Marketplace advertising software is brilliant at finding campaigns that look scalable. A Sponsored Products campaign has 5.2 ROAS. A branded keyword has 14% ACOS. A competitor ASIN target suddenly converts after a price drop. The dashboard highlights a winner, the platform suggests more budget, and the sensible next step seems obvious: scale it.
Sometimes that is exactly right. Sometimes you are simply paying a marketplace to put a paid receipt on demand you already owned.
The named mistake I see with self-service brand owners is treating attributed revenue as incremental revenue. Amazon, bol, Walmart and most retail media dashboards are attribution systems first. They tell you which ad received credit for an order inside a window. They do not automatically tell you whether the order would have happened without the ad, whether a cheaper placement could have captured it, or whether the extra click still left contribution margin after fees, returns and stock pressure.
My stance: before a campaign gets more budget, it should pass an incrementality smoke test. Not a perfect academic experiment. Not a six-week analytics project that only enterprise teams can run. A lightweight commercial check that asks one practical question: if we spend the next €500 here, are we likely creating new profit or buying our own organic demand back at a discount?
This guide is for brand owners managing marketplace ads themselves, usually from around €1.5K monthly ad spend across Amazon, bol.com, Walmart, Kaufland, Mirakl retailers or DTC support channels. At that level, you may not have enough data for every sophisticated holdout method. You do have enough spend for false positives to hurt.
What current incrementality advice gets right
The better competitor content has moved beyond simple ROAS worship. Pacvue talks about going beyond ROAS with incrementality and iROAS, which is the correct direction for mature retail media teams. Perpetua explains the risk of cannibalization in Amazon Ads and points teams toward smarter strategies that separate growth from recycled demand. Feedvisor and specialist Amazon PPC blogs discuss geo-holdouts, AMC analysis, new-to-brand metrics and branded search tests. Teikametrics, Quartile, BidX, m19 and Helium 10 all emphasize that modern advertising software should automate bids, budgets and campaign management while respecting performance signals.
That is useful. The common gap is that much of the advice assumes one of two worlds. Either you are large enough to run clean experiments with Amazon Marketing Cloud, geo splits and analysts, or you are small enough that the advice becomes “watch TACOS and reduce waste”. Most brand owners live in the middle. They spend enough for wasted attributed sales to matter, but not enough to pause a whole country for a month just to satisfy a measurement purist.
The missing operating layer is a smoke test: a repeatable pre-scale check that sits inside your advertising software and combines ad data with organic position, SKU margin, stock cover, pricing and campaign role. FiveX fits here because ad performance is not reviewed alone. It can be checked next to product profitability, marketplace fees, stock risk, Buy Box or offer status, and the rules that decide whether automation may scale.
Why ROAS can look good while incrementality is weak
ROAS is not lying. It is just answering a smaller question than the one finance cares about. If a shopper clicks your Sponsored Products ad and buys within the attribution window, the ad receives credit. That can be true even when the shopper searched your exact brand name, your product ranked first organically, your price was already best, and the ad mostly moved the click from an organic result to a paid result.
Imagine a Dutch kitchenware brand selling a frying pan on bol.com for €39.95. After commission, fulfilment, packaging, returns allowance and landed cost, the SKU has €9.20 contribution margin before advertising. The branded Sponsored Products campaign spends €420 in a week and reports €5,880 revenue. On paper that is 14.0 ROAS and roughly 7.1% ACOS. Lovely. But the product already ranks first organically for the brand query, has 4.6 stars, and 82% of ad sales come from exact branded terms. If a two-week smoke test shows that pausing the branded ad reduces paid sales but total SKU revenue only drops from €6,300 to €6,120, then the campaign created roughly €180 incremental revenue, not €5,880. At €420 spend, that is not a winner. It is expensive decoration.
Now compare that with a non-branded Amazon campaign for “ceramic non stick pan 28cm”. The campaign spends €380, reports €1,520 sales, and looks mediocre at 4.0 ROAS. But the product ranks organically on page two, receives 41% new-to-brand orders, and total SKU revenue rises by €1,300 during the test while branded sales stay stable. If the SKU keeps €8.40 margin before ads and the campaign creates 39 extra units, the incremental contribution before ad spend is about €327.60. After €380 ad spend, it is not profitable yet, but it is a much better learning candidate than the pretty branded campaign. One needs a bid and conversion fix. The other needs a budget ceiling.
The incrementality smoke test in five checks
A smoke test is not meant to prove the absolute truth forever. It is meant to stop your software from scaling obvious false positives. I use five checks before giving a campaign more budget.
1. Campaign role: defence, discovery or conquest?
Start by naming the campaign’s job. A defence campaign protects branded terms, hero ASINs or retailer shelf space. A discovery campaign learns which generic search terms deserve promotion. A conquest campaign targets competitor products or categories. These roles should not share one incrementality standard.
Defence can be valuable, but it rarely deserves unlimited scaling just because ACOS is low. Discovery can tolerate higher ACOS when it opens profitable search demand. Conquest needs stricter margin permission because clicks are often colder and more expensive. FiveX ad rules are useful here because the campaign role can influence bid ceilings, budget permission and pause thresholds instead of forcing every campaign into the same ROAS target.
2. Organic ownership: are we paying for the shelf we already own?
Check organic rank, share of voice and listing strength before scaling. If your product already sits in the first organic result for a branded query, a paid click has a higher cannibalization risk. If you are absent organically on a category term but ads are creating conversion data, the same ACOS may be more acceptable.
A practical rule: branded exact campaigns that already rank top two organically need a ceiling unless total SKU revenue grows with paid spend. Generic terms ranking below position eight can receive more learning budget when margin and stock allow it. Competitor targets should be judged by incremental units and halo, not by attributed ROAS alone.
3. SKU margin: can the next click afford to be incremental?
Incrementality is only useful when connected to contribution margin. A 20% incremental lift on a weak-margin SKU can still destroy profit. A smaller lift on a high-margin SKU may deserve budget.
Use a simple headroom calculation. If a product sells for €32, keeps €7.60 contribution margin before ads, and converts one paid click in every 12 clicks, the break-even CPC is about €0.63. If your current CPC is €0.88, the campaign needs either better conversion, higher price, lower fees, lower bid or a very strong strategic reason. FiveX’s product profitability view helps make this visible before ad automation treats the target as healthy.
4. Stock and price context: are we scaling into an operational trap?
A campaign can be incremental and still wrong to scale today. If stock cover is eight days, increasing ads may create a stockout that kills organic rank and future revenue. If price has just increased by 12%, yesterday’s conversion rate may no longer be reliable. If the Buy Box is unstable or a bol.com offer is losing delivery promise, extra ad spend can buy clicks into a weaker conversion environment.
This is one of the biggest misses in generic incrementality content. It treats ads as a measurement problem. Operators know it is also a timing problem. In FiveX, stock, pricing, marketplace performance and ad spend can sit in the same cockpit, so the smoke test can say: “good signal, bad week to scale.”
5. Total revenue response: did the business move, or only the attribution line?
The final check is the simplest. When paid spend changes, did total SKU or product-family revenue move in the same direction after normal seasonality? If paid revenue rises but total revenue is flat, you may be shifting credit. If paid revenue falls but total revenue holds, the campaign was probably over-claiming. If paid spend rises, total revenue rises and contribution margin stays positive, you have a scale candidate.
You do not need perfect certainty for every decision. You need enough evidence to stop obvious waste and enough structure to repeat the decision next week.
A practical two-week test plan
Here is the simplest version I would run for a self-service brand.
Week 0: classify. Split campaigns into defence, discovery and conquest. Tag branded exact terms separately. Record SKU contribution margin, stock cover, organic rank, average CPC, conversion rate, attributed revenue and total SKU revenue. Do not start with 40 metrics. Start with the numbers that decide whether the next euro can safely spend.
Week 1: cap the likely cannibals. Do not blindly pause every branded campaign. That is brave in the same way removing your brakes is brave. Instead, cap or reduce the most suspicious branded campaigns by 30% to 50% while leaving listing, price and stock stable. Watch total SKU revenue, organic clicks where available, TACOS and contribution margin. If attributed revenue drops but total revenue barely moves, keep the cap.
Week 2: fund the likely creators. Move a controlled amount of budget into one or two discovery or conquest campaigns with weak organic ownership but acceptable margin headroom. For a €1.5K monthly advertiser, this might be only €150 to €250 of test budget. For a €12K monthly advertiser, it may be €1,000. The point is not to flood the account. The point is to compare total revenue response against the budget you removed from low-incrementality areas.
Decision Friday: write the rule. If the campaign passes, give it a clear scaling rule: increase budget by 15% when total SKU revenue and contribution margin both improve for seven days, stock cover is above 21 days, and CPC stays below the break-even ceiling. If it fails, write the opposite: cap spend, lower bids, add negatives, or keep it as a small defence layer only.
Three scenarios where the answer changes
Scenario 1: the branded bestseller. A skincare brand spends €900 per month on a branded Amazon campaign with 18 ROAS. The hero serum has €11.50 contribution margin before ads and ranks first organically for the brand name. After a 40% budget cap, paid sales fall by €2,400, but total serum revenue falls by only €260. The rule is not “turn off brand forever”. The rule is “brand defence gets a ceiling, not scale budget.” The saved €360 goes into non-branded terms where the product ranks below page one.
Scenario 2: the category climber. A sports nutrition brand spends €600 testing “electrolyte powder sugar free”. ACOS is 38%, which looks high against a 25% account target. But organic rank improves from position 23 to 11, new-to-brand share is 54%, and total product-family revenue rises by €1,900. The SKU has enough stock for 47 days and €12.80 margin before ads. The campaign gets permission to scale slowly, even though its dashboard ACOS is not pretty yet.
Scenario 3: the stockout trap. A toy brand sees a Walmart Connect campaign jump from 3.1 to 5.6 ROAS after a creator video drives demand. The campaign is incremental, but stock cover is six days and replenishment lands in three weeks. Scaling would win the week and damage the month. The rule is to hold budget, protect availability, and reopen the scale gate when stock cover is back above 21 days.
Where advertising software should help
The smoke test should not live in a forgotten spreadsheet. Good self-service advertising software should make it operational.
First, it should connect campaign performance to SKU-level profit. ROAS, ACOS and CPC need to sit next to marketplace fees, fulfilment cost, landed cost, returns allowance and contribution margin. Otherwise automation optimizes the ad account while the business absorbs the surprise.
Second, it should connect ads to availability and price. A campaign with strong incrementality but weak stock deserves a different action than a campaign with weak incrementality and plenty of stock. FiveX helps by bringing advertising, inventory and pricing context into one decision layer.
Third, it should turn the decision into a rule. If a campaign is capped because branded demand is mostly organic, that logic should be visible next month. If a discovery campaign earned scale permission because total revenue moved, the software should remember the threshold. FiveX’s ad automation and AI recommendations are strongest when they operate inside these commercial guardrails instead of chasing yesterday’s attributed sales alone.
The operator takeaway
Incrementality does not have to be intimidating. You do not need to solve causal measurement perfectly before making better marketplace ad decisions. You only need to stop giving more budget to campaigns that cannot prove they are doing a commercial job.
Use attribution to see what received credit. Use incrementality smoke tests to decide what deserves the next euro. That small distinction is where self-service brand owners protect margin, avoid branded-search ego spend, and let advertising software scale the campaigns that actually create demand.
If you manage Amazon, bol.com, Walmart or other marketplace ads from inside your own team, make this the weekly question: did the campaign create profitable demand, or did it simply collect the receipt? The answer should decide your budget before the platform’s recommendation does.