht2labs.com
 Learning Record Store

Connect systems together, prove the impact of training and make

better decisions on your future learning design.

Learning Analytics Business Rules Identity Management

Cloud Deployment SSL Secured Encrypted at Rest

Find out more at HT2Labs.com/learning-locker