Black Friday Discount Depth: The Loss You Book in Advance

Break-even ROAS is 1 divided by contribution margin. Every point of Black Friday discount comes straight out of the margin, so every point raises the ROAS a channel has to clear before it makes money. On a product with a 50% contribution margin at list price, a 30% discount moves break-even from 2.0 to 3.5. That loss is booked when the discount is chosen, before the first ad is served.

Loss and Avoidance is the strongest of the Octalysis black hat drives, and the useful thing about this one is that it is entirely arithmetic. No vendor, no data, one table.

The table

Assume variable cost (goods, shipping, payment, packaging) of 50 on a list price of 100, so contribution margin at list is 50%. Discount the price and the cost stays where it is:

Rebuild it with your own cost line. The shape does not change: the last ten points of discount cost more than the first twenty, because the margin they come out of is already thin.

Why the comparison in the plan is wrong twice

Most Black Friday plans compare a channel's reported ROAS against last year's target, and last year's target was set at last year's discount. If the discount deepens from 20% to 30%, the target on the same product moves from 2.67 to 3.50, and a channel that "beat target" at 3.0 has lost money. Tariffs changed your unit economics, not your ROAS targets covers the same mechanism for landed cost; the discount version is worse because it is chosen, and chosen late.

The second error compounds it. The ROAS being compared against the target is the platform's, and The Price of Being Found spends a chapter on why the platform's figure overstates: it counts conversions it can associate with its own inventory inside its own window, and the sum across platforms exceeds the orders shipped. So the plan compares an overstated return against an understated target. Both errors point the same way.

The number that survives the discount

Contribution margin is a fact from the store and the cost sheet. Break-even ROAS follows from it. The only remaining unknown is the channel's incremental return, and that is the number the plan needs and the platform cannot supply. A holdout that starts by 2 October answers it for one channel with an interval. A causal read on the GA4 export estimates it for every channel from the data you already hold, with intervals, so the comparison in the plan is incremental ROAS against break-even at this year's discount, not reported ROAS against last year's.

The reference price is part of the same arithmetic

In the EU the "was" price on the banner has to be the lowest price applied in the previous 30 days, which for a 27 November reduction means from 28 October. A flash discount in early November lowers the reference and shrinks the advertised reduction; a deeper discount to compensate deepens the margin loss in the table above. The EU line on Black Friday urgency has the rule. The pricing calendar and the margin table are the same decision.

What to do this week

• If you own the budget: rebuild the table with your real variable cost, put this year's planned discount in it, and read off the break-even ROAS. That number replaces last year's target in every channel line.
• If you have to defend it: ask which figure the plan compares against break-even. If the answer is the platform's ROAS, bring the claim ratio for last Cyber Week to the same meeting.

The Black Friday 2026 calendar has the dates this arithmetic hangs on.

Arithmetic as of 9 September 2026 on an illustrative cost structure; substitute your own. The platform overstatement mechanism is from The Price of Being Found (Edition 2.10), Chapter 9.