Method 1
Platform lift studies
Brand Lift, Search Lift, Conversion Lift: the audience is randomly split into an exposed group and a control group, then results are compared to isolate the media's incremental effect.
Measurement.
Social is accelerating. Measurement and attribution become the key point: knowing what each euro invested really produces, to arbitrate budgets with confidence.
Page published in French.
The faster Social accelerates, the more measurement and attribution matter.
We combine platform tools, geographic tests, media mix modeling and incrementality protocols, on both paid and organic levers.
Four methods
Method 2
Geographic tests
Regions are assigned to test or control, and spend is modulated through geotargeting. Randomization protects against hidden biases; GeoLift and Meridian GeoX provide tooling for design and analysis.
Method 3
Media mix modeling (MMM)
A statistical model on aggregated data estimates each channel's contribution, with lagged effects and saturation. It does not depend on user data and is calibrated with experiments.
Method 4
Incrementality protocols, paid and organic
Holdout, go-dark, heavy-up on paid levers. For organic, a piece of content cannot be excluded from a zone: we proceed with a staggered launch or with geotargeted paid promotion of the content.
Triangulation
No single method is enough.
Experiments provide causal proof, the model generalizes. The platforms themselves recommend combining them and using lift studies to calibrate the MMM.
Safeguards
Rigor is decided before the first euro is spent.
01
Define before launching
A clear objective and an analysis fixed in advance: an analysis modified after the fact undermines the test.
02
Simulate power
Without prior simulation, confidence intervals are too wide to decide.
03
Choosing the right granularity
Units that are too fine cause leakage between areas and blur the control.
04
Repeat
A single test is a snapshot: regular experiments calibrate the tools.
05
Make assumptions explicit
An MMM rests on causal assumptions: they are to be discussed and verified through experimentation.
Case studies
Browse our case studies.
Bouygues Telecom
Bouygues Telecom: less than €1 per in-store visit thanks to Snapchat Promoted Places
Read the case →
ba&sh
ba&sh: winning new audiences and creating incrementality
Read the case →
AMI Paris
AMI Paris: a full-funnel strategy serving the brand and the business
Read the case →
Printemps
Printemps: activating product feed data on Meta
Read the case →Collaboration