Find the channels with room to grow.

Your attribution says €7.60 back per euro. Only €1.90 is incremental.

Nupact · headroom read
Revenue back per €1 spent
Meta · Retargeting-DEOver-credited
Internal attribution
€7.60
Tested · March audience holdout
€1.90±0.40
Meta · Advantage+ Sales-DERoom to grow
Internal attribution
€2.10
Tested · May lift study
€3.10±0.50
Why it matters

The budget meeting, with proof.

If you run marketing

“Which channels do I scale, and can I show why?”

A tested return per channel, instead of four dashboards that all claim the same conversions. When finance challenges the Meta budget, you put the test on the table.

What you getMore budget for the channels that earn it.
If you sign off the budget

“What does the next euro return, really?”

One number per channel, reconciled to your revenue, with your attribution’s number beside it. Your reporting stays as it is. The test decides the budget.

What you getThe real ROI of your marketing spend.
If you run the company

“Are we growing on real demand, or buying sales we already had?”

Budgets move on measured incrementality, kept fresh by a standing calendar. The top 1% of advertisers run this with in-house teams. You get it without building one.

What you getMore growth from the same budget.
What changes

Tested, then corrected every day.

Test results correct your attribution numbers, so every budget decision you make after the test is a better one.

Nupact · conversions, last 30 days, corrected
Conversions, your attributionConversions, incremental after the test
Meta · Retargeting-DEaudience holdout, 26 weeks ago retest booked
1,900 = 7,600 × 0.25
1.9incremental ROAS
Meta · Advantage+ Sales-DElift study, 17 weeks ago retest running
7,970 = 5,400 × 1.48
3.1incremental ROAS
Google · Search-DE, non-brandgo-dark, 9 weeks ago
6,860 = 9,140 × 0.75
2.6incremental ROAS
TikTok · Spark-DEbenchmark, unmeasured
1.3–2.2k = 1,100 × 1.2–2.0
1.9–3.2benchmark ROAS
How a test works

Three questions, three tests.

A country splits into regions. Some get the change, the rest do not, and the difference in your own revenue is the answer. Which change depends on what you are asking.

Channel offRunning as usual
Go-dark

Is the money we already spend earning its return?

You learnHow much revenue disappears when the channel stops, and how much was coming anyway
New channel liveHeld back
Holdback

Does this new channel add anything?

You learnThe lift the launch created, measured before you roll it out everywhere
Spend doubledRunning as usual
Heavy-up

Does the next euro pay back?

You learnThe return on the extra spend, and how far you can scale before it stops paying

More in the guides: how to run a geo lift test and incrementality testing vs MMM.

Then it feeds the system

One test. Every decision after it.

A test result does not sit in a deck. It flows into the system that already runs your accounts, so every budget proposal, every pacing call and every alert starts from the corrected number until the next test replaces it.

Nupact · approvals1 pending
PendingMon 08:40
Move €1,360/day from Meta · Retargeting-DE to Meta · Advantage+ Sales-DE
Retargeting-DE tested at €1.90 back per euro (±0.40) in March’s audience holdout, below the €2.40 break-even: most of those buyers were coming back anyway. Advantage+ Sales-DE tested at €3.10 incremental in May’s lift study, above the €2.10 your attribution credits it, and is capped by its budget.
Forecast incremental revenue at tested returns, remaining flight
range +€15k to +€65k within the tests’ ± ranges
+€41k
ApproveEdit & approveReject
Day one

Which channels can be tested, and which cannot, before you start.

Before anything is promised, Nupact does the maths on your own regions, history and spend, channel by channel. Some channels can be tested. Some can only be read through the platform. Some are too small to read at all, and the check says so.

Nupact · feasibility checkShowing 4 of 7
ChannelVerdictSpend shareTest designFirst result
Google · Search-DE, non-brandsearch adsWe can test it31%Regional go-dark, half the regions, eight weeks. Brand search excluded; Performance Max read alongside, since it picks up paused queriesWeek 10
Meta · Advantage+ Sales-DEsocial adsPlatform runs it, we check28%A regional split means restructuring the campaign. Meta’s own lift study is cleaner: Meta runs it, we check the result against your revenueWeek 8
Meta · Retargeting-DEsocial ads, own audienceWe can test it9%Holdout on your own customer list: half of it is excluded from your Meta retargeting for six weeksWeek 8
TikTok · Spark-DEshort video adsToo small to test5%Too small for a regional test. TikTok’s own lift study where available; otherwise a benchmark, flagged as unmeasured—
What you see

A year of tests, planned in advance.

A standing calendar keeps every channel’s number current as seasons and channels change. It is what the top 1% of advertisers run in-house.

Nupact · measurement calendar
JanFebMarAprMayJunJulAugSepOctNovDec
Corrected attribution
Every channel, every day, with the test its factor came from and its age
Weekly estimate
Platform lift studies
Google
Meta
Meta
Audience holdouts
Retargeting
Retargeting
Regional tests
Search · heavy-up
Search · go-dark
YouTube · holdback

The corrected attribution and the weekly estimate run all year. Lift studies are run by the platforms and checked by Nupact. Holdouts and regional tests are run by Nupact.

Questions

Incrementality testing: what teams ask first.

What is incrementality testing?

It measures the revenue a channel causes, not the revenue it is credited with. Some regions, or part of your customer list, get a change and a comparable group does not. The revenue gap between the two is what the channel caused.

Incrementality vs attribution: what is the difference?

Attribution shares out credit for the sales that happened, including sales that would have happened anyway. Incrementality asks which sales you would lose without the channel. Nupact keeps your attribution and corrects each tested channel’s number with its result.

What is incremental ROAS (iROAS)?

The revenue a channel adds per euro spent, measured against a control group. Retargeting can show €7.60 per euro in your attribution and test at €1.90 ±0.40, under a €2.40 break-even, while Advantage+ shows €2.10 and tests at €3.10 ±0.50. The budget moves to the channel with room to grow.

How long does an incrementality test take?

A regional test runs about eight weeks, plus a cool-down if conversions lag. A holdout on your customer list runs about six. Running longer buys little precision: the channel’s spend relative to your revenue, and a longer clean history before the test, matter more. The geo test precision guide shows how much. The free feasibility check gives the length for each of your channels.

What happens after a test?

Your attribution reports stay as they are, with the tested number beside them. Every budget proposal, pacing call and alert uses that number, shown with its range and its age. The calendar books a retest before it goes stale.

What if a test is inconclusive?

That is a normal result at mid-market scale. It means the range is too wide to say which side of break-even the channel sits on, though it still bounds the effect, and you hear it plainly. The budget holds, and Nupact proposes the next read: a longer clean history, a repeat to pool with this one, or the platform’s own lift study.

Which channels can be tested?

Each by its own route, depending on your regions, history, spend and campaign setup. Non-brand search and YouTube can be tested by region, retargeting with a holdout on your customer list, and Meta Advantage+ and Performance Max through Meta’s and Google’s own lift studies, which Nupact checks against your revenue. Channels too small to read get the platform’s lift study where available, otherwise a benchmark, flagged as unmeasured. As a rule of thumb, a regional read clear enough to move budget needs a channel spending about 7.5% of that country’s revenue. The free check tells you which is which before any test budget is spent.

Bring one channel. We run the feasibility check for free.

Bring the regional history of one channel and we run the check: what can be read at your scale, how long it takes, and what each design would cost you.

Founder-led, your setup on screen.