Skip to main content
All CollectionsConcepts
Understanding Geolift
Understanding Geolift

What is Geolift and how do we use it to attribute post-view sales?

Updated over a week ago

What is a Geolift?

The Geolift is a measurement technique based on AB testing, a classic among the measurement methods used in data science. It is based on robust statistical calculations and allows one to prove, without any ambiguity and with total reliability, the precise business increment (or the absence of increment) that a media brings.

⚠ To date, we are only able to deploy the Geolift for Meta and Google Search media (especially to evaluate Branding or Generic Campaigns).

We use Geolift to provide you with post-view data, which gives insight into the value brought by your offline marketing campaigns (ads in public spaces, radio ads, word-of-mouth, etc.).

The Geolift is the most efficient method we have to complete your attribution data from online interactions (clicks and views on online ads, affiliation, influence…) with offline interaction data. Combining the two gives Quanticfy the ability to show you exhaustive sales attribution data.

How does it work?

The principle of a Geolift is to randomly select a sample of exposed and non-exposed geographical areas in a 50/50 split and to verify that the sample is representative in terms of population, revenue, and customer value. It follows this process:

  • Step 1: we establish a random sample of exposed and non-exposed geographical areas in a 50/50 split, and we make sure the sample is representative and unbiased

⚠ Our role, as experts and statistical guarantors, is to constitute these samples in such a way that there is no bias (total randomness) nor any contact migration between the two populations.

  • Step 2: once the postal codes are defined for each population, we communicate them to you. We will ask you to upload those postal codes in the Google and/or Meta campaign manager interface (we will guide you through this process).

  • Step 3: the non-exposed areas are turned off in the Facebook campaign management console.

  • Step 4: we conduct the test for a duration of 1 to 6 months

  • Step 5: the statistical calculation can measure the incremental value of new customers brought in by exposure to ads. This allows us to measure the incremental acquisition via this channel and determine the post-view.

⚠ Our Geolift is not the same as the Meta Geolift. We offer an unbiased solution because we conduct the Geolift as an external party relying on your CRM data (from Shopify) to establish population samples with similar consumption patterns. This allows Quantic Factory to provide you with accurate A/B test results.

How do we use Geolift results to estimate offline attributed sales?

To know how many sales were brought by offline ads, we go through the following steps:

  • Step 1: We use our attribution model to look at the online attributed sales ("clicks") for Meta (sales that were brought through Facebook and Instagram ads online)

  • Step 2: we do the Geolift and see how many sales your business generated (thanks to offline and online ads). The Geolift gives us the offline ("views") part we were missing before.

  • Step 3: we look at the total number of online-lead sales ("clicks")

  • Step 4: we deduce what is the number of offline generated sales ("views")

We then combine the post-view data with attribution data following our sales measurement method, to provide you with exhaustive data about your sale's attribution.

When and how requesting a Geolift?

If a significant share of your sales cannot be attributed to post-click (clicks on online ads), it is interesting to request a Geolift that will provide you with the missing data. Even if you don't heavily invest in offline ads (radio, TV, physical billboards, etc.), word-of-mouth and other offline channels might be driving sales and it is impossible to estimate how much until you do a Geolift.

You can contact us through email or open a ticket using the chatbot at the bottom right of this page to request a Geolift.

Did this answer your question?