Customer Journey Framework

Adam Wills – GSMA, Mobile for Development Impact

September 2015

1
 Introduction
Background to the framework Three service examples
• This framework references work from 3 recent service
• Our earlier publication Getting the most of your data in M4D
Concepts evaluations in order to provide concrete examples. The services
identified that mobile service usage data is largely a “missed
A missed grounded in made reference to are:
opportunity” across the Mobile for Development sector and that
opportunity real examples
more can – and should – be done to support organisations
make greater value of their existing data.
HNI - offers public service ACRE Africa – offers micro- NextDrop - offers water supply
information via Airtel 3-2-1 on a insurance, using SMS and timing information to urban
• In particular, we found that a significant number of range of topics such as health, mobile money, where the citizens in India, using SMS &
A framework agriculture and gender in the insurance premium is paid on IVR over basic phones, as well
organisations across the sector lacked appropriate tools,
to address local language via IVR, SMS & behalf of the farmer by a seed as aggregating water service
frameworks and approaches to dealing with such data
the need USSD company as part of the purchase information from ‘valvemen’
• – this framework is designed in response to this need. of a packet of seeds using IVR & smarthpones

What are the primary barriers to using your existing service data M4D Impact provided bespoke evaluations for:
for deeper analysis?
6% Lack of appropriate skill sets

28% Lack of appropriate tools

39%
Lack of frameworks, or approaches
to analysing data
Lack of time or funds
28% Human Network NextDrop
ACRE Africa
International
Source: M4D Impact survey, 2014

2
 Objectives of the Framework
What we will cover… What we won’t cover…

Helping managers understand their service A one size fits all approach doesn’t exist
An approach to performance to a greater level of depth A one-size fits all It’s more important to think about ‘what to measure’ in
assessing operations customer journey each case

Analytical approaches vary case-by-case (e.g. based on

Using existing Understanding mobile usage data’s delivery technology)
potential value in driving new operational In depth examples of It’s more important to grasp the overall approach, then
operational service
insights particular analyses investigate details of running analysis when a need is clear
data

We advise on open-source and freely available tools (e.g.

Generated through associated analysis that
Identifying barriers to Prescription of which iPython)
identifies areas of focus to unblock barriers
service success analysis tools to use Organisations invest in different tools & skillsets & need to
to success
adapt them across frameworks

Target audience: Those managing delivery of mobile services targeting Flexibility is key: This approach is designed to be flexible across deployments in
underserved populations different sectors, accounting for differences in business model and service designs

3
 Elements of the framework
Our approach to conducting evaluations that support M4D
In 9 sequential steps
services

This document outlines M4D Impact’s approach to mobile service

M4D Impact’s 1 Establishing a plan of action
evaluations, built from our knowledge of conducting evaluations with a
approach diverse range of mobile products aimed at low income populations in
developing markets 2 Mapping out the business model

3 Setting Commercial/Social Objectives

Our approach makes use of the raw mobile usage data that these mobile
From raw mobile service providers have access to, but are often unable to extract the full 4 Assessing available data sources
usage data value from
5 Constructing the Customer Journey
The approach also draws on customer-centric qualitative research
Using customer- techniques coupled with data analytics on mobile usage data, linking 6 Using data analytics to identify bottlenecks
centric methods findings with behavioral drivers
7 Developing qualitative research briefs

Most importantly, the approach is driven by the business needs of the

8 Analysing and presenting results
Driven by business product, linked to clear commercial and social success factors.
needs Recommendations are given in the context of a rich understanding of the
business model underlying the product 9 Monitoring, implementing & evaluating

4
 1
Establishing a plan of action High Level Work Plan

Month 1 Month 2 Month 3

Start with a plan of action, appropriate resourcing and expertise

Senior Business Data & Qualitative & Final
Preliminary
leadership problem Business customer report
report
project buy- statement model diagnostic
The high level plan on the right outlines the basic in – sign off & data diagnostic
Follow the high level approach we took to running different evaluations, samples
plan received
it can be adapted across different cases
1. Business model analysis
Secondary research /
business model & Business
Key milestones include project buy-in and objectives sketch model &
secondary research, running in-depth diagnostics, objectives
finalised
Key milestones and delivering the reports back to business Synthesis &
decision makers Preliminary report write Final report
up - Lead write up
2. Data analytics

We recommend a team with three areas of Data

Data
testing
expertise: extraction
& cleaning
•Business model analysis / strategy Customer Synthesis &
3 areas of expertise •Data analytics - with database & large dataset journey Preliminary
analysis
analysis experience report write up - Final report
3. Qualitative research analytics write up
•Qualitative research – with ethnographic/
Human Centered Design (HCD) experience Qualitative Synthesis &
brief Preliminary
preparation report write up - Final report
Note: The timeline for turning around a project is qualitative write up
variable between 2 – 3 months

Note: The plan assumes report output is equivalent to ‘full report’ as for
HNI, and that this is not the team’s only area of focus.
 Example: Planning two in-market visits with ACRE Africa
Month 1 Month 2 Month 3
Diagnostic & Qualitative engagements are both key components

Senior Business Data & Qualitative & Final

Preliminary
leadership problem Business customer report
report
Data & business model Clarify objectives with
project buy- statement model diagnostic
ACRE Africa team & data
in – sign off diagnostic
• We created a plan in advance with the ACRE Africa samples
Deep dive on all aspects of received
team; interviewed marketing, product, and other
business model
team leads, the mobile operator and seed company 1. Business model analysis
to investigate the agriculture insurance product
Secondary research /
Interview all key partners
business model & Business
• This and early data analysis clarified that driving objectives sketch model &
registration at points of distribution was critical to Extract all data / objectives
investigate preliminary analysis finalised
Synthesis &
Preliminary report write Final report
up - Lead write up
2. Data analytics
Qualitative Research & Customer Investigate Data
Data
Insights testing
extraction
& cleaning
• A bespoke research engagement focused on a key Customer Synthesis &
distribution point, and the interaction between farmers journey Preliminary
analysis report write up - Final report
and agricultural merchants during the sale of a packet of
seeds 3. Qualitative research analytics write up

Qualitative Synthesis &

• This assessed behavior and attitudes that might limit brief Preliminary
farmer registrations, and opportunities for driving preparation report write up - Final report
greater future registrations qualitative write up

6
 2
Mapping out the business model
Service business model must be broken down into ‘Mobile’ components exhibit common business model
components patterns
• We suggest using the business model canvas below as a way to break down and • There are some common features across mobile service models, especially
investigate the mobile service when considering distribution, customer relationships, and benefits (e.g.
considered from the mobile operator perspective)
• Models vary widely in the M4D space, and without specific understanding of the
structure, evaluations will risk leaning on false assumptions about how the service
works

How do you
How do you interact? What are the benefits? (MNO)
How do you do it?
Who will help How do you interact? Who do you Customer relationship (over technology
KEY ACTIVITIES Direct benefits
you? help? help? channel)
CUSTOMER
PARTNERS VALUE RELATIONSHIP CUSTOMERS
PROPOSITION Increase Increase
Mobile Mobile
connections ARPU
infrastructure Money

What do you
How do you reach Machine to SMS, USSD,
need? Reduce Reduce
them? Machine IVR, data,
KEY churn costs
RESOURCES (M2M) apps
DISTRIBUTION

How do you reach them? Indirect benefits

Distribution channel
What are the costs? What are the benefits?
Above the Improve Improve value
Agent brand of mobile
COSTS REVENUE Line (ATL)
network perception ownership
marketing

Below the Increase

Partner
Line (BTL) ability to use
network
Customer Value Distribution & Key activities Cost/Rev marketing mobile
Partners relationships and resources
Proposition structure

7
 NextDrop, a story of three customers and three models

Services and their history are not as simple as

expected on first glance Water utility
Registering users
Utility:
Direct
relationship
Mapping the
(BWSSB): - Better information with end-users
distribution
Required to sign off on the network as they can
network B2G:
on ND rolling out potential issues report issues
with water from
• NextDrop offer water supply timing information to urban citizens in India. While the in a city Collecting info from - Better image
ND Water utility
valvemen + - Engineers less
value proposition sounds simple, their business model has pivoted multiple times building ties disturbed by calls Shared branding
(BWSSB)
Grant from end-users with water
providers/funder
• The initial model was Business to Customer (B2C), but it was too hard to balance the s Collecting feedback utility of FMCG
from users company
cost/revenue equation with monthly subscription fees from end-users - GSMA End-users:
B2C:
- VC Save time as they
- Development are better informed
• The model then pivoted to Business to government (B2G), where the water utility Innovation Lab NextDrop
End-users
(government led) pay for an employee monitoring system to better monitor their Valvemen: - Media
IT providers: Historical data Less disturbed by
network - M-governance
IVR platform and Location identification phone calls from website
SMS platform process (GEO end-users - Referrals from
• The model pivoted yet again, when an Fast-Moving Consumer Goods (FMCG) CODER) existing B2B:
company showed interest to access the large base of users, offering a Business to Other suppliers: Platform + algorithms FMCG: customers
Smartphone (tech IP) Access to a large - Past ND rep
Business (B2B) sponsorship model FMCG
supplier (for customer base of (door sign up)
Valveman app/IVR
valvemen) loyal users
Human capital (<100)

Utility:
Valvemen monitoring system
contract
Lesson: Staff : IT Valvemen incentives:
-Don’t leave any rock un-turned, the model will be more complex (bulk of the platform: Airtime (Hubli) and FMCG: Co-branding contract
cost so IVR + smartphones (new roll
than you expect far) SMS outs) Grants (e.g. GSMA or Development Lab): counted as
-Few models in this space are mature enough to straightforwardly revenue

replicate, so don’t blindly assume similarity

8
 3
Setting Commercial & Social Objectives
Identify central commercial and social objectives Break down the business using a pyramid
before anything else approach around key levers

• In conjunction with understanding the business model the headline social and • Then break the logical dependencies for achieving these commercial and
commercial objectives for the mobile product must be clearly and social objectives down into mutually exclusive and collectively
unambiguously defined exhaustive factors

• This process should be conducted in close discussion with employees • These sub-factors should be sufficiently granular & measureable, e.g.
who have a deep understanding of the product in order to ensure the right number of users registered, or value of transactions paid by brokers to
metrics are isolated farmers in value chain

• By examining relationships in the system, levers for creating the desired

value can be better identified

Key
Commercial Value Social Value
Optimise what? levers

Mutually Exclusive &

Commercial Value Social Value Collectively Exhaustive

Granular factor Granular factor Granular factor

…. 9
 Breaking down an agricultural micro-insurance product:
it’s all about registration
Driving registrations is key across both commercial
and social dimensions

Social + Commercial Impact

• ACRE Africa offers micro-insurance where the insurance premium is
paid on behalf of the farmer by a seed company as part of the
purchase of a packet of seeds. The farmer simply needs to register by = # registered packets * XX% of average premium – overheads (fixed)
sending an SMS with a unique code at the point of planting, and is
insured against the circumstance of no rainfall inside a defined
planting window.
How can ACRE increase profits and impact?
• The commercial value is straightforwardly tied to profits, while the
social value is tied to the number of farmers who have increased
mitigation against lack of rainfall. Both factors increase with farmer Key
registrations: increasing the number of farmers insured & increasing
Drive the number registered packets lever
profits since costs are fixed with a fixed average premium revenue
associated with each packet registered.

Social Commercial
Lesson:
- Keep it simple, you need to be clear about the commercial/social
objectives and what you’re optimising Increases the number of farmers Increases profits for ACRE
- Don’t try to measure more than a few key things at first insured against seed germination
failure

10
 4
Assessing available data sources
Different service models and delivery channels will Run early tests with available data to assess what
create different ‘data opportunities’ behavior can be analysed

• With objectives established the feasibility of what Start with raw • Generally raw transaction logs will look something like
can be done with the data must be assessed. For transaction the simplified, dummy example below. Logs record
Feasibility mobile services that have gone live and have logs unique customer identifiers, timestamps, and actions
customers, we look for a record of all customer The exact time the The exact time the The choice made,
Sessions are User IDs
which may require
transactions, i.e. a ‘free record’ of behavior. unique, the same (scrambled) can customer dials the
service (e.g.
user made a choice
(e.g. information, or another table to
user can have generally be linked
IVR/USSD) given to send money) decipher (e.g. 1 =
multiple sessions to mobile numbers
health info)
• This record will naturally differ across delivery
Technology / technologies and service types. For example the Session ID User ID Call Time Choice Time Choice ID
service type data records will be different across IVR, USSD,
1 71beb 01/10/14 00:15
SMS & Data/Web, as well as between mobile money
and information services. 2
c5717 01/10/14 03:30 01/10/14 04:56 1

• Generally data is available from different sources, 3

Different data check service line agreements with providers, e.g.
c5717 02/10/14 00:15 02/10/14 00:44 15
sources platform vendors or operators if in direct partnership
Then assess • Once the meaning of fields and supplementary data is
feasible given, discussions can begin on what is possible to
analysis analyse
Note: for another example considering mobile money transaction
data see the technical notes from the GSMA’s Mobile Money team, E.g. In the above, we can segment the user base by how many registered users accessed the service last
month, how many ‘made a choice’ and how many accessed & made a choice more than XX times – the
available on request decision about whether such a segmentation is valuable depends on the previous exercises with business
model and commercial/social objective mapping.

11
 Examining what’s possible with IVR data – a
refined notion of ‘active’ use
3 unique users in this Choices are empty
IVR data is readily available HNI, but which success metrics are snippet of the where the user
possible to distill from the data? transaction record doesn’t pass the
IVR Information
service
(IDs scrambled, not
MSISDNs)
home menu, leaving
null values
The dummy snippet on the right shows the transaction log of an IVR table. It shows
3 unique users, making choices (or not) to access information from menus and then
individual messages (choice ID) on an IVR system.

A basic calculation to understand valuable user behavior is ‘active use’, e.g. has the user Duration of
dialed the service in the last month? 3/3 users would be active on this definition in our Session ID User ID Call Time* Choice Time Choice ID call after
choice (sec)**
example – but can we do better?

More nuanced behaviors can be looked at with the data: 1 71beb 01/10/14 00:15

Have users called in multiple times? 2 c5717 01/10/14 00:15 01/10/14 00:44 1 11
Frequency
Note: 2/3 users dialed in multiple times in our example

Have the users looked at multiple content areas or just one 3 c5717 01/10/14 03:30 01/10/14 04:56 1 32
Diversity Note: 0/3 users looked at multiple content areas in our example
4 d0fc5 01/10/14 03:34
Have the users listened to messages in full or not?
Engagement Note: 1/3 users have listened to a message in full in our example
5 71beb 02/10/14 00:15 02/10/14 00:44 13 20
In HNI’s case, listening to messages in full (engagement) is a logical precondition of social behaviour change,
and since the commercial value was seen to be based on indirect revenue benefits to the operator,
engagement was also a good fit for tracking commercial value * a user doesn’t access content by just dialing in, they must first make a choice to listen to content
** a user must listen for >30 seconds to listen to a message “in full”

Check your data is clean: in another service example it was discovered that Lesson: Make sure you look at the fields and experiment, this is critical in

! timestamps for the same event differed across datasets – the most basic issue
can lose you a week if you don’t spend enough time to understand the data first
understanding what customer segments will be feasible to construct with the
data

12
 5
Constructing the customer journey
Data is a key dependency, and unique service Use a basic model to start – but tailor to fit your
designs must be factored service
The journey below is a useful basic model to start thinking about the
The customer journey is a set of events that define customer’s journey from a state of low awareness to regular use of a
key experiences in the life cycles of customers. product – it can and should be adapted to fit different service models
The Customers will progress through such a journey in
Customer their evolving use of a service. The aim is to have Non-Aware Aware Understand Register Trial Use

Journey them gravitate towards an ideal use case. Data can User is not
aware of
User has
become
User is
aware &
User has
registered
User has
tried the
User is
regularly
be used to quantify how many customers sit at the service aware of the understands for the service, using the
service the service’s service maybe just service
different points of the journey value once

With the data constraint & unique service features in mind,

Assessing available data sources provides a the most appropriate customer journey segments must be
natural constraint on what journey segments can Tailoring identified - the key is that later stages of the journey
The data and cannot be described (e.g. SMS is limited, to fit correspond with the kind of value that the service is
constraint aiming to create, which depends on the social and
IVR is much better since more customer behavior
commercial objectives outlined beforehand
is visible in service usage logs)
Commercial Correspon
Also consider the unique features of a particular value
Social Value ds with Ideal
Usage
Unique service that may be important to include in the
service journey – e.g. a pre-registration phase or a post-
features trial phase Understa
Non-Aware Aware nd ??? ??? ???

13
 Example: Segmentation of customers who register
and have repeat use of the insurance
How can ACRE Africa increase profits and
Driving registrations is key, how can this be best reflected in
impact?
the customer journey map?
= how can ACRE Africa drive the number registered
In the case of ACRE Africa’s micro-insurance product, driving registration was key:
packets ?
Many farmers had access to the product - having bought a packet of insured
Access seeds – but had not registered. This unique part of the service model was worth
highlighting as ‘access’
Unique element Ideal
Register Farmers who registered at least once would still drive crucial service value
Usage
Repeat
An ideal farmer would be one that repeats registration for the product season-
on-season.

Non-Aware Aware Understand Access Register Repeat

Users have sent an Users have re-

Users have bought a SMS to the short registered a packet
Users have potential Users have become Users have become packet of seeds code and of seeds after payout
to access the service aware of the service aware of the service, giving them free successfully decision, either
but are not aware of and understand its access to the service registered for the within the same or
it value
insurance next season

! Different types of ‘user’ require different journeys: In the case of NextDrop, we defined two types of customer journeys for two
different user types of their mobile services. Different usage, data & categories were used to define each journey

14
 6
Using data analytics to identify bottlenecks
Use the customer journey framework to show where use Establish hypotheses around what causes each
is being obstructed bottleneck

• The customer journey framework (constructed at the previous stage) • With a priority bottleneck in mind, brainstorm hypotheses around
coupled with data analytics can be used to quantify the percentage of what could be causing customers to get stuck at this stage
the entire customer base falling in specific segments – e.g. % of
base not going past trial stage • For example, suppose ‘trial’ was the main bottleneck, is this due to
usability issues, understanding the service, or something else? These
• This will create a data-driven view of where users are ‘stuck on a hypotheses are critical to inform the qualitative briefs that deep
journey toward ideal use’ – use this view to address the biggest dive on customer’s behavioral and attitudinal drivers
bottlenecks to ‘ideal use’

Technical
Ideal failures
Usage Use XX% Lack of
?
Low
utility/value awareness
? ?

Trial XX%
Usability Poor
? Barrier @ Trial Understanding
?
Register XX%

15
 Understanding where usage falls off –
the home menu problem
The majority of the user base falls … upon further investigation with the data,
For HNI, the cursory segment was into the ‘cursory stage’, making it a ‘key issues’ are highlighted which can inform
natural place to consider making
‘stuck at home’ service changes…
priority actions

For HNI’s information service in Madagascar a customer journey

Segment / Stage* % base Key issues Priorities
included the following categories:
Close to one third of calls made by
repeat users do not pass the home
•Cursory – user calls the IVR menu, but never listens to a final Repeat
14.8% menu

Medium / Low
message
•Occasional – called <5 times & accessed at least one message Engaged 51.4% of calls made by repeat users
end up in a message (a figure close to
Improve IVR
tree design
•Repeat – called >=5 times & accessed at least one message occasional users) and content

45% of the user base are in the cursory segment and 84% of that Repeat
Non-
0.5% Only 46.4% of calls made by repeat
users end up in a message listened to
at over 75%
segment exit at the home menu. Why? Some early hypotheses might Engaged
be:

Home menu too

long
Voice is not
appealing
User dials wrong
number
Occasional 34.9% 65.5% of calls made by occasional Boost

Medium
users end up in a message repeat
Engaged behavior
Chances to listen to a message in full Improve IVR
End-user research can help confirm/refute hypotheses is less than half of repeat users tree design

Lesson:
Occasional
Non-
4.6% and content

Engaged
-Oftentimes the biggest bottleneck segment and the key issue within it Educate
83.8% of the calls exit at home users about
is identifiable from the analytics

High
the service
-Good segmentation and data analytics should make the priority issue
much clearer
Cursory
Non-
45.2% 19.5% of paying users are in this
segment (probably by mistake) Improve
IVR home
Engaged
menu

* note: ‘engaged’ users are those that have listened to over 75% a final message in the period
 Methodology notes: Use open source free analysis
tools until more advanced tools are required
The tools used for analytics will vary across organisations, but freely
available tools are powerful

Free & • Powerful and freely available tools to conduct a customer journey
powerful analysis on large data sets are available e.g., iPython, R
tools

• A strong community of practitioners for such tools exist, with a

Strong online multitude of tutorials & material online to support those willing to
communities invest in learning more advanced analytical toolsets

• All analysis conducted for examples in this toolkit used

We chose iPython, the advantage of this approach is that well-documented
iPython python code in ‘notebooks’ can be effectively passed between
individuals across teams/organistions

Each must • More advanced tools may be required later on, but investments in
choose their new team skillets/ software must be considered with a long term
approach view in mind*

Useful Links:
- Python basics (review Gilles)
- IPython intro (review Gilles)
- Review of analytical tools (ensure iPython, R included, preferably with others like Tableau)

17
*We assume that most organisations creating mobile products and services will have some kind of coding expertise in-house, making IPython a strong candidate tool since it should have cross-over with existing expertise
 Developing qualitative research briefs from 7
bottlenecks identified
Appreciate the limitations of data analytics, Deep dive on more granular journeys and
qualitative work can add critical insights behaviors/attitudes that have no digital trace

• Data analytics are good for seeing trends and Create a brief with a qualitative research team that investigates key
correlations, but not the causal factors that barriers and opportunities for overcoming the bottleneck identified
Limits of
underlie them through customer journey data analytics – different problems call for
data different techniques, so choose the method that fits best:
• Good at enabling segmentation of the entire
analytics
user base, but not at getting a rich picture of
typical individuals in different segments Community
Focus groups or
immersions/ Ethnography Depth interviews
workshops
intercepts
• Understand thought processes/motivations
Need for underpinning an individual’s relationship with the Flexible/ wide-angled/ in-context Structured topics/ solution focused
qualitative wider world – behavioural
research • Understanding how people’s behaviour and Flexible, Deep-dive into Shorter, more Shorter, more
respondent-led needs, attitudes and structured structured group
attitudes are situated in the wider world –
detailed behaviours individual sessions sessions
anthropological Broad community
of subject matter
/context focus Congruent sample

• Examples of techniques that can be applied:

Across a Valuable for
Valuable where Valuable where
range of little is known
Valuable for
research
focused solution
observing testing and
about subject objectives are
techniques Community Focus groups matter and
behaviour in
more focused,
structured idea
Depth action and generation, with
immersions/ Ethnography or community less exploratory
interviews understanding groups of people
intercepts workshops dynamics are yet depth is also
the role of context who share similar
desired desirable
characteristics

18
 Ethnography at the point of sale
5 
Re-purchase
1

The farmer re- Pre-purchase
What happens between the merchant and purchases a packet of The farmer hears about
seeds of same brand the relevant seeds and
farmer during the sale of a packet of seeds? after positive product the insurance service
experience
• Registration was identified through data
analytics as the key barrier to adoption for
2

What could
drive more ACRE Africa’s micro-insurance product, but
uptake? questions remained about what was driving the
low uptake Seed
Purchase
The farmer buys a
authenticity packet(s) of seeds from
• In response, a qualitative scope of work that focused on concern their local
A
barriers and opportunities to driving uptake of the service agrovet
qualitative
was created. In particular, it included examination of
brief
‘agrovets’ – local merchants who sell seeds and other
scoped No
farming inputs to small-holders incentive
3
Immersion at points
Ethnographic depths
Depths with

Register
of sale agrovets
3 x distribution points 6 x half-day The farmer registers by
farm visits 3 x extended depth sending SMS to a short
Observation & Intercept
interviews with agrovets code at time of
interviews


planting

Findings at the ‘purchase’ stage of a farmers’ seed journey (shown on

the right) showed that agrovets had no incentive to make farmers Replanting
A If there is no rain, the
aware of the product, and both farmers and agrovets were concerned farmer is paid in time
qualitative with seed authenticity which potentially confused or detracted from the with mobile money to
brief micro-insurance value proposition. The lack of incentives and the buy new seeds and
scoped replant
authenticity concern are barriers and opportunities to driving Barrier at Registration
greater product registration 4
19
 8
Analysing and presenting results
Present these as part of a larger framework of actionable
Pull together findings across all aspects of the project
recommendations

Keep commercial • When presenting results, the social and

and social objectives commercial objectives established at the project Social + Commercial Impact
outset should be of central focus – how do
top of mind
recommendations support social/commercial
goals?
• Customer journey analysis establishes barriers to 45.2%
Analytics
Use analytics to ideal usage, it should be used to quantify where
Cursory insights
clarify pain points customer engagement is lost – what are the most
significant bottlenecks?

Recommendations framework
• Recommendations must be linked to changes to
Target the business
the product or wider business model – what is the
model weakest areas of the business model to focus Customer Distributi
Key
Cost/Rev
Biz
activitie
on? Value
Proposit
Partners
on &
relationsh
s/ Key
resourc
structur
e
model
ion ips
es
analysis
• Qualitative research should provide an evidence
Provide qualitative base from which implementation actions are
evidence inspired – what are the key insights that support
recommendations?
Qualitative
• Actions proposed should have expected outcomes
Community
immersions/ Ethnography
Depth Focus groups Insights
interviews or workshops
intercepts

Measureable that can be linked to impact on social/commercial

outcomes for actions objectives stated – how can recommendations
be made accountable through re- Specific implementation ideas
measurement?

20
 Reviewing valvemen incentives to drive
active usage
Social + Commercial Impact
One key recommendation area for NextDrop
put valvemen in focus = How can NextDrop drive more registered end users? More end user
feedback? And more active valvemen usage?
NextDrop have to capture high quality data from valvemen
who enter information over the phone as they patrol the city,
manually opening and closing valves which control the water Recommendations framework
supply to citizens - how can NextDrop drive more quality Only 9%
Of Registered users
(active) valvemen usage of the mobile service? are Consistent
Users
Brand Registration Valvemen Accuracy Infrastructure
The earlier phase of the project determined that
Objective
NextDrop needed to drive more registered end users,
Active end user feedback, and active valvemen usage to drive Reframe brand & Create scalable Review valvemen Improve accuracy of Explore automation
customer value registration process incentives core product features in network
valvemen social & commercial impact Value
proposition Proposition

Recommendations were delivered as part of a full

Specific implementation ideas
package. Dealing with valvemen incentives was one
Recommendatio area of focus. Using data analytics a clear benchmark  Show them that it will eventually help them to save time
V1 Must fit with the
n Valvemen for improvement was established – 9% of the routine – i.e., not  Ensure they have sufficient data & battery to use and mitigate against
incentive registered base were exhibiting the ideal ‘consistent’ take too much time scenarios in which these are depleted (e.g. provide top-ups at beginning of
routine, limit data use, instruct on phone charging practices)
use SMARTPHONE
APP
V2 Must be practical –  Provide a protective hull with the smartphone to avoid it being damaged
Qualitative work showed that any mobile solution must How can NextDrop i.e. not risk damage
make changes to
 Conduct user testing in ‘mucky environments’ with valvemen
meet key valvemen usability/utility criteria. With the key during mucky work
Detailed Actions improve against
Across valvemen objective of driving more valvemen usage, valvemen
implementation ideas were sketched with the usability requirements? V3
Must be risk free –  Have phone heavily branded to reduce risk of stealing
usability / utility i.e. not expose them
criteria in mind and the benchmark of improving against to abuse/ phone
 Create casing to conceal phone in everyday use as much as possible
the 9% consistent user ratio. theft
 Direct users calling valvemen to NextDrop (e.g. an automatic message
Must help ‘deflect
V4 sent) to reduce harassment
customer
complaints– i.e.  Consider social incentives (e.g. valveman of the month highlighted on
reduce complaints utility website and by SMS to the residents of the area)
 9
Monitoring, implementing & evaluating
The approach is most effective when repeated, not Use the customer journey across each stage of an
just as one-off analysis ongoing cycle

The customer journey developed for the Use the customer journey
By framing recommendations to service providers service can impact all stages of the cycle from to monitor the service at
using the data-driven baseline of the customer constant monitoring, implementing and a more granular level
Measurement – measurement
approach journey there automatically is a future evaluating actions made on the service approach
measurement approach

Monitor
• Changes to the service can be conducted as MEASUREMENT APPROACH
controlled experiments with expected outcomes on
Controlled the quality of the user base, translated into
experiments indicators on the customer journey – e.g. Non-Aware Aware Understand Register Trial Use

expect a XX% reduction in users stuck at trial

stage (note: using ratios is better than aggregate Implement
Evaluate
numbers)

• With these indicators as ‘lines in the sand’

Brainstorm service
Create cyclic implementation changes can be evaluated by re- Evaluate the actions
actions and associated
analysing recent service data logs. By determining made and make
expected outcomes
evaluation refinements accordingly
– controlled
how effective changes have been against – closing the loop
expected outcomes new actions can be refined experiments

22
Following up with all 3 service provider examples
• HNI changed the way they monitored
service performance, measuring the number
From users
of engaged users listening to messages in
Monitor to
full as opposed to user numbers alone– a
engagement
much better fit with their stated success
Changing key monitoring metrics
KPIs
Enhancing• In addition, they used the Sankey
business representation of the customer
intelligence engagement journey (shown right) to
monitor the performance of the service over
time
Who How How How Who
Changes •to On review of the recommendations, ACRE will do do do do
help you you you you
Implement the Africa planned tests with multiple you do help
Customer
inter help
? it? ? act? ?
registration registration card designs to determine which What

Value
Proposition How
PART KEY VALU CUST CUST
card would drive the strongest registration uptake NER
do
ACT E
do
OM OM
Making business model S
you
IVITI
nee
PR
Distribution
you
ER
 relationships
reac
&
ERS
by communicating the right value ES
d?
OP REL
h
changes proposition What are the costs?
KEY
OSI
TIO
ATI
the
WhatONS
are the benefits?
m?
Incentivising RES
N HIP
COSTS REVENUE
OUR DIST
the
• Similar implementation experiments were CES RIB
UTI
distribution ON
planned for the distribution channel,
channel
incentivising agrovets with airtime to drive
farmer registrations of the product Received
Remove
Removing consent >5000
Evaluate •
unnecessary NextDrop responded to recommendations to Pieces of Solicited
barrier
consent & remove the consent barrier entirely, while feedback in a few
reframing reframing the messaging to share ‘water 1,312 weeks of making
Seeing uplift in end-user status updates’ to get users more involved Users changes
feedback (12% of base) in Reframe
solicited feedback message
•
To see strong They saw sizeable increase in the amount segment
user of feedback users gave in just a few weeks,
feedback providing a boost in one of the success 23
metrics identified for the service
 What other elements would you like us to add?
This toolkit outlines our 9-step approach 1 2 3 4 5
how would you like to see us build on it?
1 Establishing a plan of action
• This toolkit provides a holistic overview of our approach – in 9
succinct steps – in conducting customer journey evaluations 2 Mapping out the business model
9 step across a range of diverse mobile service types with existing
approach
mobile service data – all with a common theme of trying to 3 Setting Commercial/Social Objectives
achieve commercial and social impact in developing markets
4 Assessing available data sources

5 Constructing the Customer Journey

• We hope the approach is valuable to individuals across a wide
… which
range of organisations in the Mobile for Development space in 6
could be Using data analytics to identify bottlenecks
thinking about how to make better use of their data, but realise
added to
that some may want more detail, in which case please get in 7 Developing qualitative research briefs
touch
8 Analysing and presenting results

If there are any elements you would like to see added to this framework 9 Monitoring, implementing & evaluating
Please email [email protected]

24
 Read our evaluation case studies

ACRE Africa

NextDrop
Human Network
International

m4dimpact.com/analysis/case-studies

25
Read our Appendix for More Information

http://www.m4dimpact.com/analysis/insights

26
Supported by: Contacts
Mobile for Development Impact
[email protected]

27