Causal Lift FAQ

Causal Lift is our geo-based incrementality testing solution that helps marketers measure the true impact of their campaigns by comparing treated and control regions. Instead of relying on clicks or attribution models, Causal Lift isolates what actually drove conversions or sales. To learn more, check out this article.

A Geo Lift experiment reveals the true incremental impact of your advertising campaigns on outcomes like conversions, customer acquisition, or traffic. We typically translate that into actionable metrics such as:

These metrics help you understand how much you're truly paying for incremental results—filtering out what would have happened anyway.
At a high level, a causal lift experiment can uncover opportunities to reallocate budget – for instance, you might discover a campaign isn’t actually driving incremental sales, allowing you to invest it elsewhere or discover the incremental value is undervalued compared to traditional attribution allowing you to increase budget for that channel.

Causal Lift operates at the campaign level. You can test a single campaign or a group of campaigns together. However, if multiple campaigns are grouped into one test, the results will be aggregated, and we won’t be able to isolate performance by individual campaign.

Because the attribution is based on location and not cookies, we can measure sales from all channels as long as we can track the location of a sale. We can measure conversions from Amazon Seller Central natively and even physical retail if you provide us with the data.

We rely on standardized geographic partitions with sufficient independence to minimize spillover:

We use a stratified randomization approach. Regions are grouped into “bins” based on historical performance (e.g., past sales or conversions). Within each bin, regions are randomly assigned to either the test or control group. This ensures a balanced design—test and control groups are similar in size, business importance, and historical behavior. To learn more about the approach behind causal lift, check out this article.

Before the campaign begins, we verify that test and control regions are statistically similar in terms of volume and trend using historical data. During the analysis, we also use synthetic controls to account for any remaining differences, creating a reliable counterfactual to compare against.

We calculate confidence intervals around the estimated lift to quantify uncertainty. These intervals show the range where the true lift likely falls. If the confidence interval includes zero (or is too wide), it suggests the result may not be statistically significant.
Additionally, we use simulation and sampling-based error estimation before and after the test to ensure the findings are robust—not just a result of random variation.

Geo Lift relies on detecting outcome differences across regions. For tests focused on new customer acquisition, we recommend a baseline volume of at least 100 acquisitions per day.
Additionally, the campaign or set of campaigns being tested should be impactful enough to generate a detectable signal above the noise. If the expected lift is too small relative to natural geographic variability, results may be inconclusive.

In some cases, we will recommend focusing on the traffic outcome (Sessions) to measure impact, in this case, volume is typically large enough to run an experiment.

There are cases where measuring incremental traffic is both strategic and more practical than measuring conversions.

A common example is testing SEM brand campaigns. The core question is:

“What’s the true added value of bidding on our own brand name in search results?”

By measuring sessions, we can determine whether users still reach your site organically when paid ads are turned off. Traffic reacts faster than conversions, so this type of test can deliver directional insights in as little as 2 weeks.

You might also choose to optimize for traffic in cases where:

When appropriate, we can estimate incremental conversions by applying your historical conversion rate to the observed lift in sessions—giving you a directional sense of performance while keeping testing efficient.

Running a Causal Lift test with Polar is designed to be collaborative, low-risk, and insight-driven. Here's what the process looks like:

From start to finish, we manage the technical heavy lifting—so you can focus on learning what truly drives incremental results.

The setup is simple and lightweight. You just need to have:

If the experiment is focused on traffic outcomes, we’ll also need the Polar Pixel installed on your site to accurately measure sessions by region.

Once these are in place, everything else—from geo-targeting to test tracking—is handled on our side.

Test duration depends on the type of campaign and the volume of data available.

We simulate your expected results upfront and recommend the shortest duration that will deliver a reliable confidence interval based on your goals.

We are opened to discussing any options for Geo-lift experimentation. For example, if you want to assess incremental impact on physical sales, or test a TV campaign, we can assist you in the design, implementation and evaluation.