How Much Should You Scale for BFCM? A Geo Test Says +50%, Not +100%

Two brands walk into Black Friday week with the same plan: spend more.

The first one doubles the budget on Wednesday, watches the orders pour in, and calls it a triumph. Revenue is up, the dashboard is green, everyone's buzzing. The second one only lifts spend by half, keeps the rest of the budget sitting in reserve, and ends the week with less top-line revenue but more actual profit in the bank.

Here's the thing - most founders are running the first playbook and assuming it's the smart one. I want to walk you through a piece of testing that suggests the second brand made the better call, and then give you a way to make that call yourself without paying for an expensive measurement tool.

The test that started this

I came across a clean little experiment from a fairly large DTC brand recently. They had a suspicion I've had for years: that the marginal return on ad spend during BFCM isn't nearly as incremental as it feels.

The logic is simple when you say it out loud. Peak week is when you're harvesting demand you've already built. People have known about the product for months. They were always going to buy. So when you pour extra money in during those six days, a chunk of it is just paying to reach people who'd have converted anyway.

To test it, they split their home country into three cells:

• 10% of the country got business-as-usual spend (their baseline).
• 45% got spend scaled up by 50%.
• 45% got spend scaled up by 100%.

This ran across the core six days of BFCM, on one product line, on Meta only. Not the whole account, not every channel. One clean read on one category.

What the three cells actually showed

The +50% cell was the standout. Scaling up by half was clearly incremental and clearly profitable. The extra dollars brought in extra orders at a cost per order that still made sense.

The +100% cell is where it got interesting. The read was noisy, and the honest version is that the confidence interval around it was wide enough to drive a truck through. But the directional signal said that doubling was probably no longer profitable on the margin. The cost of each additional order at that level had crept past the point where it paid for itself.

So the takeaway wasn't "don't scale". They still made a great return overall. The takeaway was about where the ceiling sits. Somewhere between +50% and +100%, the incremental dollar stops earning its keep.

I'll be honest about the messiness here, because it matters. The team running it stressed that the 100% result was uncertain enough that it could genuinely go either way on a re-run. They deliberately kept the baseline holdout tiny at 10% because a bigger holdout would have meant deliberately under-spending across a third of the country during the best week of the year, which is a costly thing to do on purpose. The cleaner design would've been three even thirds. They knew that. They traded some statistical clarity to avoid leaving money on the table.

That trade-off is the whole reason most brands can't just copy this.

Why you probably can't run this exact test

A geo holdout test like this needs scale and it needs a measurement partner. You're carving up a country, you're running it through an incrementality platform, and you've usually got a strategist on the other end helping you read a confidence interval so you don't over-trust a noisy cell. That stack costs real money every month.

If you're a brand doing a few million a year, that's a hard expense to justify for one read a year. And running the test on a category that's ten times bigger than your test category gets genuinely expensive, because the cost of deliberately holding back spend scales with the size of the thing you're holding back.

So most of you are not going to run a three-cell geo test this November. That's fine. The lesson still transfers. You just need a cheaper way to find your own ceiling.

The practical version: pre-set your tiers before the week starts

Here's what I'd do in your seat, and roughly what we set up for clients heading into peak.

Decide your scaling tiers in advance, off your normal daily spend. Don't improvise on the day with adrenaline running. Write them down a fortnight out.

Something like:

• Tier 1: +50% on your baseline daily spend. This is your default. You start here.
• Tier 2: +75%. You only step up to this if Tier 1 is holding profitably.
• Tier 3: +100%. You treat this as the exception, not the plan, and only if the numbers are still clearly green.

The reason I anchor the default at +50% rather than double is exactly what that geo test showed. For most brands, +50% is the band where you're confident the spend is pulling its weight. Doubling might be fine, but you've got to earn your way up to it rather than assuming it.

One nuance worth stealing: if you do step up a category that's much bigger or much more proven than the one you tested your nerve on, don't halve your old aggression. Halving is too blunt a cut. Trimming to something like 70-75% of what you used to do is usually the more sensible read, because a bigger, more established line carries demand that a small one doesn't.

Watch marginal CPA daily, not blended ROAS

The metric that tells you whether to climb a tier is not your blended ROAS for the day. Blended ROAS during BFCM looks amazing for everyone, because all that harvested demand flatters it. It'll happily tell you to keep spending right past the point where you should stop.

What you actually want to watch is the cost of the additional orders the extra spend is buying. Marginal CPA, roughly.

You can approximate it without fancy tooling. Take yesterday at your old baseline spend and the orders it produced. Take today at the higher spend and the orders it produced. The extra spend divided by the extra orders is your rough marginal cost per order. If that number is comfortably inside your break-even CPA, you've got room to climb a tier. If it's drifting up towards break-even, you've found your ceiling for the week. Hold there.

It's not lab-grade. Day-of-week effects and promo timing muddy it. But it's directionally honest in a way that blended ROAS simply isn't during peak.

Budget-flighting rules we use during peak weeks

A few simple rules keep this from turning into chaos at 11pm on Black Friday:

• Set the tiers before the week, change them only on data. No moving the goalposts because a competitor dropped a sale.
• Step up, never leap. Move one tier at a time and give it a day to read. Jumping straight from baseline to double is how you blow past your ceiling without noticing.
• A green blended number is not permission to climb. Only marginal CPA staying healthy is.
• Hold the line on supply too. Scaling spend into stock you can't ship is a worse outcome than leaving a sale on the table. The customer who has to wait usually still buys. The one who gets a botched order often doesn't.

That last one is the quiet trap. It's tempting to grab every possible sale during the surge, but you're not actually ahead if you burn down your supply chain to hit a bigger top-line number. Those rushed orders come at thinner margins and a service cost, while the same demand met calmly in January often converts at full margin. Sometimes the most profitable thing you can do during peak is pull back a notch.

Where to from here

You don't need a six-figure measurement stack to act on this. You need a number you trust for break-even CPA, three pre-set spending tiers off your baseline, and the discipline to only climb a tier when the marginal cost of the extra orders still clears that break-even.

If you want a clearer picture of where your real ceiling sits before the rush, that's exactly the kind of thing a Signal/Noise Audit digs into - we'll map your unit economics and your account history so you walk into peak week knowing which tier is your default and which one is a bridge too far. No pressure either way.

What's your honest read on last BFCM: did doubling the budget actually make you more money, or did it just make the revenue chart look good?