US009202227B2

(12) United States Patent (10) Patent No.: US 9.202,227 B2

KahloW (45) Date of Patent: Dec. 1, 2015
(54) SALES PREDICTION SYSTEMS AND 7.620,655 B2 11/2009 Larsson et al.
METHODS 7,725,346 B2 5, 2010 Gruhl et al.
7.991,640 B2 8/2011 Shapira et al.
(71) Applicant: 6 Sense Insights, Inc., San Francisco, 8,234,148 B2 7/2012 Shapira et al.
8,244,714 B1 8, 2012 Collins
CA (US) 8.412,847 B2 4/2013 Longo
8,417,557 B2 4/2013 Shapira et al.
(72) Inventor: Amanda Kahlow, San Francisco, CA 2002/0169657 A1* 1 1/2002 Singh et al. ..................... 70.5/10
(US) 2007/0027741 A1 2/2007 Gruhl et al. ..................... 70.5/10

(73) Assignee: 6 Sense Insights, Inc., San Francisco, FOREIGN PATENT DOCUMENTS
CA (US) WO 2007 112411 A2 10, 2007
(*) Notice: Subject to any disclaimer, the term of this * cited by examiner
patent is extended or adjusted under 35
U.S.C. 154(b) by 172 days.
Primary Examiner — Sujay Koneru
(21) Appl. No.: 13/762,041 (74) Attorney, Agent, or Firm — Procopio Cory Har
(22) Filed: Feb. 7, 2013 greaves & Savitch LLP
(65) Prior Publication Data
(57) ABSTRACT
US 2013/0204663 A1 Aug. 8, 2013
A computer implemented sales prediction system collects
Related U.S. Application Data data relating to events of visitors showing an interest in a
(60) Provisional application No. 61/595,819, filed on Feb. client company from plural data sources, an organization
7, 2012. module which organizes collected data into different event
types and separates the collected event counts in each event
(51) Int. Cl. type between non-recent events and recent events occurring
G06O 10/00 (2012.01) within a predetermined time period, a first processing module
G06O 30/00 (2012.01) which periodically calculates weighting for each event type
G06O 30/02 (2012.01) based on recent events and non-recent events for the event
(52) U.S. Cl. type compared to totals for other selected event types, a
CPC .................................. G06O 30/0202 (2013.01) second processing module which periodically calculates
(58) Field of Classification Search sales prediction scores for each visitor and companies with
CPC .................................................. G06Q30/0202 which visitors are associated based on the accumulated event
USPC ................................................ 705/7.11 7.42 data and weighting, and a reporting and data extract module
See application file for complete search history. which is configured to detect variation in sales prediction
scores over time to identify spikes which can predict upcom
(56) References Cited ing sales and to provide predicted sales information and leads
U.S. PATENT DOCUMENTS to the client company.
7,085,682 B1 8, 2006 Heller et al.
7,349,827 B1 3/2008 Heller et al. 11 Claims, 9 Drawing Sheets

so

Identify data associated with an

'inson diffeet channes

Compute channel-Q scores

815& Accmulate the chaie-CR

scores (accumulated Q score)

Oetermine a spike from another

accumulated Q score to the
accimulated Q Score

- Provide a sale prediction

Sinc
U.S. Patent Dec. 1, 2015 Sheet 1 of 9 US 9.202,227 B2

ESP Architecture
14
Consuming Systems

Web analytic CMS

Anonymous Company Score
vistors Targeting
Search
engines
Mkt automation Companies Individual SCOres Mkt automation

Social Contact/individual Chat

| " Chat
contacts) attributes o

CRM

Ad network
Customers ompany attributes Ad networks

Personalization
CRM
One-to-One

IP Lookup service

IP address Domains Companies Firmagraphi

attributes
Geos

FIG. 1
U.S. Patent Dec. 1, 2015 Sheet 2 of 9 US 9.202,227 B2

Company Company Company

A B C

Data Receiving / Input Module

Data Management Module

Data Organization Module

Data Storage
Module

Event Type Weighting Processor Module

Sales Prediction Score Processing

Module

Report Generating Module

FIG. 2
U.S. Patent Dec. 1, 2015 Sheet 3 of 9 US 9.202,227 B2

Ve|S-Nou3.Á?edSnO3g)
OS3Á|dTu?
U.S. Patent Dec. 1, 2015 Sheet 4 of 9 US 9.202,227 B2

BSE|- -

V|SaeÁ3udnf?hoecaSuO3g)
•000 0’es •000 0’ws
duroºs dor 'zs dor ºts
U.S. Patent Dec. 1, 2015 Sheet 5 Of 9 US 9.202,227 B2
U.S. Patent Dec. 1, 2015 Sheet 6 of 9 US 9.202,227 B2

,
!

Eugºd~~

|
·

?*)×waeº:5$.
U.S. Patent Dec. 1, 2015 Sheet 7 Of 9 US 9.202,227 B2

saeuvonpqi Bunax?e
møeonuasd?oeuqndpves.oanluexos
U.S. Patent Dec. 1, 2015 Sheet 8 of 9 US 9.202,227 B2

800

805 ldentify data associated with an

entity, the data comes from
different channels

810 Compute channel-Q Scores

815 ACCumulate the Channel-O

scores (accumulated Q score)

820
Determine a spike from another
accumulated Q Score to the
accumulated Q Score

825 Provide a sale prediction

End

FIG. 8
U.S. Patent Dec. 1, 2015 Sheet 9 Of 9 US 9.202,227 B2

50

Processor
560

Secondary
Main Memory Memory 570
565
Internal Medium

( I/O Interface
) 575

585
605
Communication Network/External
Interface 590 U Medium 610
600

( ) Baseband
62O
Radio
615
Antenna
609

FIG. 9
 US 9,202,227 B2
1. 2
SALES PREDICTION SYSTEMIS AND events in a predetermined current time period, Such as the last
METHODS X days. Types of visitor activity detected prior to previous
sales (for the same and other similar companies) can be a
BACKGROUND powerful predictor of future sales.
In one embodiment, the system first uses collected data
1. Field of the Invention regarding visitor activity of one or more companies that have
The present invention relates generally to a system and purchased from the client company in the past to calculate a
method of predicting future sales, leads, and opportunities sales predictive model. The model includes the predictive
based on information gathered from multiple sources. value or weighting for each activity as a sales predictor based
2. Related Art 10 on prior sales to the known company or companies, and this
Analytic systems and lead scoring methods are known model is then used when calculating sales prediction scores
for other visitors and visitor companies to the client company.
which score levels of engagement, but these generally do not In other words, this system uses past purchase behavior of
predict sales. They also only use past purchase history of companies to predict when other companies might buy in the
known contacts only. They do not score companies levels of future
engagement or likelihood to purchase. All is done at the 15 In another aspect, a computer implemented method for
known contact level. In some cases, only limited Sources are predicting future sales to company visitors is provided, in
used for prediction purposes. For example, U.S. Pat. No. which data on company visitor activity is collected via stan
7,725,346 uses online public discussions to predict sales. dard means or visitor data from plural data sources or chan
They also do not score (or predict) what products or services nels, including sales data, online chat and Social media, com
companies will purchase. pany website search log files, web log files, website tags,
customer relationship management systems, and e-mail mar
SUMMARY keting data, and the collected data is organized into different
event types and the collected event counts are separated
Embodiments described herein provide for a sales predic between recent events and non-recent events, following ana
tion system and method which looks for various types of 25 lyzing the collected and organized data by calculating weight
interactions from multiple channels of data in order to predict ing for each event type as a sales predictor based on other
possible future sales. companies buying and behavior prior to buying, and recent
According to one embodiment, a computer implemented events and non-recent events for the event types compared to
sales prediction system is provided which comprises a non other event types, calculating sales prediction scores based on
transitory computer readable medium configured to store 30 the accumulated event data and weighting, and creating
computer executable programmed modules, a processor.com reports and extracts of data which track the variation in cal
municatively coupled with the non-transitory computer read culated sales prediction scores over time and identify spikes
able medium and configured to execute programmed mod in the sales prediction score for visitors as well as companies
ules comprising a data management module which collects with which the visitors are associated, which can then be used
data relating to activities of known or unknown visitors activi 35 for marketing purposes.
ties to one or more client companies from plural data sources, Spikes or increases in the calculated sales prediction scores
an organization module which organizes collected data into using this system and method have been found to be a statis
different event types and separates the collected event counts tically accurate predictor of upcoming sales increases, and the
in each event type between recent events and non-recent scores are also useful in identifying potential leads for mar
events, a first processing module which periodically system 40 keting purposes.
atically and automatically calculates weighting for each event Other features and advantages of the present invention will
type and each action or response taken based on recent events become more readily apparent to those of ordinary skill in the
and non-recent events for the event type compared to totals art after reviewing the following detailed description and
for other selected event types or actions or responses taken, a accompanying drawings.
second processing module which periodically calculates 45
sales prediction scores for each visitor and for each company BRIEF DESCRIPTION OF THE DRAWINGS
(all visitors from that company) based on the accumulated
event data and weighting, and a reporting module which is The details of the present invention, both as to its structure
configured to detect variation in sales prediction scores over and operation, may be gleaned in part by study of the accom
time to identify spikes (increases in scores compared to pre 50 panying drawings, in which like reference numerals refer to
vious periods of time) which can predict upcoming sales and like parts, and in which:
opportunities and to provide predicted sales and marketing FIG. 1 is a block diagram of one embodiment of a sales
information to each client company. prediction system;
The term “visitors' includes both known individuals or FIG. 2 is a block diagram illustrating the system architec
contacts and unknown individuals, and the visitors may be 55 ture in more detail;
independent individuals or may be associated with compa FIGS. 3 to 6 illustrate examples of sales prediction scores
nies which may be predicted to make purchases in the future, for several different potential customer companies (and the
based on activity levels. The actions of all “visitors' associ average score for all identified visitors of that company) of a
ated with a particular company are considered together to client company using the system of FIGS. 1 and 2 to predict
predict the likelihood of a future purchase by the associated 60 future sales, leads, and opportunities and obtain visitor based
company. The “visitor activity” includes actions of all types marketing information;
of visitors to the client company, its website, and identified FIG. 7 illustrates a method of triggering sales alerts based
interest following marketing activities such as targeted emails on spikes in sales prediction scores;
and the like. A visitor is identified in various ways, including FIG. 8 shows an example of a process implementation; and
IP look up services to identify the IP address for visitors over 65 FIG.9 is a block diagram illustrating an example processor
the Internet, followed by mapping of the IP address to based system that may be used in connection with various
domains and associated companies. “Recent events are embodiments described herein.
 US 9,202,227 B2
3 4
DETAILED DESCRIPTION 4. Who they are D&B company firmagraphic attributes
(vertical, geography, size of business, industry, annual rev
Certain embodiments as disclosed herein provide for a enue, etc.).
system and method for predicting future sales, leads and The collected data may include, but is not limited to, web
opportunities based on monitoring and analyzing data relat visiting data, call data, customer relationship management
ing to specific visitors and their companies in order to create (CRM), ad network, email, campaign, search, Social, market
sales prediction scores. ing automation, chat, etc. The detected contact entities or
After reading this description it will become apparent to visitors may be anonymous prospective customers or known
one skilled in the art how to implement the invention in prior customers or contacts. As illustrated in FIG. 1, the
various alternative embodiments and alternative applications. 10 system also uses an IP lookup service 12 Such as Demand
However, although various embodiments of the present Base to look up IP addresses, identify domain names and the
invention will be described herein, it is understood that these
associated companies, and to obtain and store firmagraphic
embodiments are presented by way of example only, and not attributes and geographical information on the detected new
limitation.
company, for later distribution to the particular client com
In the following description, the term "client company”
15 pany for which the data is being collected. The collected data
is processed in processing system or central processor 14
means a company using the system and method described according to the Software program instructions to produce
below to predict future sales of their products or services and various types of sales prediction reports and data extracts
to provide marketing insight, marketing leads, and sales which are transmitted to consuming systems 15 of the client
opportunities and sales leads, the term “visitor” means indi companies, such as CMS (customer management systems),
viduals (known individuals or contacts and unknown indi prospect targeting engines, marketing automation systems,
viduals) which make contact with the client company or chat, CRM, ad networks, and personalization one-to-one sys
engages in activity with the client companies’ digital chan temS.
nels or show interest in the client company in one or more of As illustrated in FIG. 2, the sales prediction system 10 may
multiple different ways and which may purchase from the 25 generate sales prediction information as illustrated in FIG. 1
client company in the future. Visitors may be independent for a plurality of different client companies 20, such as com
individuals or may be found to be associated with companies, pany A, company B, company C, etc., as illustrated in FIG. 2.
i.e. visitor companies, where the term “visitor company” Each client company 20 provides data files to the data input
means a company with which one or more visitors are asso module 11 of computer or processor 14. The files may include
ciated. The term "Indicator” or “visitor event means a 30 activity data, communication data, and/or other data. The
detected event or activity initiated by a visitor which relates to system may also be programmed to load third party client
the client companies products or services sold. Such events company specific data or data about the client companies
cover a large number of different possible event types, includ relative products or services over the Internet, for example
ing doing a search on a company website, doing a search on a from Social network sources and other sources.
search engine Such as Google, looking at identified pages on 35 The collected data files are stored in data storage module 22
the website, clicking on a company ad, opening a marketing and are also provided to data management module 24. In one
e-mail, filling out a company information form, mentioning embodiment, up to billions or more of rows of granular record
the company on on-line chat or Social networks, contacting level data and Supplementary lookup attribute data (i.e. other
the company on the company chat page or social network data Such as campaign and cost data) may be loaded from any
activity, attending a company trade show, seminar, presenta 40 one or more of the following data sources of the respective
tion or any detectable company related event, and may client companies, with separate storage and processing of
include both web data and offline interaction data provided in data for each of the client companies being handled:
client company data files, as discussed in more detail below. 1. Sales and bookings Data (such as SalesForce, Sibel)
FIGS. 1 and 2 are block diagrams illustrating one embodi 2. Web log files (raw server logs)
ment of a computer implemented sales prediction system 10 45 3. Website tags (such as SiteCatalyst, or Webtrends)
in which one or more processors are programmed with a 4. Registration data
non-transitory computer readable medium in order to carry 5. Customer Relationship management systems
out the steps described below. This system and method may 6. Marketing Automation systems—Lead data
be used to predict upcoming sales increases and also as a 7. Social media data
marketing tool to develop new customer prospects, contacts, 50 8. Search log files
and sales opportunities. 9. Online interactive Chat data (such as LivePerson)
System 10 collects multiple channels of visitor activity 10. Campaign spend data
data and/or communication data, Such as the types of data 11 11. Campaign lookup data
listed on the left hand side in FIG. 1, relating to a particular 12. Ad network data (such as Doubleclick, Atlas)
client company or multiple client companies, and then scores 55 13. IP to company lookup sources (such as Demandbase)
visitors of the client company detected in the activity data, The system is designed to handle terabytes of data, so there
including both individuals and companies, in order to deter is little or no limitation on the number of data sources which
mine the following for each detected individual visitor and for may be used or how much data can be loaded. The data is
each company visitor as a whole (taking into account all loaded into the data management module 24. In one embodi
individual visitor activities found to be associated with a 60 ment, this module uses a multi-channel analytic engine Such
company): as Adobe Insight or Hadoop, but other databases may be used
1. The likelihood to purchase. for this purpose in other embodiments.
2. How far along in the sales cycle the prospective customer The above data sources are connected via multiple keys.
(company or individual) is. Are they in the early stage of Keys include but are not limited to: web Cookies, contactids,
buying or late stages? 65 IP address, DUN's number, company name/address and coun
3.What products/services/technologies (goods sold) compa try code matching. All data is connected at every level where
nies are interested in buying. there is a possible connection. All data is stored at the most
 US 9,202,227 B2
5 6
granular level. Nothing is pre-aggregated. This allows the “visitor company' means a previously known or unknown
processing system to see and analyze indicators prior to buy company with which one or more visitors are associated.
ing in order to perform the analysis described below. Unlike other models, the sales predictions produced by this
The data is first organized and dimensionalized in data system are all based on relative increases (spikes) in scores
organization module 25. Standard extract, transform and load 5 over time per visitor company. The system produces a sales
(ETL) processes such as those provided in Adobe R. Insight, prediction score as described in more detail below. If a small
Hadoop or others may be used to used to create dimensions visitor company has say a low score of 5 one month, and then
out of the raw data. These dimensions are then sorted into goes up to a significantly higher score 50 in the following
action “Indicators' in event type weighting module 26. This is month, this indicates a 10x increase month over month. The
where the predictive model begins. Indicators are at two lev 10
system predicts that this company is going to purchase 10x
els, specifically category indicators and element level indica more than they have in the past. If their average booking was
previously S1000 a month, the system predicts they are going
tors. Thus, each individual element type has an individual to buy S10,000 (10x increase). If a large visitor company say
indicator, for example, viewing of a specific web marketing has an average monthly score of 100 then increases to 1000,
page on the client company website and is also assigned to a the system predicts they are going to buy 10x more than they
“category' indicator, which in this case is “viewed a web 15 have in the past. In this case, if their average booking was
marketing page', i.e. a count whenever any of the company’s S100,000 a month, the system predicts that they are going to
web marketing pages is viewed. Thus, each detected indicator buy $1,000,000 in future months.
or contact eventis assigned to the specific individual indicator The following is one example of a weighting technique
and category indicator in data storage. used to create an initial sales predictive model. Each category
There may be many category indicators (hundreds in and individual element or indicator is weighted based on
Some cases, less in others) for each channel of data. For overall counts by any dimensions in order to create the initial
example in the web channel, a few indicators would be: sales predictive model. Using the web source Page dimension
viewed a web marketing page, started a session on a web as an example, one conversion point might be a count of web
marketing page, watched a video, downloaded a white paper pages of a client company website visited by visitors prior to
orpdf document, clicked to the partner locator tool, etc. Email 25 a sale, as indicated in the example below.
category indicators include opening an email, responding to Example of Weighting Calculation for an Event
the email offer, etc. The processor also evaluates hundreds of Event=Pages Visited before a SALE (or other conver
thousands (millions) of element level indicators. Examples of Sion point can be used).
element level indicators are each individual page on a web In this case, the weighting for each page is P1/PT, where P1 is
site, each video on the site, each white paper, each email offer, 30
the number of times a page was visited before sale and PT is
each banner offer, each click event, etc. the total number of all website pages visited before the sale.
Each category indicator and individual element level indi This is calculated below for an example of two visitors to a
cator is assigned a weighting by weighting processor module website, where history indicates that several pages were vis
as described in more detail below. The system does not arbi ited, followed by a sale, i.e. a purchase by that visitor. The
trarily weight these indicators as is the case with some prior 35
visitors may be associated with a known contact, may be
art sales prediction scoring engines. Instead, weighting is associated with a known company or just scored as an
assigned based on linking stored visitor behavior to later sales unknown visitor not connected to a contact or company.
data in the clients CRM systems, making this system a more Visitor 1
accurate predictor of potential future sales. As noted above,
data is connected in the database by commonkeys. Sales data 40
for visitor companies such as business to business (B2B) PAGES: A-> B-> A-> C->E-> D-> SALE
companies is stored by company IDs, commonly a D&B
DUN's number (or company name/address/country code). A.
B
1
1
1
Anonymous visitor activity is connected by IP address to their C 1
companies using Demandbase (or other similar vendor) 45 D 1
lookup files, or to anonymous individual contacts. The system E 1
can therefore identify, for example, billions of interactions
from unknown visitors and derive an interaction behavior to
TOTAL PAGES before sale=6
later sales data in CRM systems. In other words, the system WEIGHT Calculation per PAGE:
looks for companies that have purchased in the past then 50
calculates the weighting based on what they did prior to
buying. This is then used to locate other companies showing
similar behavior patterns in order to predict future sales and
identify new leads and sales opportunities.
Statistical methods may be used in order to determine the 55
element level and category indicator weights for an initial
sales predicting model for each new client company. This
method looks at companies that have purchased from the
client company in the past, what actions did those companies Visitor 2
take prior to buying, and in what order and what % of their 60
PAGES:
overall actions? Actions that show a stronger relationship to A-sA-s-s-C-s-s-B-sD-s-s-A-s3-sC-sk-sSALE
buying are thereby weighted higher than those that do not TOTAL PAGES=11
show a relationship to a later purchase. When other new
visitors exhibit similar behavior, the system predicts that they WEIGHT Calculation per page visited:
too are getting ready to buy. The new visitors may be associ 65
ated with companies or may be independent individuals, i.e.
individuals for which no link is found to a company. The term
 US 9,202,227 B2
8
time, where the visit happened recently, is given a count of 2
for the recent visit. A sales prediction score Q is then calcu
lated in processing module 28 by multiplying each element
count by the calculated predictive weighting for that element,
and calculating the total score for all elements, as follows:
Sales Prediction Score Q=(A element count A
weight)+(B element count B weight)+(C element
K=1.11–0.09 count C weight)+(D element count D
weight)+(E element count E weight)+(Hele
TOTAL PAGE CALCULATION for 2 VISITORS ment count H weight)+(Kelement count'K
10 weight)+(A element recent count A weight)+(B
element recent count B weight)+(C element
recent count C weight)+(D element recent
count D weight)+(E element recent count E
weight)+(H element recent count H weight)+(K
element recent count Kweight).
15
In other words, Q=(CX--RX)*weight, where CX is the raw
count or total number of times the indicator has been hit, and
RX is the recent count. Thus, each recent count is counted
twice, one in the raw count, and once in the recent count.
Target scoring can be done for visitors who are individuals
and companies, to indicate the likelihood of purchasing any
K=0.09:1=0.09
client company product or a specific product, service or tech
nology. The following example is for a client company (ABC
Other examples of methods for performing the weighting Company) which has two product lines, specifically printers
include machine learning statistical methods such as but not 25
and tablet computers:
limited to decision trees, logistic regression, and linear ABC Company Scores:
regression. Q Score
Activities of any number of visitors may be used to deter (Likely to purchase): 155
mine weightings based on past sales. For example, weights Printer interest score: 33
may be produced from one sales transaction or two or more 30
Tablet interest score: 120
sales transactions (in which case, the weights for multiple Individual Y Scores:
sales transactions may be average weights). The weightings Likely to purchase score: 35
by event and by event category are then stored and used to Printer interest score: 0
calculate sales predicting scores for other visitors and visitor Tablet interest score: 400
companies based on activities similar to those of other visitors 35
In this case overall scores are calculated, along with Q
prior to a sale. The predictive model may be periodically scores for visitor or company activities associated with the
updated using new sales information and activities prior to specific product lines.
sales. Table 2 below is an example of an individual cookies
In some implementations, weights may be computed and/ calculation only at the category level (for simplicity of expla
or produced in associated with one or more products or ser 40
nation). The sales estimate processing system and method
vices. For example, generating weights from historic data carries out the calculations of weighting and sales prediction
associated with the one or more products or services. scores at both the category and element level. A lookup file is
In one embodiment, the system may be programmed to used to load the element weights back into date management/
double the value of an action/indicator count if the indicator data organization modules and the calculations are done for
was “recent event, where “recent’ means within a certain 45
rows of unique indicator elements (e.g., up to billions or more
current recent time window. The recency flag can be set to any rows). The final column is the Score or Q-(Non-recent
time window, for example within the last seven days, or a count--(Recent count2))*W (where W-weight), with total
shorter or longer period. In other words, a page visited one Q=504.3.
TABLE 2
Recency: Did the visitor
Non-Recent show interest in the last
Weights Normalized Activity (7) days?
Reach: Campaign Ad 1.O O.S9/o 1 1 3
Reach: SearchEngineToWM 1.O O.S9/o O
Reach: SocialMediaToWM 1.O O.S9/o O
Reach: WMCiscoSearch 2.0 1.0% O
Reach: WMLandingPage 1.O O.S9/o O
Usage:WMPages 1.O O.S9/o 5 3 11
Usage: AdWMPages O.1 O.1% 5 3 1
Usage: NonWMPages O.2 O.1% 5 3 2
Usage: Logins O.O O.0% 5 3 O
Usage: MobilePages 1.O O.S9/o 5 3 11
Content: Complete Video S.O 2.6% 5 3 55
Content: VideoPlay S.O 2.6% 5 3 55
Content: WMEndOf Tree S.O 2.6% 5 3 55
Content: CPRRegister S.O 2.6% O
Content: DiscussionForums S.O 2.6% O
Content: GistComplete 2O.O 10.5% 1 1 60
 US 9,202,227 B2
9 10
TABLE 2-continued
Recency: Did the visitor
Non-Recent show interest in the last
Weights Normalized Activity (7) days?
Content: Keycode 7.0 3.7% 1 1 21
Content: MyCisco S.O 2.6% O
Content: News 1.O O.S90 O
Content: OreoClick 1O.O S.2% O
Content: Podcastt S.O 2.6% O
Content: RSSReadClicks 7.0 3.7% O
Content: RSSSubscribers S.O 2.6% O
Content: SocialSharing 7.0 3.7% O
Content: ThreeDModel 6.O 3.1% O
Connection: BuyNow 2O.O 10.5% O
Connection: Chat 1S.O 7.8% 2 2 90
Connection: Email 1O.O S.2% 2 2 60
Connection: PartnerLocator 2O.O 10.5% 1 40
Connection: RequestPrice 2O.O 10.5% 1 40

Total 100.0% 47 33 SO4.3

The resultant sales prediction scores are stored for indi company, calculated as explained above. The visitor company
viduals (cookies, contactids) and by visitor company (name, scores are based on events initiated by one or more “visitors'
DUNS number), so that sales personnel can follow up on to the client company which are associated with the visitor
promising sales leads and marketing tools and engines can 25
company. The vertical bars represent the monetary amount of
target that visitor (individual or company) with a targeted purchases or bookings made by the company, while calcu
message. A total sales prediction score for a particular visitor lated scores over time are shown as data points and are con
(individual or company) can be tracked over time using the nected by the score line. It has been found that the score line
above calculation, with periodical updates of counts, as well spikes up prior to a booking, with the size of the spike pre
as updates of weighting based on counts in categories or 30 dicting the relative future booking amount or dollar value.
individual contact events. Updates are not connected to book The scores may be calculated at different time intervals
ing but are performed periodically regardless of sales or depending on the type of business in which the company is
bookings. involved. In the illustrated examples, the client companies for
A similar methodology is used to determine what product which contacts are being tracked are B2B companies with
category and what individual product an individual or com 35 6-12 month sales cycles, so the scores are calculated monthly,
pany is interested in and is likely to buy. The above method is and the expectation is that a score spike might be expected
followed to calculate scores, however indicators are only used 3-12 months prior to the booking. A spike in the calculated
if they map to a particular product or product category. For score is a relatively good predictor of a booking to come,
example, if scores are being scored for say Laptops for an IT while the size of the spike provides an indication of the size of
company, the score is based only on pages and videos and 40 the future booking, as seen by the arrowed lines between a
events that relate to the product topic. spike and a subsequent booking in FIGS. 3 to 6.
The accumulated estimated sales prediction scores for In FIGS. 3 to 6, each of the visitor companies being scored
companies and individuals showing interest in a client com is a prospective customer of the client company, say visitor
pany over time are stored in database 22 and then converted companies 1, 2, 3 and 4 which are seen to contact client
into various reports and data extracts by report generating 45 company A. For example, for visitor Company 1, the system
module 30. Reports and data extracts are communicated to looks at a selected number of people from visitor Company 1,
various downstream systems of the respective client compa using cookies or unknown people mapped to the IP address of
nies 20, as indicated in FIGS. 1 and 2. Scores may be visitor Company 1, and as well as known contacts at visitor
extracted at the individual/contact level or at the company Company 1, and tracks events over time as described above,
level. Extracts are sent to the client company’s content man 50 using the previously generated weighting model as well as the
agement and web or demand generation personalization sys accumulated event data in order to calculate Successive sales
tems, marketing automation systems, CRM platforms, sent prediction scores. The same process is carried out for visitor
directly as alerts to sales representatives, and so on. Alerts Company 2, visitor Company 3, etc. The scores in FIGS. 3 to
may be sent in the event of a detected spike in scores for a 6 are the combined or aggregated scores for all tracked visi
particular visitor, product line, or product, for appropriate 55 tors/contacts from the respective visitor Company 1, 2, 3 and
follow up by company personnel. The downstream use cases 4. Spikes in the sales predicting scores are communicated to
for sales and marketing are numerous. The foregoing method the client companies.
is carried out periodically for each client company to provide As illustrated in FIG. 3 for visitor Company 1, which is a
them with sales prediction scores. Small business with 100 employees, sales spikes of purchases
FIGS. 3 to 6 illustrate some graphical examples of report 60 of products or services from client company A occurred a few
ing the tracked contact scores for Small, mid-size, and large months after spikes in the calculated scores, as seen by the
visitor companies interacting with a client company. The arrows between the spikes and Subsequent purchases. The
client companies these are built for are all B2B companies same sales prediction can be seen for visitor Company 2 (a
with an average of 6 to 12 month sales cycles. The X axis in small business with 218 employees) in FIG. 4, visitor Com
FIGS. 3 to 6 represents time by month, while the left hand 65 pany 3 (a mid-size company with 1350 employees) in FIG. 5,
Vertical axis represents total sales and the right hand vertical and visitor Company 4 (a large enterprise with 160,000
axis is the calculated sales prediction score or Q for the visitor employees) in FIG. 6. The largest spike in FIG. 3 was fol
 US 9,202,227 B2
11 12
lowed by a booking 17 times larger than the previous average tistical relationship to later buying, and this model is used to
booking for this visitor company. predict sales and provide new leads and sales opportunity
The sales prediction value of this system and method has information to a client company based on other individuals or
been shown by using the system to map historical activity to companies exhibiting the same pattern of behavior.
sales or bookings over a two year period, tracking a very large Thus, when a new visitor or visitor company is found to
number of companies. In this period, the system was set up to demonstrate the same pattern of behavior, the model can
monitor for booking or sales spikes in the nine months fol predict they too are going to buy. The system looks at past
lowing a scoring spike indicating increased contact activity. interaction history, and if increasing by X% in their activity,
With the score tracking graphs created as described above and the system predicts a purchase of X times higher than their
illustrated in FIGS. 3 to 6, there are four possible outcomes: 10 past purchase history. For several client companies, this sys
1. No scoring spike—no booking spike. tem has been statistically validated to predict future purchases
2. Scoring spike-no booking spike following scoring spike. with significant accuracy.
3. Scoring spike booking spike following scoring spike. This model connects data from not just one source at a time
4. No scoring spike booking spike. but multiple sources of interactive data at the record/visitor
Statistical analysis showed that, for a company with a 15 level/company level. The system does not pre-aggregate any
minimum score of 50 and a spike of 10x or greater than the data. The predictions are based on interactions and provide an
average score spike, 70% booked 5x or greater than average advance prediction of when a customer is going to purchase,
booking size. Of this group, 38% booked 20x or greater their how much is purchased, and what is purchased. Models can
average booking size. 27% booked at 50x or greater their be built up over time to predict purchases on a customer
average booking size. specific basis, not compared to every prospective customer
This system and method can determine a company or indi but based on that individual (or companies) past interactions.
vidual’s level of interest in a client company, what particular The predictive models are continuously updated based on
products or services they are interested in, the likelihood of newly detected scores and Subsequent sales data for each
purchase and when a purchase might occur, as well as how detected visitor.
much might be purchased based on past purchase history. It 25 There are many “propensity to buy”, “wallet share' etc
also allows profiling of a client company's customers and models built by most corporate database marketing compa
prospective customers. This allows a client company using nies. These models use past purchase history, and are not
the sales prediction system described above to do better tar based on current customer interactions, unlike the systems
geted marketing, for example based on sales prediction score and method described above.
triggers, personalized marketing based on detected product 30 When the system identifies purchase intent, product inter
interests, and allows generation of leads and telemarketing est is identified, and the information can then be sent to the
lists of companies and individuals which have shown an client company for which the information was identified, and
interest in the client company’s products or services. FIG. 7 then used for a plethora of downstream sales and marketing
illustrates one technique for triggering sales alerts. As illus activities, as illustrated in FIGS. 1 and 7. These activities may
trated in the graph in the lower left hand corner of FIG. 7, an 35 include, for example:
activity spike or score spike indicates a potential new oppor Personalizing the web experience—send to the company's
tunity for the client company, and a decision engine detects CRM or other personalization tools:
Such activity spikes and triggers an alert, which may include Generate new leads/opportunities—send into CRM, and
sending a personalized list of highly engaged contacts and Marketing automation systems (salesforce)
companies to the appropriate next step sales contact. Such as 40 Generate targeted telemarketing lists—the system gener
outbound telemarketing, sales assistant managers, and com ates lists which may be exported at the contact id level
pany partners. the generated product interest results may also be used as a
This system can also be used to measure marketing effec universal benchmark measure of Success for all marketing
tiveness, Such as the effectiveness of marketing mix, market efforts (channel attribution models, campaign effectiveness,
ing attribution, campaigns of various types, website content, 45 content, messaging etc)
tools, features, innovations, and the like. the sales prediction scores generated as described below
Companies already do “lead scoring, but only after some are also used to forecast company business, for example what
one is a known contact. The system and method described products, Verticals, geos, market segments are going up or
above identifies contacts before they are known to the client down
company, and creates individual and company scores based 50 The sales prediction system and method described above
on anonymous activity. In order to do this, the system maps IP uses statistical methods to determine what companies have
addresses to companies using a lookup file of the type pro purchased in the past and what actions they took prior to
vided by vendors such as Demandbase, Inc., as described buying as a percentage of their overall actions. Actions having
above. a stronger relationship to buying are weighted higher. Similar
This system and method does not just produce scores based 55 behavior patterns of other companies can then be used to
on detected engagement, but instead builds predictive models predict future purchases by those companies with the same
based on records of activity. Engagement scoring could mean accuracy.
that a company is just looking for tech Support, or something FIG. 8 shows an example of a process implementation.
unrelated to buying. Most prior art scoring models use arbi Process 800 may start with block 805 where data produced
trary weightings. They say they “think” X action should be 60 within a time period and associated with an entity (e.g., an
scored with a given weight factor. The model created by the individual or company) is identified. The data may be col
system and method described above specifically looks for the lected from different sources or channels. Each collection or
types of interactions that predict a sale. In order to build the occurrence of an activity (e.g., a webpage visit, an email
weightings, the system takes data from customers that have message, a text message, etc.) may be referred to as a “data
purchased in the past and looks for what they do prior to 65 page. At block 810, a channel-Q score or page-Q(p) score is
buying. A predictive statistical model and weights are then computed for each of the channels of data. The channel-Q
created around the interactions that are found to have a sta score may be an aggregate (e.g., an average) of the individual
 US 9,202,227 B2
13 14
Q Scores associated with the data pages of the channel (page general purpose processor, a digital signal processor
Q(p) score) using a formula as described above, as follows: (“DSP”), an ASIC, FPGA or other programmable logic
page-g(p) score number of occurrences of data page device, discrete gate or transistor logic, discrete hardware
(p)xpage weight (w) components, or any combination thereof designed to perform
In some implementations, a recently occurred data page the functions described herein. A general-purpose processor
(detected visitor activity) within a certain time period prior to can be hardware such as a microprocessor, or any processor,
the computation of the channel-Q score may be counted more controller, or microcontroller, and the processing can be per
(e.g., twice or three times, etc.), as indicated in Table 2 above, formed on a single piece of hardware or distributed across
using the following formula for counting each recent occur 10
multiple servers or running on multiple computers that are
rence twice, for example: housed in a local area or dispersed across different geographic
page-g(p) score=total number of occurrences of data
locations. A processor can also be implemented as a combi
page(p)+number of recent occurrences of date nation of computing devices, for example, a combination of a
page (p)xpage weight (w) DSP and a microprocessor, a plurality of microprocessors,
In this way, recent occurrences are included both in the total 15 one or more microprocessors in conjunction with a DSP core,
number and the number of recent occurrences, and are thus or any other Such configuration.
counted twice. The recent occurrences may be multiplied by FIG. 9 is a block diagram illustrating an example of a
2, for example, if they are to be given even more weight in the computer system 550 that may be used in connection with
calculation. various embodiments described herein. For example, the
At block 815, the channel-Q scores for two or more of the computer system 550 may be used as the microprocessor or
channels of data may be computed to produce, for example, central processing unit 14 which receives the activity data and
an accumulated Q score. At block 820, the accumulated Q which uses that data and the created models for contacts to
score may be analyzed with one or more previously computed generate output sales prediction reports for client company
accumulated Q Scores (e.g., accumulated Q scores computed customers or prospective customers, as well as marketing
for different time periods) to determine a movement or trend, 25 alerts or the like (consumption data), as described above.
Such as a spike. If a spike occurs (e.g., spike above a certain However, other computer systems and/or architectures may
threshold), the spike may be provided and/or reported as a be used, as will be clear to those skilled in the art.
sale prediction. The sale prediction may be sent to one or The computer system 550 preferably includes one or more
more recipients in alerts (e.g., alert email messages, text processors, such as processor 560. Additional processors may
messages, and/or Voice messages). 30
be provided, such as an auxiliary processor to manage input/
In some examples, process 800 may be implemented with output, an auxiliary processor to perform floating point math
different, fewer, or more blocks. Process 800 may be imple ematical operations, a special-purpose microprocessor hav
mented as computer executable instructions, which can be ing an architecture Suitable for fast execution of signal
stored on a medium, loaded onto one or more processors of processing algorithms (e.g., digital signal processor), a slave
one or more computing devices, and executed as a computer 35
processor Subordinate to the main processing system (e.g.,
implemented method.
Process 800 is one example of a method of weighting back-end processor), an additional microprocessor or con
accumulated data on visitor activities related to client com troller for dual or multiple processor systems, or a coproces
panies in order to predict possible future sales and opportu sor. Such auxiliary processors may be discrete processors or
nities. Other methods of weighting may alternatively be used 40 may be integrated with the processor 560.
in other embodiments to perform the weighting and predic The processor 560 is preferably connected to a communi
tion, for example machine learning statistical methods such cation bus 555. The communication bus 555 may include a
as but not limited to decision trees, logistic regression, and data channel for facilitating information transfer between
linear regression. storage and other peripheral components of the computer
Those of skill in the art will appreciate that the various 45 system 550. The communication bus 555 further may provide
illustrative logical blocks, modules, circuits, and method a set of signals used for communication with the processor
steps described in connection with the above described fig 560, including a data bus, address bus, and control bus (not
ures and the embodiments disclosed herein can often be shown). The communication bus 555 may comprise any stan
implemented as electronic hardware, computer Software, or dard or non-standard bus architecture such as, for example,
combinations of both. To clearly illustrate this interchange 50 bus architectures compliant with industry standard architec
ability of hardware and software, various illustrative compo ture (“ISA), extended industry standard architecture
nents, blocks, modules, circuits, and steps have been (“EISA), Micro Channel Architecture (“MCA'), peripheral
described above generally in terms of their functionality. component interconnect (“PCI) local bus, or standards pro
Whether such functionality is implemented as hardware or mulgated by the Institute of Electrical and Electronics Engi
Software depends upon the particular application and design 55 neers (“IEEE) including IEEE 488 general-purpose inter
constraints imposed on the overall system. Skilled persons face bus (“GPIB), IEEE 696/S-100, and the like.
can implement the described functionality in varying ways Computer system 550 preferably includes a main memory
for each particular application, but such implementation deci 565 and may also include a secondary memory 570. The main
sions should not be interpreted as causing a departure from memory 565 provides storage of instructions and data in a
the scope of the invention. In addition, the grouping of func 60 computer readable medium for programs executing on the
tions within a module, block, circuit or step is for ease of processor 560, such as the program modules described above
description. Specific functions or steps can be moved from in connection with FIGS. 1 and 2. The main memory 565 is
one module, block or circuit to another without departing typically semiconductor-based memory Such as dynamic ran
from the invention. dom access memory (“DRAM) and/or static random access
Moreover, the various illustrative logical blocks, modules, 65 memory (“SRAM). Other semiconductor-based memory
and methods described in connection with the embodiments types include, for example, synchronous dynamic random
disclosed herein can be implemented or performed with a access memory (“SDRAM), Rambus dynamic random
 US 9,202,227 B2
15 16
access memory (“RDRAM), ferroelectric random access tional phone line, cellular phone link, wireless data commu
memory ("FRAM), and the like, including read only nication link, radio frequency (“RF) link, or infrared link,
memory (“ROM). just to name a few.
The secondary memory 570 may optionally include a hard Computer executable code (i.e., computer programs or
disk drive 575 and/or a removable storage drive, for example software) is stored in the main memory 565 and/or the sec
a floppy disk drive, a magnetic tape drive, a compact disc ondary memory 570. Computer programs can also be
(“CD) drive, a digital versatile disc (“DVD) drive, etc. The received via communication interface 590 and stored in the
removable storage drive reads from and/or writes to a remov main memory 565 and/or the secondary memory 570. Such
able storage medium in a well-known manner. The removable 10
computer programs, when executed, enable the computer
storage medium may be, for example, a floppy disk, magnetic system 550 to perform the various functions of the embodi
tape, CD, DVD, etc. ments described above.
The removable storage medium is a non-transitory com In this description, the term “computer readable medium’
puter readable medium having stored thereon computer is used to refer to any non-transitory computer readable stor
executable code (i.e., software) and/or data. The computer 15 age media used to provide computer executable code (e.g.,
Software or data stored on the removable storage medium is Software and computer programs) to the computer system
read into the computer system 550 as electrical communica 550. Examples of these media include main memory 565,
tion signals for execution by the processor. secondary memory 570 (including hard disk drive 575,
In alternative embodiments, secondary memory 570 may removable storage medium, and external storage medium
include other similar means for allowing computer programs 610), and any peripheral device communicatively coupled
or other data or instructions to be loaded into the computer with communication interface 590 (including a network
system 550. Such means may include, for example, an exter information server or other network device). These non-tran
nal storage medium 610 and an interface 600. Examples of sitory computer readable mediums are means for providing
external storage medium 610 may include an external hard executable code, programming instructions, and Software to
disk drive or an external optical drive, or and external mag 25 the computer system 550.
neto-optical drive. In an embodiment that is implemented using software, the
Other examples of secondary memory 570 may include Software may be stored on a computer readable medium and
semiconductor-based memory Such as programmable read loaded into computer system 550 by way of removable stor
only memory (“PROM), erasable programmable read-only 30
age drive 610, interface 585, or communication interface 590.
memory (“EPROM), electrically erasable read-only In Such an embodiment, the Software is loaded into the com
memory (“EEPROM), or flash memory (block oriented puter system 550 in the form of electrical communication
memory similar to EEPROM). Also included are any other signals 605. The software, when executed by the processor
removable storage units and interfaces, which allow software 560, preferably causes the processor 560 to perform the fea
and data to be transferred from the removable storage unit to 35 tures and functions previously described herein.
the computer system 550. Various embodiments may also be implemented primarily
Computer system 550 may also include a communication in hardware using, for example, components such as applica
interface 590. The communication interface 590 allows soft tion specific integrated circuits (ASICs'), or field program
ware and data to be transferred between computer system 550 mable gate arrays (“FPGAs). Implementation of a hardware
and external devices (e.g. printers), networks, or information 40 state machine capable of performing the functions described
Sources. For example, computer Software or executable code herein will also be apparent to those skilled in the relevant art.
may be transferred to computer system 550 from a network Various embodiments may also be implemented using a com
server via communication interface 590. Examples of com bination of both hardware and software.
munication interface 590 include a modem, a network inter The system 550 also includes optional wireless communi
face card (“NIC), a communications port, a PCMCIA slot 45 cation components that facilitate wireless communication
and card, an infrared interface, and an IEEE 1394 fire-wire, over a Voice and over a data network. The wireless commu
just to name a few. Communication interface may also com nication components may comprise an antenna system 620, a
prise Bluetooth or other wireless communication interface 34 radio system 615 and/or a baseband system 610. In the com
of FIG. 3. munication device 550, radio frequency (“RF) signals, if
Communication interface 590 preferably implements 50 generated, are transmitted and received over the air by the
industry promulgated protocol standards, such as Ethernet antenna system 620 under the management of the radio sys
IEEE 802 standards, Fiber Channel, digital subscriber line tem 615.
(“DSL), asynchronous digital subscriber line (ADSL). In one embodiment, the antenna system 620, if present,
frame relay, asynchronous transfer mode (ATM), inte may comprise one or more antennae and one or more multi
grated digital services network (“ISDN), personal commu 55 plexors (not shown) that perform a Switching function to
nications services (“PCS”), transmission control protocol/ provide the antenna system 620 with transmit and receive
Internet protocol (“TCP/IP), serial line Internet protocol/ signal paths. In the receive path, received RF signals can be
point to point protocol (“SLIP/PPP), and so on, but may also coupled from a multiplexor to a low noise amplifier (not
implement customized or non-standard interface protocols as shown) that amplifies the received RF signal and sends the
well. 60 amplified signal to the radio system 615.
Software and data transferred via communication interface In alternative embodiments, the radio system 615 may
590 are generally in the form of electrical communication comprise one or more radios that are configured to commu
signals 605. These signals are preferably provided to com nication over various frequencies. In one embodiment, the
munication interface 590 via a communication channel 600. radio system 615 may combine a demodulator (not shown)
Communication channel 600 carries signals 605 and can be 65 and modulator (not shown) in one integrated circuit (“IC).
implemented using a variety of wired or wireless communi The demodulator and modulator can also be separate compo
cation means including wire or cable, fiber optics, conven nents. In the incoming path, the demodulator Strips away the
 US 9,202,227 B2
17 18
RF carrier signal leaving a baseband receive audio signal, historical sales data, of a number of occurrences of the
which is sent from the radio system 615 to the baseband data page(p) over a number of occurrences of one or
system 610. more data pages associated with one or more previous
The baseband system 610 is also communicatively coupled sales associated with the at least one product or service,
with the processor 560. The central processing unit 560 has where the one or more data pages include the data page
access to data storage areas 565 and 570. The central process (p);
ing unit 560 is preferably configured to execute instructions accumulating the channel-Q score for each of the channels
(i.e., computer programs or software) that can be stored in the of data (accumulated Q Score);
memory 565 or the secondary memory 570. Computer pro storing the accumulated Q Score;
grams can also be received from the baseband processor 610 10 comparing the accumulated Q Score to a prior accumulated
and stored in the data storage area 565 or in secondary Q Score computed for a previous time period to the
memory, or executed upon receipt. Such computer programs, accumulated Q Score;
when executed, enable the computer system 550 to perform determining if the accumulated Q Score produces a spike
the various functions of the present invention as previously from the prior accumulated Q Score;
described. For example, data storage areas 565 may include 15 determining if the spike is above a selected threshold; and
various software modules (not shown) that perform the vari sending a sale prediction alert to at least one recipient on
ous functions of the present invention as previously detection of a spike above the selected threshold, the
described. alert including Stored contact information for visitors
The above description of the disclosed embodiments is identified in connection with the accumulated Q score.
provided to enable any person skilled in the art to make or use 2. The method of claim 1, wherein the entity comprises one
the invention. Various modifications to these embodiments or more names associated with a company.
will be readily apparent to those skilled in the art, and the 3. The method of claim 1, wherein weight (w) is recom
generic principles described herein can be applied to other puted periodically.
embodiments without departing from the spirit or scope of the 4. The method of claim 1, wherein the number of occur
invention. Thus, it is to be understood that the description and 25 rences of data page(p) in the formula include double counting
drawings presented herein represent a presently preferred the data page(p) for recent occurrences of the data page(p)
embodiment of the invention and are therefore representative within a predetermined time period prior to the computation
of the subject matter which is broadly contemplated by the of the page-Q(p) score, whereby the number of occurrences
present invention. It is further understood that the scope of the of data page(p) used in the formula is equal to the total
present invention fully encompasses other embodiments that 30 number of occurrences of date page (p) plus n times the
may become obvious to those skilled in the art. number of recent occurrences of data page(p), where n is one
What is claimed is: O Oe.
1. A computer-implemented method, wherein one or more 5. A non-transitory computer readable medium having
computing devices comprising storage and a processor are stored therein computer executable instructions for:
programmed to perform steps comprising: 35 storing historical sales data associated with at least one
storing historical sales data associated with at least one product or service of an entity;
product or service of an entity; identifying and storing data produced within a time period
identifying and storing data produced within a time period and associated with the at least one product or service,
and associated with the at least one product or service, the data comprising visitor activities in connection with
the data comprising visitor activities in connection with 40 plural data sources, the data sources comprising at least
plural data sources, the data sources comprising at least website page look up data on at least one website asso
website page look up data on at least one website asso ciated with the entity, and at least one of Social media
ciated with the entity, and at least one of Social media data, e-mail data, search log data, and advertisement
data, e-mail data, search log data, and advertisement look up data, and the visitors being known or unknown
look up data, and the visitors being known or unknown 45 to the entity:
to the entity: the data being from a plurality of channels, the data stored
the data being from a plurality of channels, the data stored on respective channels corresponding to respective dif
on respective channels corresponding to respective dif ferent visitor activities, and data from each of the chan
ferent visitor activities, and data from each of the chan nels comprises at least one data page(p), where p=1:
nels comprises at least one data page(p), where p=1: 50 obtaining and storing contact information for visitors asso
obtaining and storing contact information for visitors asso ciated with identified visitor activities corresponding to
ciated with identified visitor activities corresponding to the respective channels, wherein obtaining the contact
the respective channels, wherein obtaining the contact information comprises, for at least one or more of the
information comprises, for at least one or more of the visitors, mapping an Internet Protocol (IP) address of the
visitors, mapping an Internet Protocol (IP) address of the 55 visitor to contact information for the visitor using a
visitor to contact information for the visitor using a lookup file;
lookup file; computing one channel-Q score for each of the channels of
computing one channel-Q score for each of the channels of data, wherein the channel-Q score is an aggregate of at
data, wherein the channel-Q score is an aggregate of at least one page-Q(p) score for the at least one data page
least one page-Q(p) score for the at least one data page 60 (p), wherein the at least one page-Q(p) score is com
(p), wherein the score is computed using formula: puted using formula:
page-g(p) score number of occurrences of data page page-g(p) score number of occurrences of data page
(p)xpage weight (w) (p)xpage weight (w)
where the number of occurrences is a number of times 65 where the number of occurrences is a number of times the
the data page(p) occurs within the current time period, data page(p) occurs within the current time period, and
and the weight (w) represents a ratio, based on stored the weight (w) represents a ratio, based on stored his
 US 9,202,227 B2
19 20
torical data, of a number of occurrences of the data the respective channels, wherein obtaining the contact
page(p) over a number of occurrences of one or more information comprises, for at least one or more of the
data pages associated with one or more previous sales visitors, mapping an Internet Protocol (IP) address of the
associated with the at least one product or service, where visitor to contact information for the visitor using a
the one or more data pages include the data page(p); lookup file;
accumulating the channel-Q score for each of the channels computing one channel-Q score for each of the channels of
of data (accumulated Q Score); data, wherein the channel-Q score is an aggregate of at
storing the accumulated Q Score; least one page-Q(p) score for the at least one data page
comparing the accumulated Q Score to a prior accumulated (p), wherein the at least one page-Q(p) score is com
Q score computed for a previous time period to the 10 puted using formula:
accumulated Q Score; page-g(p) score number of occurrences of data page
determining if the accumulated Q Score produces a spike (p)xpage weight (w)
from the prior accumulated Q Score; where the number of occurrences is a number of times the
determining if the spike is above a selected threshold; and data page(p) occurs within the current time period, and
sending a sale prediction alert to at least one recipient on 15
detection of a spike above the selected threshold, the the weight (w) represents a ratio, based on stored his
alert including Stored contact information for visitors torical data, of a number of occurrences of the data
identified in connection with the accumulated Q score. page(p) over a number of occurrences of one or more
6. The computer readable medium of claim 5, wherein the data pages associated with one or more previous sales
entity comprises one or more names associated with a com associated with the at least one product or service, where
pany. the one or more data pages include the data page(p);
7. The computer readable medium of claim 5, wherein page accumulating the channel-Q score for each of the channels
weight (w) is recomputed from time to time. of data (accumulated Q Score);
8. At least one computing device comprising storage and a storing the accumulated Q Score;
processor configured to perform: 25 comparing the accumulated Q Score to a prior accumulated
storing historical sales data associated with at least one Q Score computed for a previous time period to the
product or service of an entity; accumulated Q Score;
identifying and storing data produced within a time period determining if the accumulated Q Score produces a spike
and associated with the at least one product or service, from the prior accumulated Q Score;
the data comprising visitor activities in connection with 30 determining if the spike is above a selected threshold; and
plural data sources, the data sources comprising at least sending a sale prediction alert to at least one recipient on
website page look up data on at least one website asso detection of a spike above the selected threshold, the
ciated with the entity, and at least one of Social media alert including Stored contact information for visitors
data, e-mail data, search log data, and advertisement identified in connection with the accumulated Q score.
look up data, and the visitors being known or unknown 35 9. The at least one computing device of claim 8, wherein
to the entity: the entity comprises one or more names associated with a
company.
the data being from a plurality of channels, the data stored 10. The at least one computing device of claim 8, wherein
on respective channels corresponding to respective dif
ferent visitor activities, and data from each of the chan the page weight (w) is recomputed from time to time.
nels comprises at least one data page(p), where p=1: 40 11. The at least one computing device of claim 8, wherein
obtaining and storing contact information for visitors asso the historic data is produced within another time period.
ciated with identified visitor activities corresponding to k k k k k