Five Questions Before You Trust Your Q4 Numbers

Most brands enter peak trading unable to answer a simple question: did last year's Q4 spending actually grow the business, or did it take credit for growth that was going to happen anyway. Attribution reports are confident, detailed and answer a different question entirely, which is why they are least useful in the period when the stakes are highest. You do not need an econometrics team to close that gap. You need five questions and the willingness to accept the answers.

Incrementality is the revenue that happened because of your marketing rather than alongside it. Attribution tells you which touchpoint preceded a sale; incrementality asks whether that sale would have occurred anyway. In Q4 the difference is at its widest, because purchase intent is elevated for reasons that have nothing to do with your advertising.

Why Is This Worse During Peak?

Because the season does a great deal of your selling for you. A shopper who already intended to buy something in your category, on a weekend when they expect discounts, will find a way to purchase whether or not they saw your retargeting ad. That ad will still be credited, and your reported return on ad spend will look excellent.

The consequence is a specific and expensive failure mode: the campaigns that look best during Q4 are frequently the ones doing the least work, because they are the ones closest to a purchase that was already going to happen. Budget then flows towards them, away from the activity that was actually creating demand, and the following year starts from a smaller base. This is the argument underneath our post on incremental attribution.

Question One: Would This Customer Have Purchased Anyway?

Asked campaign by campaign, honestly. Retargeting someone who added to basket forty minutes ago is not creating demand, it is reminding an existing intention. That has value, but it is a conversion-cost question rather than a growth question, and it should be judged and budgeted differently from prospecting.

A useful test is to ask what would have to be true for this campaign to be creating demand rather than harvesting it. If the honest answer involves the audience already knowing you, wanting the product and having a deadline, you are harvesting, and harvesting is cheap but finite.

Question Two: What Happened to Blended Revenue as Spend Increased?

This is the single most informative check available, it uses data you already hold, and it takes twenty minutes.

Plot total business revenue against total marketing spend by week. If spend rose materially and blended revenue did not move with it, the extra money was largely buying conversions that would have happened anyway. Platform reports will disagree, because each channel is counting the same sales. The store cannot be argued with. This is why we argue for reporting against MER rather than platform return, as set out in our post on MER vs ROAS and in the missing metrics in agency reporting.

Question Three: Did New Customer Volume Actually Increase?

Not revenue, not orders: new customers, counted. Q4 revenue rises easily because existing customers buy more and buy at a discount, so a quarter can post record numbers while acquiring fewer genuinely new people than the quarter before it.

That distinction decides whether peak built anything. New customers are the asset the following year trades on; returning customers buying on discount are this year's revenue brought forward. Count both, separately, every week, using the approach in our new customer CAC guide.

Question Four: What Happened to Contribution Margin?

Per order, net of discount, shipping and fees. This is where the uncomfortable answers usually live.

Revenue can rise, conversion rate can climb, platform ROAS can improve and contribution margin per order can still fall, because discount depth increased and the customer mix shifted. A record Q4 that makes less money than a quieter one is not unusual and it is not a paradox, it is arithmetic, and it only becomes visible if somebody is tracking margin per order rather than revenue. The full case is in our Q4 profit trap post.

Question Five: Did the Improvement Persist?

Look at the four weeks after the promotional window closed. If performance collapsed back to where it started, the season pulled demand forward rather than creating it, and some of what you counted as growth was borrowed from January.

That is not automatically bad, because pulling demand forward at good margin is a legitimate thing to do. It is only bad if you did not know you were doing it, and then set next year's targets against a number that included the borrowing. The January side of this is covered in post-holiday retention.

Do You Need a Formal Test?

Not to get most of the value. Geo holdouts and structured incrementality tests are the rigorous route, they are worth doing, and they need scale, clean conditions and planning. None of those exist during peak, which is why anything structured belongs in September and October rather than November. Our marketing mix modelling guide covers the more formal end.

The five questions are the version every brand can run, on data it already has, in an afternoon. They will not give you a precise incrementality coefficient. They will reliably tell you whether the story your dashboards are telling is the same as the story your bank account is telling, which is the decision-grade version of the question.

What Do You Do With Uncomfortable Answers?

Move money earlier and wider. If the honest read is that peak spend mostly harvested existing intent, the response is to spend more in the window where demand is being formed rather than converted, which for most categories means October and the weeks before the promotional period. That shift costs nothing extra and buys cheaper impressions, per our Q4 CPM inflation post.

The second response is to stop rewarding the campaigns that look best. Judge prospecting on new customer volume and blended revenue movement, and judge retargeting on cost per order rather than on return, so the two are not competing in a comparison that retargeting will always win and always deserve to win for the wrong reasons.

The Bottom Line

Attributed revenue is not incremental revenue, and the gap is widest in the quarter where the numbers matter most. Ask whether the customer would have bought anyway, whether blended revenue moved with spend, whether new customer volume actually rose, what happened to margin per order, and whether any of it persisted. Five questions, an afternoon, and considerably more clarity than another attribution model would have given you.

Our Biggest Q4 Guide Lands Soon

We are finishing the full Q4 Profit Playbook, including the incrementality framework and the scorecard we use to review client peak trading. Our newsletter list gets it first. Join at webtopia.co/newsletter to be first to get it, plus every platform change that matters each Tuesday in Beyond the Clicks.

Want These Questions Answered With Your Data?

Book a call and we will run the five checks against your account and tell you what your spend is really buying, part of how we work as an ecommerce marketing agency and an ecommerce paid media agency.

Frequently Asked Questions

What is incrementality in ecommerce marketing?

The revenue that happened because of your marketing rather than alongside it. Attribution tells you which touchpoint preceded a sale; incrementality asks whether the sale would have occurred anyway.

Why does incrementality matter more in Q4?

Because purchase intent is elevated for reasons unrelated to your advertising, so attributed performance looks strongest exactly when it is least trustworthy.

Do you need a formal incrementality test?

Not to get most of the value. Five structured questions asked against data you already hold will catch most cases where reported and real performance have diverged.

What is the simplest incrementality check?

Compare total business revenue against total marketing spend over time. If spend rose materially and blended revenue did not, the extra spend largely bought conversions that would have happened anyway.

When should incrementality be tested?

Before peak. Structured tests need clean conditions and a stable reading period, neither of which exists once promotional trading starts.

What should you do if the answers are uncomfortable?

Shift budget towards activity that moved blended outcomes and away from activity that only moved attributed ones, which usually means more demand creation earlier in the quarter.