Shiva Sharma
Gran.Reports Book a call

Incrementality test designer and calculator: lift, iCPA, iROAS

Platforms report what they touched, not what they caused. Design a holdout test that can actually detect a lift, then read what the ads really added.

A flagship tool with Gran: free to use, and a free account saves your report as a PDF.

Currency
I want to
1Your audience
People in the campaign's audience each week: reach for Meta, or users in the test regions for a geo test.
Share of the audience that converts in a week if they see no ads. Your organic or holdout baseline.
%
₹
₹
2The test
Relative lift in conversion rate from the ads. 5–20% is typical for prospecting on Meta and YouTube.
%
%

Share of the audience that sees no ads.

%
Confidence

Holdout size bigger holdouts finish sooner

Cost of the test

What to do next

Result–See result

Private by design: calculations run on your device. Nothing is stored unless you create a report in your Gran workspace.

Built and reviewed by Shiva Sharma. Updated .

How it is calculated

Design mode works out how many users a two-group test needs to detect your expected lift, then divides by weekly reach. A test always runs at least two full weeks, so weekday patterns even out.

n = [ z(1−α/2)·√(p̄(1−p̄)(1/h + 1/(1−h))) + z(power)·√(p₀(1−p₀)/h + p₁(1−p₁)/(1−h)) ]² ÷ (p₁ − p₀)²
Lift                     = (exposed rate − holdout rate) ÷ holdout rate
Incremental conversions  = exposed users × (exposed rate − holdout rate)
Incremental CPA          = spend ÷ incremental conversions

Read mode compares the exposed and holdout groups with a two-proportion z-test and gives the lift with a confidence range. Comparing incremental conversions with what the platform reported gives the share of reported conversions the ads actually caused.

Questions

What is an incrementality test?

An experiment that holds back ads from a random group of people or regions and compares their conversions with those who saw the ads. The difference is what the ads caused, as opposed to what they were credited with.

How big should the holdout be?

Big enough to detect the lift you expect in a reasonable time. Ten percent is common for large audiences; smaller audiences need 20 to 50 percent. The designer shows the weeks needed for each size.

What is a good incremental ROAS?

One that clears your break-even ROAS. Incremental ROAS is usually well below platform-reported ROAS, often by half or more for retargeting and brand search.

If growth is the conversation, let’s have it.

Advisory conversations, speaking, podcasts and media. Tell me what you are working on and I will reply personally within two working days.

Book a conversation hello@shivasharma.co