mobile attribution modeling - open analytics nyc

24
The Channel-less Consumer TALK ABOUT US USING #FUTUREM Identifying Attribution within Mobile

Upload: open-analytics

Post on 12-May-2015

1.126 views

Category:

Business


1 download

TRANSCRIPT

Page 1: Mobile attribution modeling - open analytics nyc

The Channel-less Consumer

TALK ABOUT US USING

#FUTUREM

Identifying Attribution within Mobile

Page 2: Mobile attribution modeling - open analytics nyc

Michael KaushanskyAnalytics and Insights, Havas Media, [email protected]@kaushansky

2

New Approach to Consumer Centric Modeling

Page 3: Mobile attribution modeling - open analytics nyc

• Deterioration in key performance metrics (leads & sales)

• We looked for answers: why wasn’t our existing media mix working? Why wasn’t TV, search and display across screens generating the same volume of leads?

• Simulations of existing media mix models pointed to a shift to digital though unclear where. Current models were not reflective of emerging trends in media.

Our global Auto client had experienced a recent deteriorating trend

3

Page 4: Mobile attribution modeling - open analytics nyc

• Click to edit Master text styles– Second level

• Third level– Fourth level

» Fifth level

Media Consumption

The Path to Purchase

The Role of the Consumer Consumer Expectations

What we experienced was a shift in marketing dynamics

Page 5: Mobile attribution modeling - open analytics nyc

Adding to the shift - the emergence of the multi-channel marketplace

2 explosive and complementary markets Local Advertising: $35B in 2015

Mobile Commerce: $39B in 2016

Location, Location, Location 70% of mobile revenue tied to location by

2015 40% of mobile search has local intent

1) BIA/Kelsey, U.S. Local Media Forecast, 03/20122) eMarketer, US Mobile Commerce Forecast: Capitalizing On Consumers’ Urgent Needs, 01/20123) Forrester Report, US Online Sales Forecast, 20094) BIA/Kelsey, U.S. Local Media Annual Forecast (2010-2015), 06/2011

Page 6: Mobile attribution modeling - open analytics nyc

Cross Channel Shopping will be 6X of Online Retail alone

Source: Paypal Media Network

Page 7: Mobile attribution modeling - open analytics nyc

Multi-channel brings a new complexity to the purchase funnel

Source: Paypal Media Network

Nearly 1/3 of retailers credited smartphones with driving traffic to physical stores in 2011, up from 1/5 in 2010-Retail Systems Research, ‘The 21st Century Store: The Search for Relevance”, 6/2011

15% of US online shoppers made a holiday purchase via their mobile, growing 3X over Holiday 2010- Baynote, ‘2nd Annual Baynote Holiday Shopping Survey’, 1/2012- IBM, ‘Benchmark December holiday report’, 1/2012

Page 8: Mobile attribution modeling - open analytics nyc

The purchase funnel becomes… the purchase pretzel

Source: Paypal Media Network

Page 9: Mobile attribution modeling - open analytics nyc

Existing model failed to identify the shift towards digital

Investment

Sales

BlackBox

Consumer journey is unique

Journey metrics are essential to understand the impact of media

Page 10: Mobile attribution modeling - open analytics nyc

Our objective was clear…update the model! Enhance our existing model to address the

unexpected digital shift

Approach – divided the journey into stages, defined success metrics across each stage & linked the metrics to show progression throughout the journey

Methodology – developed 4 models which explained key drivers at each stage, i.e. what drove search, site traffic, leads and sales

Page 11: Mobile attribution modeling - open analytics nyc

Of Course it will work

Page 12: Mobile attribution modeling - open analytics nyc

We modeled the consumer journey across the 5 stages

Stage Success Metric

Awareness Organic Branded Search

Engagement Page Views

Consideration Digital Leads

Shopping Showroom Traffic

Purchase Unit Sales

1

2

3

4

5

Page 13: Mobile attribution modeling - open analytics nyc

Consumer Stages

Awareness

Engagement

Consideration

Shopping

Purchase

Hierarchical Modeling explained impact at each stage

4 Independent ECONOMETRIC Models

Known Strong Existing Correlation

Media Mix

TVPrintOOHDisplaySearchMobileOnline videoPaid socialSocial

Page 14: Mobile attribution modeling - open analytics nyc

We identified 12+ data sources across 52 weeks

Structured Equation

Source Data Begins Timing Data Descriptions Metrics Available

Digital Media (DFA) 2009 daily actual media spend by channel (OLA, SEM) impressions, clicks, hvtsKantar Evaliant 2010 weekly competitor web impressions and spend impressions, spendKantar Stradegy 2011 weekly offl ine GRP & spend GRP, spend

Sysomos 2009 weekly social media buzz data buzz volume, sentimentHitwise/Compete last year daily site visitation and duration visitation, duration

Google Insights 2004 daily total search trends (paid & organic) search trend indexAutodata/Wards 2008 monthly monthly sales data by brand/model/category sales figures

Google Analytics/ Omniture 2009 daily Website/model/configurator/RFQ pageviews visits, configs, RFQ, Model PVHarte-Hanks 2001 weekly owners, prospects, email/mail campaign, etc.

Urban Science 2009 weekly leads by channel by week leadsManufacturer 2009 weekly dealer visits dealer visits

Third Party Sites 2009 monthly Comp model PV, day/week/month model pageviewsBrand Tracker 2010 Semi attitudinal data by month consideration, intent

Census 2000 monthly economic data by Geo economic data

Yt= α + β1Mt + β2Et + β3St + β4Tt + εt, Y (represents the dependent variable, e.g.,sales)M (media), E (engagement), S (searches), T (store traffic), etc.

Page 15: Mobile attribution modeling - open analytics nyc

Results

As suspected… results pointed media consumption varying at each stage

We identified 5 key insights

Page 16: Mobile attribution modeling - open analytics nyc

1. Brand equity dominates & TV is king of paid (36% contribution from TV)Stage

Awareness

Engagement

Consideration

Shopping

Purchase

Page 17: Mobile attribution modeling - open analytics nyc

2. More channels at play in driving site traffic, digital media is the workhorse (33% contribution)

Stage

Awareness

Engagement

Consideration

Shopping

Purchase

Page 18: Mobile attribution modeling - open analytics nyc

3. Paid Search was essential in driving leads influenced by DRTV (paid/Mobile search + DRTV @ 46%)

Stage

Awareness

Engagement

Consideration

Shopping

Purchase

Page 19: Mobile attribution modeling - open analytics nyc

4. Leads were the most significant driver of showroom traffic (40% contribution)

Stage

Awareness

Engagement

Consideration

Shopping

Purchase

Page 20: Mobile attribution modeling - open analytics nyc

5. Showroom traffic was the best proxy of units sales

Stage

Awareness

Engagement

Consideration

Shopping

Purchase

1/1/2

010

2/1/2

010

3/1/2

010

4/1/2

010

5/1/2

010

6/1/2

010

7/1/2

010

8/1/2

010

9/1/2

010

10/1/2

010

11/1/2

010

12/1/2

010

1/1/2

011

2/1/2

011

3/1/2

011

4/1/2

011

5/1/2

011

6/1/2

0112400

2600

2800

3000

3200

3400

3600

3800

3500

4000

4500

5000

5500

6000

6500

7000

7500

8000

Show-rooming was responsible for 82% of sales

Sale

s U

nits

Unit Sales

Showroom Traffic

Page 21: Mobile attribution modeling - open analytics nyc

Moving the consumer further down journey increasingly relies on paid media

Awareness

Engagement

Consideration

Shopping

0%20%40%60%80%

100%

MediaBase

1 2 3 4 5 sales

Page 22: Mobile attribution modeling - open analytics nyc

The new models made our simulations more true-to-life… (what-if scenarios) identified dimensioning returns & ROI

Page 23: Mobile attribution modeling - open analytics nyc

What we’ve learned…

1. Consistently / frequently assess the journey of your consumers

2. Digital media is significant in the middle of the journey

3. Mobile’s role was stronger than previously thought

4. Promotional messaging is effective if consistent at each stage

5. Exogenous data (weather & holidays) have minimal impact

6. Going dark beyond 5-weeks would be detrimental

7. Optimizing media at each stage points to a 12% improvement

Page 24: Mobile attribution modeling - open analytics nyc

Ramifications and things you should consider

Do Your HomeworkDetermine consumer behavior by stage/funnelWeight social influence along journey

Set up for measurement successEstablish KPIs along purchase pathCRM listsPaid, owned and earned media

Unify 3 screensCreative + MediaMerge screen identifiers and interactions

Crunch dataWhat does it tell you?Weight activity against conversion contributionOptimize accordingly

Future ConsiderationWeight social influence along the journeyInvest in approach & tech to unify screens