which test can i run

Platform lift, holdout, geo or pause: find out which one your spend and set-up allow, and what to fix to open up a better one.

Platform you want to test Spend on it in the last 12 months $ Conversions you optimise for, last 90 days Budget you would put into a brand lift study $ Conversions per week from the ads Test length (weeks) Share you would hold out % Smallest lift worth knowing % relative

Conversion tracking is set up, checked, and de-duplicated I can withhold ads from a random share of my audience I can switch ads on or off by region I can pause ads for two weeks or more

Answers stay in your browser. Thresholds are shown on the page with where they came from.

best option for you

Your options

Test	Can you run it?	Why

Ranked by how much of the design the platform handles for you and how clean the comparison is. A test that is possible but underpowered is marked, not hidden.

What to fix to qualify

Thresholds and where they came from

Rule	Value	Source quality
Meta conversion lift: spend in the last year	$5,000	Reported by secondary guides. Not confirmed on Meta's own pages.
Meta conversion lift: optimised conversions in the last 90 days	500	Same.
Meta conversion lift: tracking set up properly	required	Same.
Meta brand lift: campaign budget	$30,000	Reported by secondary sources. Not confirmed on Meta's own pages.
Geo test: annual ad spend	about $500,000	One 2026 secondary source, a US rule of thumb. Weak.
Manual holdout and pause test	no platform threshold	Your own arithmetic, shown above.
Google, TikTok and others	not encoded	No threshold was verified. Check the platform's current rules.

Last checked 2026-10-23. Platforms change these rules without notice. Verify before you plan around any of them.

How it works

The options

Platform conversion lift: the platform randomises your audience and counts conversions in both groups for you. The cleanest and least work, if you qualify.

Platform brand lift: a survey-based test of awareness and similar measures, not sales. Included because it has its own minimum budget.

Manual holdout: you withhold ads from a random share of your audience and compare. It needs you to control who sees ads, plus enough conversions to see a difference.

Geo test: you switch ads off in some regions and compare. There are few regions and they differ, so it needs more spend and more care.

Pause test: you stop ads for a while and watch what happens. It is always possible and always the weakest, because seasonality and everything else changes at the same time.

The holdout arithmetic

With T conversions among people who see the ads, a holdout of share h would have about C = T x h / (1 - h) conversions if the ads did nothing. The log of the ratio between the two groups has a standard error of about sqrt(1/T + 1/C). At 95% confidence and 80% power:

smallest lift seen = exp((1.960 + 0.842) x sqrt(1/T + 1/C)) - 1

If that is at or below the lift you care about, the holdout is yes. Up to one and a half times it, underpowered. Beyond that, no. For a fuller plan, use Holdout Sizer.

Read this before you plan around the answer

The thresholds are the weak part. Most come from guides that cite the platform, not from the platform. Treat "meets the reported thresholds" as "worth checking", not as a promise.

A test you cannot run is not a reason to trust the dashboard. If nothing is available yet, the fix list says what would open one up.

The holdout figure assumes a clean split. Leakage, repeat buyers and seasonality all make real tests noisier.