mtl+ecommerce #40 - media buying for ecommerce using data and machine learning

41
DIGITAL MEDIA + E-COMMERCE MTL+ECOMMERCE JUNE 28, 2017

Upload: mtlecommerce

Post on 22-Jan-2018

187 views

Category:

Marketing


2 download

TRANSCRIPT

DIGITAL MEDIA + E-COMMERCEMTL+ECOMMERCEJUNE 28, 2017

ADVISO ? JF RENAUD ?

2

► 70 Digital Experts since 2002

► Many E-commerce & lead gen clients

► Result oriented► Expertises :

► Digital media► Omnichannel Strategy and Digital

Transformation► Analytics► Content Marketing

► Entrepreneur & Business Owner

► Hyperactive► 2 x Daddy► Professor► Xillenial

STUFF I WOULD LOVE TO KNOW IF I WOULD BE RUNNING AN

E-COMMERCE PNL

3

UNDERSTANDMODERN CPA

4

Le CPM

5

CPMDIGITAL MEDIA CPA - 2005

6

Media Delivery (Ad server / DCO / DSP)

Data handling (DMP)

Data (1st, 2nd & 3rd party)

Media buy - CPM

DIGITAL MEDIA CPA - 2017

Analytics Layer (on-site, attribution)

INVEST IN AUDIENCES NOT ON MEDIA BRANDS

7

HERE ARE SOME READERS OF HUFFINGTON POST

8

… AND HERE ARE SOME READERS OF THE GAZETTE

9

REFINE MEASUREMENT

10

11

UNDERSTAND THE ROLE OF EACH INVESTMENT

12

UNDERSTAND ATTRIBUTION

13

USE AN AD SERVER

14

REM

Winner in GA

View-Through Conversion

Click-Through Conversions

USE DATA THAT HELPS

15

Winner needs more budget!

Pause losers sucking up budget

16

USE NEUROMARKETING FOR CREATIVE TESTINGBANNER TESTING AND RESULTS

WHY WE USE NEUROMARKETING: User Attention and Top-of-Mind Recall are getting harder each year. Users now become banner blind in a matter of days instead of weeks. This is why Adviso uses neuromarkteting to test creative attention and appreciation, especially for high stakes campaigns. The results in this use case speak for themselves...

Improved CTA Button: In this creative test, we clearly saw both 300x250 and 728x90 had nearly invisible CTA buttons. The creative was corrected and we saw a 20% increase on the Interaction Rate.

Improved Conv. Rate: Based on the results of CDS’ optimized banners, we saw an average of 2.56 conversions / 1000 ad impressions. This was 50% higher than what we saw with previous creative with similar budgets.

+20%

+50%

Call-to-Action ProblemTest revealed CTA in Phase I banners was almost invisible. We made impactful changes in Phase II and saw a lift in the banner interaction rate.

OPTIMIZE CONTEXT, CREATIVES AND LANDING PAGES

17

A STRUCTURED CRO PROCESS

18

FOCUSING ON CONVERSION RATE OPTIMIZATION

18

Off-Site Analysis of your Conversion Funnel is where we operate in Downstream CRO mode. Our Media Team uses tried, tested and true scientific marketing methods to cull the excess media weight at every stage of the funnel. The graphic on the left illustrates our process.

Analyze media presence

Analyze in-market media traffic

Analyze engaged media traffic

Analyze active shopping media traffic

Remove media placements with low online visits, brand awareness lift, video completions or ad viewability

Remove media placements with low cart-to-detail rates or purchase intent

Remove media placements with low brand equity lift or social interaction

Remove media placements with low buy-to-detail rates or purchase behavior

CONVERT

Downstream Conversion Rate Optimization

Discover

Consider

Convince

Cause Effect

DO NOT UNDERESTIMATE IMPACT OF LANDING PAGES

19

None of your business :)

INJECT DATA. LOTS OF IT.

20

DATA ?

21

VS

CAPTURE, ORGANIZE & DEPLOY AUDIENCE DATA

22

Device ID (Mobile)

CRM Records

Browser Cookie

Social / Dark Social

Login / Authentication

Universal ID

23

GEO DATA- Bid by proximity- Conquest on

competitors’ turf

24

PUBLIC DATA

DATA PROVIDED BY USER

Every clic is data !

► Interests / in market for► Height / weight► Sex / sexual preferences► Technology used► Physical location► Brand attachement

PROBABILISTIC DATA

26

27

3R PARTY DATA

28

DRIVE TO STORE MOBILE / OMNICHANNEL TRACKING

Cost / Store visit In-store visits rate

29

CONNECT DATA

SUPERPOWER YOUR EFFORTS

30

AUTOMATE! EXAMPLE OF BUSBUD

31

Mandate:

Automate a 500K “product” media buy architecture with a small monthly budget in 14 languages and 55 countries.

How we solved this :

Use an algorithm instead of a human to decide which route to support with media buys :

► Availability► Route conversion rate► Margin► ...and many more

SEM

GIVE YOUR CAMPAIGN MANAGERS SUPERPOWERS

32

BLINDFOLD OPTIMIZATION AKA FINDING YOUR MONEY KEYWORDS

Ponctual steps to repeat1. Extract 2 reports2. Normalize reports3. Create a pivot table 4. Make a Vlookup5. Create a 2nd pivot table6. Add some calculation

fields7. Make a 2nd Vlookup on

calculation fields8. Create final table

FROM TOOngoing script that provides CM with recos & alerts based on opportunity size / urgency

GIVE YOUR CAMPAIGN MANAGERS SUPERPOWERS

33

BLINDFOLD

New Average CTR

9.4%

+32% +/-14%

Impact probability

99.9%

0.001 valeur-p

34

New Average CPC

1.4

-26% +/-8.4%

Impact probability

99.8%

0.002 valeur-p

GIVE YOUR CAMPAIGN MANAGERS SUPERPOWERSBLINDFOLD

QUALITY SCORE OPTIMIZATION ASSISTANT► Proactively push insights to CM to improve quality scores and improve

campaigns performance

35

AUTOMATE FINDINGS AND HAVE MORE TIME TO WORK ON THEM

Sorry, confidential

PREDICT IMPACT OF A CHANGE OF BUDGET

36

Potential of gain

Simulation

Actual

CREATE AN HYPER GRANULAR KEYWORD STRUCTURE► Automate super boring work : 170 000 lines !

37

...TO REFLECT DEMAND COMPLEXITY

Keyword AVERY HIGH VOLUME

Keyword BHIGH VOLUME

Keyword CHIGH VOLUME

Large semantical distance

Small distance

Small distance Keyword DLOW VOLUME

AUTOMATE CLUSTERS

41

THANKS !Jean François Renaud

[email protected]