Once a lift or geo test is back, how do we turn its result into a budget move and a new ROAS or CPA target, and how far should one test move spend?
Our answer
Read a finished test's interval against breakeven. Scale one step (about 15 to 20% in Stella's examples) when its low end clears it, hold and retest when it straddles and cut in steps below. Move targets only partway.
- Confidence
- Medium
- Strongest proof
- 2 controlled tests
- 2 datasets, 3 practitioners, 9 opinions or claims
- Sources
- 11 sources
- 16 statements
New here? Take the 1-minute tour
Why this is the answer
3 reasons and our judgmentSize each budget step to the downside loss. In an illustrative example from Stella, an incrementality testing software maker, a 40% margin and a 90% interval of 2.1 to 3.7 times let an $8K loss budget fund a 20% step on $250K a month.
One test is too noisy to adopt outright. Single Facebook acquisition tests in the geo-holdout database of Common Thread Collective (CTC) range from 0.5 to 2.4 times around a median of 1.14. So CTC weights a first result by its confidence rather than adopting it.
A good result may not hold at higher spend. Phil Kiel's Haus test returned 1.93 incremental ROAS, and he expects that won't hold at four times the budget. So he plans a retest at the new spend and checks CPMr and frequency weekly.
Our judgment
- Read the result against breakeven, which Stella puts at 1 divided by contribution margin before ad costs (Setting a blended efficiency target). Sellforte's worked example against a 1.5 bar agrees, reading an interval of 0.1 to 1.3 as against the investment and minus 0.5 to 2.2 as resolving nothing. Elevarus's scale, reallocate and retest cutoffs agreed before launch and LiftLab's reading of an interval that includes zero point the same way.
- Move one step of about 15 to 20% of the channel's monthly spend per finished test, near the spend you tested, then retest at the new level. Stella's examples are a 15% increase for the following month and the 20% step above, and it advises a scale-up test before going far past the tested range. ATTN Agency's and LiftLab's scale tests look for where returns weaken. This sits inside the 10 to 20% monthly steps in Splitting spend across Google campaign types, and daily budget raises within the step follow How fast to scale a winning campaign.
- Reset the target in two moves. First move your incrementality factor partway from the benchmark toward the test result, weighted by its confidence. Then divide your business ROAS target by that factor, or multiply your incremental cost-per-purchase goal by it, to get the platform target. CTC's own example (a 2.0 business target and a 1.2 factor giving about 1.8) does not follow from plain division, which gives about 1.67. Eightx's retargeting example, with incremental ROAS 60% below reported and true customer acquisition cost about 2.5 times what the dashboard shows, is the same arithmetic on the cost side.
- Check the move monthly against the predicted range and contribution margin. After two months below that range, send the moved channel back to a holdout before more scaling, and keep the test's dates and promotion with the result. This follows Stella's check-date rule, CTC's rule that incremental ROAS is subordinate to contribution margin and its warning that a Black Friday target may not fit January, and Haus's simulation in which precise tests beat doing nothing 82% of the time against 62% for noisy ones.
Our judgment, built from Stella, Sellforte, Elevarus, LiftLab, Phil Kiel, ATTN Agency, Common Thread Collective, Eightx and Haus: our reading of these sources, not a finding from any one of them.
What the sources say
16 statements from 11 sourcesCTC says a ROAS target from a Black Friday lift test may not fit January; keep the test dates and promotion attached so a seasonal result does not become a permanent budget rule.
“Keep the test dates and promotional context attached to the result so a seasonal finding doesn't become a permanent budget assumption.”
In one UK brand's Haus geo test, a Meta campaign to lapsed customers returned 1.93 incremental ROAS on returning-customer sales against Meta's reported 6.2x; Haus counted 40% of the orders Meta claimed.
“a retention campaign showing 6x in Ads Manager is not a 6x campaign.”
Stella, which sells testing, says a budget call should state direction, size and a check date, with breakeven at 1 divided by margin (a 40% margin means 2.5x). Its step-size examples are illustrative.
“Cutting a channel because the point estimate looked soft is how brands kill channels that were working.”
Sellforte says geo tests often need 30% to 50% of eligible regions held out, and screens tests by wanting expected lift at least twice the detectable lift. Its examples are hypothetical.
“In many practical designs, holding out roughly 30% to 50% of eligible geographies creates a stronger comparison.”
Common Thread Collective says a blended ROAS target is not the same as the on-platform target: with a 2.0 business need and Meta's 7-day click incrementality factor of 1.2, the on-platform target is about 1.8.
“Your blended ROAS target (what the business needs) is not the same as your on-platform bid target.”
CTC's H1 2026 report puts Meta's average incrementality across its clients at 113.4%, from below 25% to above 200%. It blames heavy reliance on Meta and advises diversifying, not cutting Meta.
“If your Meta incrementality is low, the answer is rarely to cut Meta. It is to diversify”
Haus simulated a year of test-driven budget moves 36 million times: precise tests beat doing nothing 82% of the time vs 62% for noisy ones, and 15 noisy tests a year did no better than 6.
“The precise scenarios beat a frozen-budget baseline 82% of the time. The noisy ones, only 62%.”
CTC, which sells geo tests, says across hundreds of its tests Facebook acquisition's median incrementality factor is 1.14, with single tests ranging from 0.5x to 2.4x, so one test is not enough.
“single test results are insufficient”
CTC says even a well-calibrated incremental ROAS should answer to revenue and contribution margin; if iROAS improves while margin is flat or falling, recheck the measurement.
“Even a well-calibrated incremental return on ad spend should always be subordinate to the realities of revenue and contribution margin”
Eightx says to put measured incremental ROAS into your CAC math: if retargeting comes in 60% below what Meta reports, true CAC is about 2.5 times the dashboard figure.
“Replace ROAS with iROAS in your CAC math.”
Sets a lower floor than Sellforte's 30% to 50%: Elevarus says a holdout under 15% of regions is useless, and below $40K a month on a channel, use platform lift tests or skip testing.
“A 5% holdout on a national account produces a control group too small to compare against anything.”
LiftLab, which sells mix models and lift tests, says geo holdouts answer keep-or-cut questions, while budget planning needs the return on the next dollar from spend-level tests. No data.
“An incrementality test without a pre-committed decision is measurement theater.”
Recast, which sells a mix model, says its model constrains a channel's ROI to match a holdout test's result, weighted by the test's uncertainty. It gives no rule for how far one test should move spend.
“The model constrains its own estimate of the channels' ROI to be consistent with this measured ROI”
ATTN Agency says it tests budget scaling with a control at current spend against cells at 25%, 50% and 100% more, to find where returns diminish. It shows no results.
Cozy Earth's CMO turns test results into an incrementality factor: if Meta reports 3x for every 1x incremental, it holds a 3x in-platform target. CTC gives every channel a factor, updated only by new tests.
“if we need a three X in platform to get our one X incrementality”
WorkMagic, which sells mix modeling, says Lifepro calibrated its model with lift tests, raised Meta spend about 200% at BFCM and landed within about 2.2% of predicted ROAS; 57.9% of incremental orders came on Amazon.
“Lifepro scaled BFCM spend by 200%, with actual ROAS landing within ~2.2% of the predicted outcome”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it3 figures from cited sources
- Where each source standsA chart placing each source on the question
- The strongest disagreementThe best case against the answer, from a cited source
- How sure we areWhat its confidence rests on, and what would change the answer