Increasing Financial Returns from Cross Media Advertising Dollars with Single-Source Data Leslie Wood & Jim Spaeth Partners Media Trust, LLC 6.24 2008

Increasing Financial Returns Increasing Financial Returns from Cross Media from Cross Media Advertising Dollars with Advertising Dollars with Single-Source DataSingle-Source Data

Leslie Wood & Jim SpaethPartners

Media Trust, LLC

6.24 2008

2 Increasing Financial Returns from Cross Media Advertising Dollars with Single-source Data

Single-source data’s promise for advertising


Marketing mix models are typically driven by weekly store or market level data This is how promotions are executed Providing the sharpest possible contrast between

promoted and un-promoted sales Like a designed experiment --

Sharper indication of how advertising works

TREATMENT CONTROL

Promoted store weeks Un-promoted store weeks


TREATMENT CONTROL

Heavily Exposed 100 GRPs

Lightly Exposed50 GRPS


Weekly market or store data does not provide much contrast for sales with and without advertising exposures Every market includes individuals with a range of exposure levels Most weeks include individuals with a range of exposure levels• Even during planned hiatuses due to make-goods and

scheduling mishaps

7654321

It’s not black &whiteF

requency


As measurement progressed --Nielsen SAMI UPCBi-months 4-weeks weeklyRegions Markets Stores

And analytics progressed -- Response Bump Promotion Marketingmodels analysis models mix models

Marketing budget allocations shifted --


Share of Marketing Expenditures

Advertising Promotion





Single-source data links exposures to purchases for the same households Providing a sharp contrast between exposed and

unexposed purchases Like a designed experiment --

Will a sharper indication of how advertising works lead to a clearer view advertising’s contribution and redress the marketing allocation imbalance?

TREATMENT CONTROL

Exposed purchases Unexposed purchases


More accurate measures of advertising at work

Measures of advertising have two components: Occurrences Audiences

The precision of available occurrence data varies markedly by medium Advantaging some media over others Disadvantaging model accuracy and impartiality



Audience-data suitability for marketing mix models also varies noticeably: Television has excellent weekly data, but weekly market

level data is still estimated for all but the largest markets Internet is local-market challenged Radio, magazines, outdoor and newspapers are all

weekly-data challenged. Magazines are also local-market challenged

Although weekly magazine measurement shows promise

And all media have issues with measures of ad exposure

Advantaging some media over others

Disadvantaging model accuracy and impartiality



Single-source data measures actual ad exposures at the household level Accurate measures of exposure are linked to

subsequent purchases Can be rolled-up to markets, stores, weeks, days,

consumer or shopper segments

Providing more accurate measures of advertising at work

HH # Date: 1/1 1/2 1/3 1/4 1/5 1/6 1/7 1/8 1/9 1/10 1/11 1/12 1/13

86 TV Ad Exposure X X

86 Print Ad Exposure X

86 Category Purchase X X X X

86 Brand Purchase X X


Truer model of advertising working Single-source data can segment exposed versus unexposed

purchase occasions With actual measures of exposures Without the noise introduced by market-level averages and

imprecise exposure measures clouding advertising effects


Truer model of advertising working

3. Frequency Response

1. Initial Impact

2. Daily Retention

2+

3+

1+

100

AdImpact


Truer model of advertising working

AdImpact is a model of advertising working Takes advantage of single-source data’s direct link

between exposures and purchases

Marketing mix modeling could be enhanced with single-source data To better represent advertising value To readily integrate advertising and promotion effects

with best of breed approaches

13

Eleven marketing mix models were developed for a single brand

Price & Promotion

Price, Promotion & GRPs

Price, Promotion & Impressions

Price, Promotion, Impressions & Adstock

Price, Promotion, non-linear Impressions & Adstock

Standard MMM (Price, Promotion, non-linear GRPs & Adstock)

Standard MMM & Market Factors

Standard MMM & Frequency

Standard MMM & Frequency substitute Impressions for GRPS

Standard MMM, Reach & Frequency

Standard MMM, Reach & Frequency w/ Market Factors

Each aggregated from individual HH panel data Not individual HH models … nor market level models, but somewhere

in between

14

Two sample model fits

Eleven models by two geographies

Market factors consistently enhance standard MMM

Reach & Frequency are the best of the enhanced media models

SMM DMAPrice & Promotion 0.58 0.53Price, Promotion & GRPs 0.585 0.540Price, Promotion & Impressions 0.589 0.546Price, Promotion, Impressions & Adstock 0.593 0.553Price, Promotion, non-linear Impressions & Adstock 0.594 0.558Standard MMM (non-linear GRPs & Adstock) 0.588 0.547Standard MMM & Market Factors 0.603 0.580Standard MMM & Frequency 0.591 0.553Standard MMM & Frequency substitute Impressions for GRPS 0.594 0.559Standard MMM, Reach & Frequency 0.592 0.551Standard MMM, Reach & Frequency w/ Market Factors 0.600 0.557

r2

Eleven models by loyalty groups

Segmenting respondents by loyalty groups does not enhance the models

Market factors and reach & frequency continue to consistently enhance the standard MMM

76 Loyalty Groups 1000 Loyalty GroupsPrice & Promotion 0.547 0.525Price, Promotion & GRPs 0.547 0.525Price, Promotion & Impressions 0.548 0.525Price, Promotion, Impressions & Adstock 0.548 0.525Price, Promotion, non-linear Impressions & Adstock 0.548 0.525Standard MMM (non-linear GRPs & Adstock) 0.547 0.525Standard MMM & Market Factors 0.577 0.528Standard MMM & Frequency 0.548 0.525Standard MMM & Frequency substitute Impressions for GRPS 0.548 0.525Standard MMM, Reach & Frequency 0.548 0.525Standard MMM, Reach & Frequency w/ Market Factors 0.578 0.528

r2

Eleven models by exposure groups

Grouped by level of exposure to the brand’s advertising over the entire period

Surprisingly not better than standard markets

SMMExposure Groups

Price & Promotion 0.58 0.583Price, Promotion & GRPs 0.585 0.588Price, Promotion & Impressions 0.589 0.589Price, Promotion, Impressions & Adstock 0.593 0.590Price, Promotion, non-linear Impressions & Adstock 0.594 0.592Standard MMM (non-linear GRPs & Adstock) 0.588 0.591Standard MMM & Market Factors 0.603 0.598Standard MMM & Frequency 0.591 0.591Standard MMM & Frequency substitute Impressions for GRPS 0.594 0.592Standard MMM, Reach & Frequency 0.592 0.590Standard MMM, Reach & Frequency w/ Market Factors 0.600 0.598

r2

Working Exposures

Numerous AdImpact models have shown that advertising’s short-term sales effect diminishes dramatically within 28 days Substantiated by J.P. Jones’ STAS models This finding lead to recency planning

Those impressions reaching a consumer more than 28 days prior to a category purchase occasion are likely to have little or no effect

Those impressions reaching a consumer within a 28 day window prior to a category purchase occasion may have an effect on the brand selected We call these, “Working Exposures”

Grouped by working exposures, the exposures delivered 28 days prior to a category purchase, over the course of the full period

Working exposure groups yield consistently better models

Working Impressions as a variable further enhanced the media models

Eleven models* by Working exposure groups*Plus three with Working Impressions as a variable

Eleven models summary

Single source data enables…

Segmenting households by working exposures, rather than by market

Adding Reach & Frequency, and Working Impressions, along with a non-linear transformation of GRPs

Providing a better model fit in this case

But what does it tell us about advertising?

Markets vs. Working Exposure Groups:Volume contribution by marketing factor

Standard MMM:

80%

7% 13%

BASE PRICE/PROMOTION ADVERTISING

Better informed management decisions


Greater financial returns from advertising expenditures

Media ROI, as estimated by marketing mix models, has frequently been reported to be small or negative• No wonder advertising’s share of marketing

expenditures has fallen over the years

0

50

100

150

200

250

300

350

400

450

Brand Observations

TV

RO

I In

dex

TV Advertising ROI Index

unprofitable profitable

Source: Mike Duffy & Paul Beck, ACNielsen, “From Empirical Generalizations to Lessons Learned: Interpreting Market Response Evaluations Across Brands and Categories, MSI Metrics and Financial Performance Conference, 2006


Greater financial returns from advertising expenditures

Enhancing marketing mix models with single-source data would enable:• A sharper indication of how advertising works• More accurate measures of advertising at work• Employment of a truer model of advertising working• Result in better informed management decisions• Greater financial returns from advertising expenditures


And a shift of marketing spending back towards advertising?

Project Apollo data has provided significant learning

Which will be applied as new single-source databases emerge

Single-source enables more accurate and actionable media models

And evidences a stronger contribution of advertising to sales







Past Present Future

Documents

Increasing Financial Returns from Cross Media Advertising Dollars with Single-Source Data Leslie Wood & Jim Spaeth Partners Media Trust, LLC 6.24 2008