Stop Lighting Money on Fire on Black Friday: Is Your Next Ad Dollar Actually Incremental?

A couple of years ago I watched a brand do something that still makes me wince. Black Friday morning, revenue pacing about 30% ahead of plan, blended ROAS glowing well above target. So the call was made on the spot: pump the budgets. Facebook spend roughly doubled by lunchtime. Everyone felt like a genius.

The day finished with a great-looking MER and a profit number that was almost identical to what a flat budget would have produced. All that extra spend bought almost no extra profit. It just rode demand that was already going to convert and lit a few hours of margin on fire.

Here's the thing nobody tells you about peak. A green dashboard is not a reason to scale. During the noisiest, highest-volume days of the year, your reporting will glow no matter what you do, and that glow hides whether your next dollar is actually doing any work.

So the question I want you to sit with this Black Friday is the only one that matters when you're deciding to spend more: is that next dollar incremental? An incremental ad spend test is how you answer it instead of guessing, and most brands never run one.

Why a great MER can be lying to you in peak week

Think about who is buying on Black Friday. A big chunk of them have known about you for weeks. They saw the ads in October, they have been waiting for the sale, and they were always going to purchase. Peak is, to a large degree, a demand-harvesting moment.

That means a lot of your sales would happen at almost any spend level. When those people convert, your MER looks fantastic and your attribution lights up. But you didn't create that revenue with the marginal dollar. You captured demand that was already in the building.

This is why I get nervous when a team tells me "we're crushing our efficiency target, let's spend more." Beating your MER target is not, on its own, a reason to scale. It might just mean you're harvesting cheap demand. The real test is whether spending more makes the overall pie bigger, or just blends a great existing result down to a slightly less great one.

My take: in peak week, treat your blended numbers as a comfort blanket, not a steering wheel. They feel reassuring and they will let you overspend for hours before you notice.

The discipline: scale spend X%, revenue must rise X%

The rule we hold ourselves to is simple to say and weirdly hard to follow. If you scale spend up by some percentage, revenue has to rise by at least that same percentage for the move to be worth it.

It's basically hour-over-hour testing. You bump budgets 25%. Then you watch. Did revenue stay flat? Did it also climb 25%? Did it jump 50%? If you push spend up 25% and revenue barely moves, those extra dollars were not incremental. You just paid more for the same sales.

The trap is that this is so easy to skip. It is genuinely tempting to wake up on Black Friday, decide you're going to do a big number today, and slam the budgets without holding yourself accountable to the rise. I've done it. Most people who've run a few peaks have done it. The brands that stay disciplined are the ones who decide the bar in advance: we will only keep the extra spend if revenue moves with it.

If you only take one thing from this piece, make it this. Don't let yourself spend more without naming, out loud, the revenue increase that has to follow. Then check whether it actually showed up.

The scrappy signals that tell you the truth in real time

Hour-over-hour revenue is the headline, but it's noisy on a single day. So when we're trying to decide whether an additional push is incremental, we stack a few cheap signals on top of it. None of these is perfect. Together they get you to a confident-enough call.

• Net-new visitor rate. Of the traffic you're buying right now, how much is genuinely new versus people who already know you? If you scale and the new-visitor share holds or rises, the dollars are reaching fresh demand. If it collapses, you're paying to re-touch the same crowd.
• One-day-click ROAS. Not your 7-day blended figure, the tight one-day window. It's a rough proxy for "did this spend cause a fast purchase," and watching it move as you scale tells you whether the marginal dollar is pulling its weight.
• Percentage of first-time orders. If spending more lifts the share of orders coming from first-time customers, you're buying growth. If the extra spend just shows up as more repeat purchases, you're probably subsidising people who'd have bought anyway.
• Post-purchase survey trends. This is the scrappiest and one of the most useful. Watch the "how did you hear about us" responses, specifically the "in the last day" and "in the last week" answers. If those suddenly double as you scale, you're genuinely creating new awareness. If they're flat, the new spend isn't landing on new people.

The way it actually plays out: revenue is pacing well ahead, you're tempted to push, so you go and look. One-day-click ROAS is holding, net-new visit share is steady, and the survey's "heard about you in the last day" answers have jumped. Now you scale with conviction, maybe up 50%. If those signals were flat instead, you'd hold, take the efficiency, and let the result be great rather than diluting it.

It's an artsy science, not a clean one. But stacking these beats flying on a blended number that's designed to look good in November.

A cleaner read: split the country and watch the margin

If you want to move past in-the-moment signals and actually measure this, the strongest version is a geo-based scale test. It's the closest thing to proof you'll get during a period this noisy.

The setup is straightforward. Split your country into regions. Hold one slice at business-as-usual spend, push another slice up by 50%, push a third up by 100%. Then compare. You're not looking at platform-reported ROAS, you're looking at whether the higher-spend regions actually produced enough extra revenue to justify the extra spend.

I'll give you a made-up but representative shape of what these tests tend to show. The +50% region comes back clearly incremental and profitable. The +100% region produces a noisier read, and on the balance of evidence it's likely not profitable at the margin. In other words: scaling into peak does drive real incremental orders, but only up to a point, and that point is usually lower than the budget your gut wants to set.

That's an enormously useful thing to learn before next year. It turns "how hard do we scale on Black Friday" from a vibe into a number. Instead of tripling budgets because the dashboard is green, you flight to the level the test actually supported, maybe +50% to +70% over your pre-peak base, and you stop lighting the rest on fire.

One honest caveat: a single test on a single day carries a wide confidence interval. Treat the first read as directional, run it again next peak, and you'll build real conviction over a couple of seasons rather than betting the budget on one noisy result.

What I'd actually do this peak

If you've got a big sale coming, here's the short version of how I'd keep yourself honest.

Decide your scaling bar before the day starts. Write down the revenue rise that has to accompany any budget increase, so the call isn't made on adrenaline at 11am.

When you're tempted to push, don't look at blended ROAS first. Look at the marginal signals: net-new visit share, one-day-click ROAS, first-time-order percentage, and the survey's recent-awareness answers. Scale hard only when they move together.

And if you possibly can, run a geo split this season, even a rough one. The point isn't perfection this year. It's that next year you'll be making peak budget decisions from evidence instead of from the glow of a dashboard that was always going to look green.

Beating target feels like permission to spend. It isn't. The brands that protect their margin through peak are the ones who keep asking the boring question every time they reach for the budget slider: would this revenue have happened anyway?

If you'd value a fresh read on whether your scaling is actually buying growth or just harvesting demand you'd have won regardless, that's a big part of what we dig into in a Signal/Noise Audit. It's a calm look at where your spend is genuinely incremental and where it's quietly leaking, no pitch attached.