Incrementality helps you answer a simple question:

Did your marketing create additional sales, or would those sales have happened anyway?

For example, imagine that a campaign generated 100 sales. That number alone does not tell you whether the campaign caused all 100. Some customers may have purchased without seeing the campaign.

An incrementality test compares a group that received the advertising with a similar group that did not. If the advertised group generated 100 sales and the comparison suggests that 80 would have happened anyway, the campaign created an estimated 20 additional sales.

Those 20 sales are the campaign’s incremental impact.

Incrementality can help you understand:

Incrementality is one part of your measurement strategy. Attribution and marketing mix modeling provide other perspectives.

Attribution helps you understand where conversions received credit. Incrementality helps you determine whether the marketing caused additional conversions.

These methods work best together. Attribution supports day-to-day optimization, MMM supports broader planning, and incrementality gives you a controlled way to validate impact.

Triple Whale offers two ways to run an incrementality test. The right option depends on where you advertise and the question you want to answer.

GeoLift measures the impact of advertising across different geographic areas.

Some locations receive the advertising change, while similar locations provide a comparison. Triple Whale uses the comparison to estimate what would likely have happened without the change.

Use GeoLift when you want to measure the impact of a campaign or channel across geographic markets.

During setup, you can group locations by:

Geo Pods is selected by default. Geo Pods are groups of locations designed to create a more balanced comparison.

Learn more about Geo Pods and how GeoLift tests work.

Meta Conversion Lift measures the additional impact of your Meta ads.

Meta separates eligible people into two groups. One group can receive the ads included in the experiment, while the other does not. The difference between the groups helps estimate how many additional sales or conversions the advertising created.

Use Meta Conversion Lift when you want to test the impact of Meta advertising directly.

Your Meta ad account and selected campaigns must meet Meta’s eligibility requirements before an experiment can run.

A useful experiment begins with a clear question. Decide what you want to test and what you may do differently once you have the answer.

Before setup:

For a GeoLift experiment, choose the primary metric that matches your question:

From the Incrementality page, start a new experiment and choose GeoLift. The setup flow contains six steps.

To create a GeoLift experiment:

GeoLift is selected by default when the Test Design screen opens. Active Campaigns is also selected by default.

The GeoLift setup contains six steps:

You can select Save Draft during setup if you are not ready to complete and launch the test.

Choose the testing method and the type of campaigns you want to test.

Select GeoLift to measure marketing impact by comparing geographic regions that receive the advertising with similar regions that do not.

Then choose one of the following campaign options:

Select Next to continue.

Configure the primary metric, regional granularity, and advertising activity included in the experiment.

Choose a primary metric

Select the result you want to measure:

Revenue is selected by default.

Choose the regional granularity

Select how locations should be grouped:

Geo Pods is selected by default.

Select the test scope

Choose the channels, tactics, campaigns, or ad sets you want to include in the experiment.

You can browse the available advertising activity using:

Use Custom Categories to select from your organized campaign groups, or use Source to browse activity by advertising platform.

Review the Total Daily Spend shown for your selections. When you are ready, select Next.

Review the geographic design recommended for the experiment.

One group will receive the advertising change. The other group provides the comparison used to estimate incremental impact.

Review the proposed regions and holdout percentage before continuing.

Changing the geographic groups while the experiment is running can make the result harder to trust.

Enter the planned duration and budget for the experiment.

Review the experiment guidance shown during setup. If Triple Whale warns that the planned experiment may not be sufficiently powered, adjust the budget, duration, or geographic design before launching.

Add a name and set the timeline for your GeoLift experiment.

Name the experiment

Enter a clear Experiment Name that will help your team identify the test later.

Review the start date

Confirm the Start Date shown for the experiment.

Set the duration

Enter the number of days the experiment should run. Triple Whale recommends a minimum duration of 21 days.

The calendar displays the experiment’s start and end dates based on the selected duration.

Select the cooldown period

The Cooldown is the period after the experiment ends when results are allowed to stabilize before they are analyzed.

Review the recommended cooldown period and use the calendar to confirm the complete timeline. The calendar displays the active experiment period and cooldown period in different colors.

When the schedule is correct, select Next. If you are not ready to continue, select Save Draft.

Before launching, review:

If something is incorrect, return to the relevant step and update it before launch.

For supported GeoLift experiments, Triple Whale applies the geographic targeting changes needed to create the test and comparison groups.

To create a Meta Conversion Lift experiment:

Complete the following fields:

Select the experiment configuration that best matches your testing goals.

Optimize for Speed is selected by default.

The holdout group is the percentage of the audience that will not see the campaigns included in the experiment.

A larger holdout can help the experiment reach a result sooner, but it also prevents more eligible people from seeing the tested campaigns. A smaller holdout limits that impact but requires a longer test.

Review the experiment parameters

When you select a recommended configuration, Triple Whale automatically updates:

Confirm that the holdout percentage and test duration match your testing goals before launching.

Optional: Use a manual configuration

To enter a custom holdout percentage or test duration, turn on Manual Configuration.

Triple Whale displays a warning that custom settings may affect test reliability. Review your custom settings carefully before continuing.

When the experiment settings are correct, select Launch.

If Meta rejects the experiment, review the displayed error and correct the underlying account, campaign, permission, or eligibility issue before trying again.

After launch, use the Incrementality table to monitor the experiment’s status, Data Health, and spend delivery.

Avoid making changes based on one unusual day. Look for a sustained warning or a meaningful difference from the experiment plan.

The table can show the following statuses:

When an experiment is Under Review, it remains visible, but its complete results are not ready.

Data Health helps you identify whether an experiment is progressing as planned.

Possible statuses include:

If an experiment needs attention or is at risk, investigate the issue before relying on its result. Missing data or uneven delivery can weaken the comparison and prevent the experiment from producing a useful answer.

Use the Spend Test Tracker to compare actual delivery with the experiment plan.

The tracker can show:

GeoLift spend pacing uses the following indicators:

Focus on sustained pacing problems rather than normal changes from one day to the next.

If spend is on pace but delivery is low, review:

If spend and delivery are both below plan, review:

GeoLift results are organized into:

For a revenue-based experiment, Results Overview can include:

For an acquisition-based experiment, Results Overview can include:

An incrementality result is an estimate, not a guarantee.

Review the Probability of Direction and the displayed result range before deciding what to do next.

Probability of Direction indicates how strongly the analysis supports a positive or negative effect. A higher probability provides stronger evidence for the displayed direction.

The result range shows the uncertainty around the estimate.

A narrow range means the likely outcomes are closer together. A wide range means there is more uncertainty.

If the range includes both a meaningful positive outcome and a meaningful negative outcome, the experiment may not provide enough evidence to confidently increase or reduce spend.

The Deep Dive compares the GeoLift result with available attribution models, including:

Use this comparison to understand whether attribution and GeoLift tell a similar story.

Different results do not automatically mean that an attribution model or experiment is incorrect. Attribution assigns credit for conversions, while GeoLift estimates the additional impact caused by the marketing.

For a deeper results walkthrough, see Reading Incrementality Test Results.

Meta Conversion Lift results show the estimated additional impact of the tested campaigns and the confidence in that estimate.

Review both the estimated impact and its confidence before making a decision. A large estimated result with low confidence may look promising, but it does not provide strong enough evidence for a major budget change on its own.

For more information, see Meta Conversion Lift Experiment.

Apply the result only to the campaigns, markets, audiences, and dates included in the experiment. Do not assume that one result represents every campaign or all activity on a channel.

A reliable positive result indicates that the tested marketing created additional value.

Consider increasing spend gradually rather than making one large change. Continue monitoring performance because results may change as you scale.

A negative result suggests that the tested activity may not have created additional value during the experiment.

Before reducing spend, review:

An uncertain negative result is not the same as clear evidence that the advertising reduced performance.

An inconclusive result does not mean that the campaign had no impact. It means the experiment did not produce a clear enough answer.

Before repeating the experiment, review whether it had enough:

If attribution looks strong but incrementality is weak, the channel may be receiving credit for demand that already existed or that another activity created.

If incrementality is strong but attributed performance looks weak, your attribution model may be missing part of the channel’s contribution.

Use both perspectives to guide planning, but keep the experiment’s scope in mind.

If an experiment cannot launch, falls behind plan, or does not produce a useful result, check:

For a GeoLift experiment:

Avoid changing geographic exclusions, budgets, comparison groups, or experiment settings while an experiment is running. Unplanned changes can weaken the comparison and make the final result harder to trust.