Incrementality Testing for Beauty Brands: Run a Holdout, Find True iROAS, Then Reconcile It With MER

KPI cleanup series · Incrementality

Platform ROAS tells you what the ads touched. iROAS tells you what they caused.

By Nikki Lindgren · November 16, 2026 · 8 min read

A skincare brand spends $60,000 a month on Meta and the dashboard reports a 3.00 ROAS. Then it pauses Meta in a set of matched regions for four weeks. The regions that kept ads sold 22.0% more than the regions that did not, which works out to $90,000 of revenue the ads truly caused. That is an incremental ROAS (iROAS) of 1.50, half of what the platform claimed, and below the 1.61 this brand needs to break even.

This post is the causal chapter of our KPI cleanup series on ACOS, MER, weighted CAC and blended measurement. If those terms are new, start with our guide to ACOS, MER and weighted CAC. Buyers are searching for this: incrementality testing draws 590 searches a month at $61.83 a click, iROAS draws 480, marketing incrementality draws 210 and marginal ROAS draws 170 (Semrush US, November 2026). All ten results on page one for incrementality testing are measurement vendors. None show the math on a beauty P&L, so that is what this post does.

Why platform ROAS overstates what ads cause

Every platform takes credit for any purchase inside its attribution window. A shopper who saw a Reel, searched your brand name, clicked a Google brand ad and bought gets counted by both Meta and Google. Neither platform asks whether the shopper would have bought anyway. Add up the platform claims and they often exceed what the business earned from ads.

Incrementality asks one question: what would revenue have been without this spend? The answer is the gap between a group that saw ads and a matched group that did not. Divide that gap by spend and you get iROAS, the only ROAS that tells you whether the money worked. Our explainer on what ROAS measures covers the reported side. This post covers the causal side.

Step 1: Run a geo holdout test

A geo holdout is the most practical test for a DTC beauty brand because it uses your own order data, not the platform's. Split your markets into matched groups by past revenue, hold ads out of roughly 20% of them, and keep everything else equal: same offers, same email calendar, same prices. Run for at least four weeks so one weekly purchase cycle does not swing the result.

Here is the illustrative skincare brand at the end of a four week Meta holdout. Revenue in the control regions is scaled up to match the size of the exposed regions, so the two numbers compare like for like. Inputs are assumptions for a worked example, not a client result.

MeasureExposed regionsControl regions (scaled)Gap
Revenue$500,000$410,000$90,000
Lift over control22.0%
Meta spend$60,000$0
Meta reported revenue$180,000
ROAS3.00 reported1.50 iROAS

The incrementality factor is iROAS divided by platform ROAS: 1.50 divided by 3.00 is 50.0%. Half of the revenue Meta claimed would have arrived anyway, mostly from returning customers and brand search. Pull the order counts from Shopify or your order system, never from the ad platform, or the test inherits the bias it is meant to remove.

Step 2: Compare iROAS with break even, not with a benchmark

Break even iROAS comes from contribution margin before ads. On this brand's orders, product cost is 20.0%, shipping and fulfillment is 15.0% and payment fees are 3.0%, which leaves 62.0%. Break even iROAS is 1 divided by 0.62, or 1.61.

First order result of the Meta test
$4,200 loss

$90,000 of incremental revenue at a 62.0% margin returns $55,800 of contribution against $60,000 of spend. The dashboard showed a 3.00 ROAS. The business lost $4,200 on the first order, before repeat purchases.

Now translate the result into a target your media buyer can use every day. Divide break even iROAS by the incrementality factor: 1.61 divided by 50.0% is 3.23. Meta's reported ROAS must stay above 3.23 for this brand to break even on what Meta causes. Check the inputs on the Pennock ROAS calculator before you change a bid, and rerun the test each quarter, because the factor moves with creative, audience overlap and season.

Step 3: Reconcile iROAS with MER and weighted CAC

MER is total revenue divided by total spend. It is honest about the whole business but blind to which channel caused what. Incrementality factors fill that gap. Apply each channel's factor to its platform ROAS and the channel view finally reconciles to the blended view.

ChannelSpendPlatform ROASFactoriROASIncremental revenue
Meta$60,0003.0050.0%1.50$90,000
TikTok$25,0001.8080.0%1.44$36,000
Google brand search$15,0008.0010.0%0.80$12,000
Weighted total$100,0003.4540.0%1.38$138,000

Total revenue for the month is $400,000, so MER is 4.00. The platforms claim $345,000 of that, or 86.3%. Incrementality says ads caused $138,000, and the other $262,000 (65.5%) is baseline demand from returning customers, organic search and word of mouth. Weighted iROAS is 1.38, below the 1.61 break even, so paid media lost $14,440 on first orders in a month where MER looked strong.

Budget move the test supports
$3,968

Google brand search has the highest reported ROAS and the lowest iROAS. Moving $10,000 from brand search (0.80 iROAS) to TikTok (1.44 iROAS) adds $6,400 of incremental revenue and $3,968 of contribution, assuming TikTok holds its iROAS at the margin. Confirm that with a second holdout before moving more.

The same logic cleans up acquisition cost. Weighted CAC built on platform conversions understates the true cost of a new customer, because baseline buyers are counted as ad wins. Divide incremental spend by incremental new customers and track it next to MER every week. Our Meta vs TikTok comparison for beauty brands and the CPM calculator give you the cost inputs.

Where iROAS sits next to ACOS, MER and weighted CAC

MetricQuestion it answersWhat it counts
ACOSDid the ad order pay for itself?Ad attributed sales
MERDid the month make money?All revenue
Weighted CACWhat did the channel mix cost per customer?Customers, weighted by channel
iROASDid the spend cause the sale?Incremental revenue only
Incrementality checklist

1. Pick one channel and hold it out of about 20% of matched regions for at least four weeks.

2. Freeze offers, prices and email sends across both groups for the full test.

3. Measure revenue and new customers from your order system, not the platform.

4. Divide incremental revenue by spend for iROAS, then compare it with 1 divided by your margin.

5. Turn the result into a platform ROAS target: break even iROAS divided by the incrementality factor.

6. Retest each quarter and after any major creative or budget change.

What to do this week

Pull last month's platform ROAS for each channel and your blended MER. Add up what the platforms claim and compare it with total revenue. If the claims exceed 80.0% of revenue, at least one channel is taking credit for baseline demand. Start with the channel that has the highest reported ROAS and the most brand search overlap, run a four week holdout, and set its new ROAS target from the result.

If you want a second set of eyes, the free audit below rebuilds iROAS, MER and weighted CAC from your own orders and margins.

Frequently asked questions

What is incrementality testing?

Incrementality testing measures the sales an ad channel caused, not the sales it claimed. You hold ads back from a control group, compare it with an exposed group, and the gap is incremental revenue. In the example, the exposed regions earned $500,000 against $410,000 in a matched control, a 22.0% lift.

What is iROAS and how is it different from ROAS?

iROAS is incremental revenue divided by ad spend. Platform ROAS divides all attributed revenue by spend, including sales that would have happened anyway. In the example, Meta reported a 3.00 ROAS while the holdout showed $90,000 of incremental revenue on $60,000 of spend, an iROAS of 1.50.

What iROAS does a beauty brand need to break even?

Break even iROAS is 1 divided by contribution margin before ads. At a 62.0% margin, break even iROAS is 1.61. Below that, each $1 of spend returns less than $1 of contribution on the first order.

How do I use incrementality results with MER?

MER is total revenue divided by total spend, so it includes baseline sales that ads did not cause. Multiply each channel's platform ROAS by its incrementality factor, then compare the weighted iROAS with your break even. In the example, MER was 4.00 while weighted iROAS was 1.38.

Free audit
Find out what your ads truly cause

Our team will rebuild iROAS, MER and weighted CAC on your own orders and margins, and design the first holdout test.

Get the free audit
SEONikki Lindgren