ROAS vs incremental contribution
Platform ROAS answers a bookkeeping question: of the revenue Google can see, how much got attributed to ads? Incremental contribution answers the executive question: how much revenue exists because the ads ran that would not have existed otherwise?
The two drift apart in predictable directions. Brand Search collects demand that would mostly have arrived anyway. View-through conversions credit impressions nobody acted on. PMax blends genuinely new demand with captured existing demand into a single flattering number. At $50k–$500k+ per month, the gap between attributed and incremental is large enough to change budget decisions — which is why the question is worth asking and why the answer has to come from a test, not a dashboard.
When incrementality tests are worth the cost
A holdout costs money twice: you give up volume in the control regions, and you spend senior time designing and reading the test. That cost is justified when a real budget decision hangs on the answer — scaling PMax, defending brand spend, or setting next year's allocation — and when the measurement layer is clean enough that the result would survive a CFO's second question.
It is not justified when conversion tracking still double-counts, when offline outcomes never make it back into the account, or when nobody will act on the answer either way. Testing into a broken measurement layer produces a precise-looking number that is wrong.
Geo holdouts and practical constraints at high spend
The workhorse design for Google Ads at this spend band is the geo holdout: a set of matched regions where spend is changed or paused, compared against regions where it continues. The practical constraints are qualitative but real:
- Design: regions have to be matched on demand, seasonality, and competitive pressure, not picked alphabetically. A badly matched control group invalidates the test before it starts.
- Contamination: customers cross region lines, national media leaks into holdout geos, and a regional promotion mid-test can quietly wreck the comparison. The design has to anticipate these, not discover them in the readout.
- Duration: the window has to cover a normal demand cycle and your conversion lag. What that means in weeks depends on your business — a same-day ecommerce conversion and a 60-day B2B pipeline need different designs. We set the window from the account's own data rather than quoting a universal number.
PMax and incrementality
Performance Max is where the incrementality question gets asked most, because PMax reports one blended ROAS with limited visibility into which queries and placements produced it. A strong PMax ROAS does not tell you whether the campaign found new demand or absorbed demand your Search campaigns — brand especially — were already converting more cheaply.
Before paying for a holdout on PMax, the cheaper step is structural: understand how the campaign overlaps with your existing Search coverage and whether brand is being absorbed into the blended number. We cover the overlap question in detail on the PMax cannibalization page, and how we run PMax as a deliberate lane on the Performance Max service page.
Measurement prerequisites
Every credible incrementality test stands on the same foundation: conversion actions that mean one thing each, offline outcomes — closed revenue, qualified pipeline, booked appointments — imported back into the account, and brand separated from non-brand in reporting. Without those, a holdout measures noise.
Offline conversion import is the piece most high-spend accounts are missing, and it is the piece that moves the test's outcome metric from 'form fill' to 'revenue finance accepts.' The Business Value framework is how we keep the test's success metric tied to a business number rather than a platform one.
How an X-Audit informs whether you are ready to test
The free X-Audit reviews the measurement layer directly: conversion action hygiene, offline import status, brand-versus-non-brand separation, and whether the numbers you forward to finance are trustworthy enough to base a holdout on. The output is a findings deck that tells you plainly whether an incrementality test would produce a reliable answer on your account today — or what has to be fixed first.
You get the findings whether or not you ever become a client; ongoing management is optional. For U.S. advertisers spending $50k–$500k+ per month on Google Ads.
Questions we get
- How long should a holdout run?
- Long enough to cover a normal demand cycle for your business and to let lagged conversions land — that window is different for an ecommerce account than for a B2B pipeline with a 60-day close. The honest answer is that duration follows your conversion lag and seasonality, not a universal number of weeks. We define the window from your account's own data before any test starts, and we would rather tell you the account is not ready than run a test short enough to produce a confident wrong answer.
- Can brand Search be held out?
- Yes, and brand is often where the question is sharpest, because brand terms capture demand that already exists. A brand holdout needs care: turning brand off in a region hands that demand to whoever bids on your name next, and the test has to account for organic capture on the other side. It is a legitimate design when the measurement layer is trustworthy and the business accepts the short-term risk in the holdout regions.
- Does a strong PMax ROAS equal lift?
- No. PMax ROAS is an attribution number from inside the platform. It counts captured brand demand and view-through credit alongside genuinely incremental revenue, so a strong figure is compatible with low or even zero incrementality. The only way to know what PMax caused is a controlled test — which is why measurement readiness comes first.
- Do you sell an incrementality software product?
- No. SalesX is a Google Ads agency. Test design, execution, and readout are part of the management engagement, supported by the X-Vault automation suite and BigQuery reporting. There is no software seat to license.