Integrated Attitudinal Marketing: Improving Performance Lifts Overview.

Integrated Attitudinal Marketing:

Improving Performance Lifts

Overview

Table of Contents

1. Introduction

2. Overview and Explanation

3. Client Benefits

4. “Behind the Curtain”: A Practical Application of the Process

5. Case Studies

3

Beyond Behavioral Modeling

Typically, in the world of direct marketing, a group of individuals (customers, prospects, etc.) can be described and segmented by their lifestyle, life-stage, demographic and socio-economic characteristics—behavioral statistics, in short—for modeling or selection purposes to improve response rates.

There is another individual dimension that impacts buying decisions that was previously unavailable to the empirical world of direct mail modeling—namely attitudes.

Until now.

4

Beyond Behavioral Modeling

Twenty-Ten’s patented I AM® process, provided through Andrews Wharton, can identify and group unique scalable attitudinal groups that are not readily identifiable when using only behavioral-based research methodology.

Relevant attitudes are targeting agents that can in fact impact or incrementally predict the “yes/no” decision.

An I AM® analysis identifies strong correlations between direct marketing performance and specific, attitudinal statements (relevant to the client’s objectives). These correlations are converted into a model, which is then combined with existing traditional behavioral models to score and select those prospects with a significantly higher propensity to perform than was predicted by the base models alone.

5

A Patented Process…..

The “backbone” for I AM is a patented process as follows:

US Patent # 7,472,072

Summary: The patent covers the core capability of assigning consumers to attitudinal profiles. Specifically, the patent covers the ability to analyze large databases and assign the consumers within those databases to specific, predefined attitudinal profiles. In this way, the patent takes attitudinal profiling from being an interesting academic exercise and makes it the basis of actionable, highly effective targeted marketing. Historically, attitudinal profiles have been created via extensive market research studies, and while the attitudinal profiles have been useful, they have not been actionable. This patent allows these profiles to become actionable by attaching the identities of specific consumers, in large numbers, to the attitudinal profiles.

What We Do

We assist our clients in identifying and understanding their most important customers/prospects and then we create custom analytical tools to efficiently reach them at any touch-point.

We have successfully designed marketing initiatives that have time and time again:

Increased response rates Reduced costs Boosted overall profit Improved market share

Built broader brand awareness

Our Approach The I AM® process “actionalizes” key attitudinal insights

and integrates them with pre-existing behavioral targeting models to improve overall targeting and response.

We employ quantitative research to gather raw data on targeted need states, attitudes, behaviors and “response predispositions” that align with consumer response.

I AM® blends a client’s historical response models with the “response predisposition” attitudinal data to convert traditional “regression” analysis to a forward-focused “progression” analysis.

To achieve this, I AM® enhances flat, binary response variables into multi-dimensional, “continuous” variables that integrate and reflect key attitudinal drivers, motivators, and predictors.

Traditional Approach Used by Many

Traditional behavioral-based modeling is developed utilizing historical behavioral attributes that are inherently “binary” in nature. For example:

However, single-dimension, binary, behavioral data provides no indication of:

Degree of interest and emotional engagement in a product offering

Degree of response intent and conviction

Degree of attraction to specific product and offer

1 Responders

0 Non-Responders

Our Approach In Action We model prospects who “best fit” the profile of

responders who reflect “key” attitudinal motivators and response predictors, thus “dimensionalizing” the historical binary, behavioral data.

“1.75”Responder “1”

“.55”

“Low” Emotional involvement Attraction to product and specific offer Conviction to respond

“High”

How It Works

Current ApproachApplying Behavioral Model

Prospect Universe

Prospect Universerejected names

After Integrating I AM®* Attitudinal Model

By integrating an attitudinal ”filter” I AM is able to improve the performance of behavioral

response models by “swapping” out “weaker” prospects with “stronger”

prospects that were previously rejected

Names originally selected

Names originally overlooked

* Patented Technology

Client Benefits

1. Significant performance lifts:

2. Customized targeting solutions optimized to meet client’s specific objectives

3. Likely to identify new audience segments and good prospects being missed by existing models

4. Likely to identify new messaging insights to “tweak” segmented creative

5. Designed to “turbo charge” existing models… not replace them

6. Can apply to any touch point (DM, OBTM, E-Mail)

7. No client costs incurred until/unless model validates on client’s historical data proving it can deliver client objectives.

Potential Client Objectives by Segment

1. Objective: Customer Acquisition/Response Increase

Segment: Third party rented lists

2. Objective: Reactivation Segment: Unconverted previous “handraisers” Segment: Good customer win-back

3. Objective: Retention – Pre-Renewal Intercept Segment: Current subscribers.

4. Objective: Upsell/Cross Sell Segment: Current customers

5. Objective: Improved Lifetime Value

How the Process Works: Creating the “Connective Link”

1. Quantitative ResearchTo gather raw data on targeted need states, attitudes, behaviors and predispositions that align with consumer response

2. Score Each Respondent

Proprietary analytics to determine “tightness of fit” with the optimal target attitudinal profile that comes out of the research

• Field attitudinal e-survey to randomly selected consumers from within the Target Analytics‘ database

• Capture relevant attitudinal info.• Goal is 2k to 3k completed survey

respondents

• Field attitudinal e-survey to randomly selected consumers from within the Target Analytics‘ database

• Capture relevant attitudinal info.• Goal is 2k to 3k completed survey

respondents• Multiple analytical steps applied to

completed survey respondent data to support formal cluster analysis.

• Cluster analysis statistically isolates segment(s) of consumers who fit target attitudinal profiles consistent with client’s response/marketing objective.

• Each survey respondent is then analyzed to determine how closely they fit the target attitudinal profile, based on their survey responses

• Multiple analytical steps applied to completed survey respondent data to support formal cluster analysis.

• Cluster analysis statistically isolates segment(s) of consumers who fit target attitudinal profiles consistent with client’s response/marketing objective.

• Each survey respondent is then analyzed to determine how closely they fit the target attitudinal profile, based on their survey responses

14

How the Process Works: Creating the “Connective Link”

3. Link respondents to client transactional data and/or a

national database file

Append a large pool of individual attributes to support the generation

of predictive algorithm

4. Build probability models to predict “Tightness of Fit”

scores

The “connective” Targeting Engine (predictive algorithm)

• Appended variables are used to develop an algorithm that predicts the “tightness of fit” score

• Built off the survey respondent file, the final algorithm will be used to project a profile “tightness of fit” score for consumers on lists where the same independent variables are appended

• The projected score will provide the ability to rank consumers based on their profile “fit”

• Appended variables are used to develop an algorithm that predicts the “tightness of fit” score

• Built off the survey respondent file, the final algorithm will be used to project a profile “tightness of fit” score for consumers on lists where the same independent variables are appended

• The projected score will provide the ability to rank consumers based on their profile “fit”

• To effectively model the “dependent” variable:

a. Append customer data and transactional history available from Lowe’s internal database

b. Append robust set of “independent” modeling variables to the survey responders.

• To effectively model the “dependent” variable:

a. Append customer data and transactional history available from Lowe’s internal database

b. Append robust set of “independent” modeling variables to the survey responders.

How the Process Works:

Creating the “Connective Link”

Program Execution, and Performance Measurement

• The final predictive targeting engine is designed to maximize the “lift” in the observed “dependent” variable value within top decile of ranked survey respondents

• The targeting accuracy of each final targeting engine is then validated by applying it to a “hold-out” sample of survey respondents.

• Additional - Retro Validation:a. Client supplies mail and response

files from previously mailed campaign. Apply model and compare the scored mailed records to actual responses to validate model lift percentages.

b. Receive mail file only from previously mailed campaign and apply model. Return scored mailed records to client or their agency to validate model lift percentages.

• The final predictive targeting engine is designed to maximize the “lift” in the observed “dependent” variable value within top decile of ranked survey respondents

• The targeting accuracy of each final targeting engine is then validated by applying it to a “hold-out” sample of survey respondents.

• Additional - Retro Validation:a. Client supplies mail and response

files from previously mailed campaign. Apply model and compare the scored mailed records to actual responses to validate model lift percentages.

b. Receive mail file only from previously mailed campaign and apply model. Return scored mailed records to client or their agency to validate model lift percentages.

Timeline and Costs1. Timeline: Commencing receipt of all data assets, the survey,

model build, and validation process completes in about 8 weeks.

2. Costs:a. When and only when the client has agreed to move forward

with a “live” test do they incur costs. The “live” test client commitment would occur after all of the survey work and model build is completed and the model validations report a lift over control that meets/exceeds client’s previously agreed to objective.

b. Client costs consist of two components:

i. Model cost recoupment: One-time cost

ii. Scoring fee: Consists of a modest cpm passing fee on input and a CPM on the records selected for mail.

Vertical Product Type ObjectivePromotional

Channel Lift

Magazine Publishing Celebrity/Lifestyle Response Lift DM 186%     

Hospitality Time Share Response Lift DM 171%     

Hospitality Time Share Response Lift Email 205%     

Insurance Life Response Lift DM 271%     

Catalog Outdoor Gear/Clothing Response Lift DM 66%     

CPG Baby CareCoupon

Redemption DM 195%     

CPG Laundry DetergentCoupon

Redemption DM 366%

Case Histories

Thank you for your time and interest

Andrews Wharton

2171 Jericho Turnpike, Suite 240

Commack, NY 11725

Ph: 631.470.4546

Fax: 631.470.5683

Jack Lee, President & CEO: [email protected] Brickley, VP Sales and Marketing: [email protected]