Introduction to Optimization Group

1

A brief introduction

Prepared for:

2

Who we are …

Optimization Group is a marketing analytics firm offering the following solutions:

Traditional and on-line focus groups

Traditional survey services (CATI and on-line)

Text mining analytics

Conjoint (trade-off) analysis

Data mining and modeling

Dashboard analytics

3

Our People

– Technology “Automate and systematize complex data sets”• Systems analysts• Programmers• Database designers• Process engineers

– Marketing “Make data and analyses work in the real world”• Marketing research & consulting• Corporate brand management• Agency account service• Marketing & media database (applications focus)

Optimization Group consists of people

from two worlds:

4

Our Global Experience

US

Canada

Brazil

Mexico

UK

France

Spain

Poland

Italy

Germany

India

China

Australia

South Korea

Japan

5

Some of our Clients

6

Proprietary tools Unique Solutions

IdeaLoopz® Generating and optimizing ideas

Model Incite Finding the “marketing signal” in “noisy” data

Search Incite™ Context based text search

SiteCRM™ Brand website effectiveness

7

IdeaLoopz

Components:

– brandDelphi™ online ideation system, based on

Rand Corporation geo-political (Delphi) research

technique

– IdeaMap® online concept and messaging

optimization, rooted in conjoint analysis

– Brand Impact Analysis identifies how brand

linkage “turbocharges” specific features and benefits

8

IdeaLoopz: “The Diamond Principle”

Idea

Expansion

Optimized

Idea

Reduction

9

Case Study: Blades Servers

10

Sample

Definitions:

Company size segments were defined as follows: – Medium business = 250-999 employees– Enterprise = 1,000+ employees– Public sector = federal/state/local government, education, medical

IT Decision Maker:– Work in a IT function AND check at least one of the following as it relates to

their job:– Managing and maintaining the servers and storage environment at your site – Helping to set overall company/site strategy regarding servers and/or

storage– Evaluating and recommending new servers and storage products– Recommending or selecting the specific brand of servers and storage– Recommending or selecting the specific configuration of servers and

storage

Business Decision Maker:– Do not work directly in an IT function AND have influence over the server

and storage purchases at their company

11

12

Check the ideas you

like (the basis for the

relevance score)

Then add your own idea

or build on one input by

someone else

13

Next rate the ideas

you just checked (the

basis for the

importance score)

14

Ideal Blade Server

Overall, respondents defined the ideal Blade server as…

Q1

Key Phrase(s) % of Ideas with Word/Phrase

Price/affordable/cost 15.7%

Large/capacity/room/space 11.8%

Service/support/warranty 9.8%

Reliability/quality 9.8%

Fast/quick/time 3.9%

Easy/friendly/simple 3.9%

15

StarsNiche

Static Question Marks

RELEVANCE

IMP

OR

TA

NC

E

Idea Innovation Map

Filter on “Best Ideas”

16

Q1: Stars - Potential Differentiators

17

IdeaLoopz: “The Diamond Principle”

Idea

Expansion

Optimized

Idea

Reduction

18

The Principles of IdeaMap

1. Rooted in conjoint…determines cause and effect

2. Based on fundamental communications theory

(stimulus response)

19

Methodology Overview

Based on customer input from 1st Phase, team generated 9 “tight” attribute/benefit statements – Four categories of elements included:

• Brand/Price

• Servers

• Storage

• Better Together

Elements mixed and matched in an experimental design to form holistic concepts

Respondents evaluate concepts we analyze impact of each element

20

Key Learning

Consistent with work in the PC space among

the B2B target, language that communicates

the ability to keep things running rose to the

top…

– Upgrade/add/replace without taking down infrastructure

– Lower operational expenses – setup time drops from 12

hours to less than 30 minutes

– 24x7 support before, during and after

– Work is transferred to a spare if blade fails

21

Example of “Slicing and Dicing” the Data

Most motivating elements are shared regardless of

OS

Those with a VMS operating system find several elements significantly more motivating

– These elements have a “do more with less” theme

22

Actionable Information for You

Idea

Expansion

Optimized

Idea

Reduction

What is on your

customers’ minds?

– What are there problems?

– What would they like to

see?

What are the “hot

buttons”?

– How to position the idea

– How best to express it

– Messaging to target

segments

23

Model Incite

Optimization Group’s outsourcing solution which

uses our proprietary genetic programming based

modeling software GMAX and other statistical

techniques and tools that your projects require.

24

Classic Regression

$10 $9 $5

$4$7

$8

$3

$2$1

$60

5

10

15

20

25

30

35

40

45

50

0 1 2 3 4 5 6

RR

R

N

R

N

N

N

R

25

Statistical View Of Data

Tools like SPSS would look at the potential relationship between the likelihood of fraud and:

> income

> filing status

> married status

> SIC Code (if business) (2 digit, four digit)

> Gross Revenue

> Date of filing

> etc.

The available universe of variables is limited to only the ones the modeler has input. The limits the potential for greater insight and predictability.

26

One day while perusing the stacks at Powell's Technical, I came across an appealing title: Genetic Programming: On the Programming of Computers by Means of Natural Selection by John R. Koza. He posed an intriguing question:How can computers learn to solve problems without being explicitly programmed? In other words, how can computers be made to do what needs to be done, without being told exactly how to do it?

There is a brave, new way for computers to solve problems without being explicitly programmed and it is Genetic Programming (GP).

Koza's innovation represents an extension of the GA involving more complex structures—computer programs, rather than bit strings. Each program, like the bit strings of the GA, is measured for fitness, the most fit reproducing, the least fit dying off. Eventually, a program is found that solves the problem.

In short: One can harness the principles of Genetic Programming to create software that programs itself.

27

Genetic Programming

$10 $9 $5

$4$7

$8

$3

$2$1

$60

5

10

15

20

25

30

35

40

45

50

0 1 2 3 4 5 6

X2

X1

R

R

R

N

R

N

N

N

N

R

28

PARENT 1 PARENT 2

+ -

A + * X

B C Y Z

How GP works

29

PARENT 1 PARENT 2

+ -

A + * X

B C Y Z

OFFSPRING 1 OFFSPRING 2

+ -

A X

Y Z

*

B C

+

How GP works

30

Mining Key Data Variables

Data mining enables you to see the strength of individual

variables as well as powerful new combinations that help you

better understand your “Key” business drivers.

Variable LiftCommissions earned 375

High face amounts on policies 352

Mix of business sold 240Sales to first time customers 205

Ratio of policies issued to price quotes 200

Rate of underwriting approval 190Weeks since last activity 188

Multiple product sales to same client 170

High retention rate for policies issued 167

Policies denied in underwriting process 153

Lift is a measure of predictability.

31

Targeting your best Prospects

2,068 “unknown” alums have the same predictive variables as the top 10%

of alums who have donated $500,000.

Decile $500K Active Past

Decile Total Donors Donors Donors Unknown

1 5,704 148 1,263 2,225 2,068

2 5,704 29 660 1,919 3,096

3 5,704 17 578 1,677 3,433

4 5,704 14 496 1,435 3,759

5 5,704 7 369 1,261 4,068

6 5,704 3 335 921 4,445

7 5,704 0 280 767 4,657

8 5,704 0 125 560 5,020

9 5,704 1 160 795 4,749

10 5,704 0 98 471 5,136

Total 57,044 219 4,364 12,031 40,431

In the first decile, there are 2,068 “unknown” alums who have the same

predictive characteristics as 148 alums who have donated $500K to the

organization.

32

Customer Satisfaction Model

Our data mining revealed the variables that influence satisfaction.

New Data Combinations

Length of Time for

Call resolution

Getting through to

Cust. Service rep

Team: Durangos,

Thunderbolts

Overall satisfaction

W/rep

33

Customer Satisfaction Window

1.40 1.60 1.80 2.00 2.20

Delivery Rating

0.300

0.400

0.500

A

A

A

A

A

A

A

A

A

A

A

A

A

A

A

A

A

A

A

A

A

B

C

D

E

F

GH

I

J

K

L

M

N

O

P

QR

S

T

U

V

The Customer Satisfaction Window contrasts the perception of the

company’s delivery rating in an area against that area’s importance to

overall satisfaction (GCSI). Here is a list of the areas included in the survey.

Lowest Leverage

A Easy to Get Started

B Sales Person Support

C Easy Installation

D Quality soft/training

E Easy Info Access

F Pick-up Reliability

G Helpful Driver

H Professional Driver

I Easy Tracking

J Delivery Reliability

K Package Condition

L CSA Helpfulness

M Easy Claims Resolution

N Fair Claims Resolution

O Accurate Invoices

P Timely Invoices

Q Easy Acct. Maint.

R Easy Supplies

S Easy Website

T Easy Paperwork

U Easy Customs clear.

V Easy Preparation

Some Potential

Cost of Entry

Highest Leverage

34

Customer Satisfaction WindowThe Customer Satisfaction Window contrasts your “ability to deliver”

customer satisfaction variables against the “expected value” of those

variables.

Highest leverage

Lowest

leverage

Some

potential

Cost of entry

20 40 60 80

Ability to Deliver

-0.100

0.000

0.100

0.200

Mo

dele

d

Exp

ecte

d V

alu

e

A

B

C

D

E

F

G

H

I

J

K

L

Customer Satisfaction Window

A Time to Answer

G Number of Transfers

I Overall Rep Quality

35

Monetizing Customer Satisfaction

36

Case Study: GMAX™ and ROMI

37

Objective

Develop model and understanding of relationships between marketing expenditures and sales

Direct Mail

Catalog

Print Ads

Emails

Online Advertising

Advertising

Pricing

Customer Awareness

Customer Experience

Sales

Market Share

Total Sales $

Client

Controlled

Attitudinal

Outcomes

Sales

Outcomes

38

Print Costs

While print costs appear in the GMAX model, the relationship

is not clearly seen in graphical analysis of print costs by

themselves

Print Out of Pocket

400000300000200000100000

AL

L E

nte

rpri

se

400000000

300000000

200000000

100000000

39

Marketing Communications Variable Tree

Share of voice, print, online, and direct mail allhave an affect on sales Sales

Shipments

Prod B

Share of voice

Prod A

Share of voice

Print

Out of pocket

Direct

Mail

Print

Out of pocket Online costs

Note how Print has an impact by itself AND in combination with Direct Mail

40

Typical ROMI Output

Estimated Sales Impact per $ Invested

Type of Data Total Sales (Direct + Indirect)

Direct Mail $330 -350

Online Advertising $54

Catalog Out-of-Pocket $ $124

Print Varies by CPM “tier”

Overall (SOV) Varies

Email $82

Pricing - 1% change $22MM-$26MM

41

ROMI Model

20000000.00 30000000.00 40000000.00 50000000.00

Predicted Sales Using Model

200,000,000

300,000,000

400,000,000

500,000,000

600,000,000

700,000,000

A

A

A

A

A

A

A

A

A

A

A

AA

A

A

A

Using this model to predict sales does a very good job of matching the actual data

R-Square = 0.62

42

ROMI Simulator

Commercial Education Hospitality

Value of +1 pt in Awareness $11,777,724 $4,611,587 $847,189

Share impact 0.22% 0.35% 0.19%

Value of +1 pt in Consideration $42,152,400 11,212,500$ 2,849,408$

Share impact 0.78% 0.86% 0.63%

Value of +1 pt in ITB $53,394,000 $12,653,368 $4,506,830

Share impact 0.99% 0.97% 1.00%

Linear Effects

Value of +1 point change

43

ROMI Benefits

Identify the marketing levers which

contribute to sales– And those which don’t

Calibrate the impact to guide marketing

investment decisions

Conduct “what if” analyses– How much should I spend to achieve $X sales?

44

Search Incite™

Context based search technology

45

Typical Keyword Search

46

Search Incite Results

47

Search Incite consists of three components:

Data

Query

Ontology Algorithm

- Developed by a team of

experts over 3 ½ years (over 30 man years of work)

- Over 50,000 linguistic

elements

- Up to 500 keywords and

phrases relevant to each knowledge domain

- Customizable, scalable

and upgradeable to adapt to your changing needs.

- Inference engine

- Based on Search Incite’s

intelligent sort algorithm

- Combines linguistic

analysis with automatic pattern matching

Index

How Search Incite Works

48

Ontology Development

49

Content Selection

50

AMEX Verbatim Comments

51

Isolating Problems

52

Automated Corrective Action

Specific words, terms, phrases and issues can be

programmed for automatic intervention/handling.

PC PCPC

Police Dept Intranet

WebServer

CALEA Accreditation

Program Standard Manual

Server

Intranet Server can be hostedinternally or remotely dependingon security, IT infrastructure, andresponse time requirements

Search Incite Hardware Overview

Transfer

ProtocolF

irew

all

54

SiteCRM™

Measuring Brand Website

Effectiveness

(In partnership with crmmetrix)

5555

Site Exposure

On Site Entry

SiteCRM™

Entry Survey

Probably will

purchase

SiteCRM™

Exit Survey

Definitely

will purchase

Re-contact survey 1 week

after website visit

Purchased

Brand

On Site Exit

ROI (Purchase Tracking) Module

Purchased the brand within past 7 days

Spent $4 on most recent purchase

Media Impact – website visit influenced 50%

Estimated Web Influenced Revenue = $2.00

TV

Packaging

WOM

SearchOnline Ad

Email

Typed URL

Offline Media

Media Pull

Lift In Purchase IntentLift In Brand Health Purchase Impact=Estimated ROI

Estimated Web Influenced Revenue (aggregated)– Monthly

Total Unique Visitors/Month = 65,000

Average Estimated Web Influenced Revenue / Visitor = $2.00

Total Estimated Web Influenced Revenue = 65,000 x $2.00 = $130,000

Total Interactive Marketing Spend / Month = $105,000

Estimated ROI = 23.8%

Illustration showing the flow of website visit experience of a single visitor

Business Impact (Sales) Measurement

56

56

VISITOR QUALITY

VISITOR

QUALITY

Who are you attracting

to your website?

Six

Dimensions

SITE

PERFORMANCE

Is the site performing to

my visitors

expectations? What are

the improvements I

need to make to the

website? Are the

visitors accomplishing

their goal for coming to

the site?

BRANDING

IMPACT

Is the visit to the

website driving a

positive change in

opinion for the brand?

Is the content of the

website building a

strong brand

perception?

BUSINESS

IMPACT

Is the website driving a

lift in purchase intent?

Is it driving offline

purchase?

And brand

recommendation?

CAMPAIGN

EFFECTIVENESS

Which campaign

increases Purchase

Intent?

Which campaign drives

offline purchase?

Does the campaign

engage visitors?

CRM IMPACT

Is the website building

customer relationships?

How many of my visitors

registered for the

newsletter?

Is the content of the

website building a

positive brand

perception?

The Six Dimensions analysis, developed by crmmetrix, aims to help marketers identify what

to leverage, in order to turn your website into a powerful marketing engine.

6 Dimensions of Brand Website Effectiveness

57

What’s keeping you up at night?

58

Contact Information

Jeff Ewald

T: 248.459.1194

E: [email protected]

Kenn Devane

T: 917.208.4649

E: [email protected]

59

Intersecting marketing, science and technology™

Search Incite Software Overview

User

htm

l via

aja

x

Ontology Editor

DocumentManager

Custom Views& Reportsfor Results

DocumentSearch

Search Incite Web 2.0 SAAS/ASP Interface

Core DB

Ontology

DocumentStore

Search Incite'sIntelligent Sort Algorithm

Filter Algorithm

DocMetaData

Database Management System

Search Incite Pre-Indexer Background

Process

Customer Specific

Filter Logic and

Triggers

AutomaticImport

DBIDBI

User Auth. Filter(role/permissions)

Organization &User ManagerWeb

Browser http

External Applications SAAS/ASP Interface

Notification Queue(Email/XMPP)

DBI DBI

3rd Party

E-Mail DBI

xml-rp

c

smtpimap

Other DBMS

DocumentWarehouse