Promotion Economics: Break-Even Uplift and Black Friday P&L Tool

Download the workbook: an Excel tool, pre-filled with a fictional retailer’s Black Friday. No form, no email, no gate.

Most promotion reviews report one number: sales during the event, against the same days last year. It is almost always up, and it answers none of the questions that decide whether the promotion should run again.

A discount gives away margin on every unit, including the units that would have sold at full price anyway. It borrows sales from the weeks after. It takes some sales from lines that were not on offer. It may clear stock that would otherwise have needed a deeper markdown later. A promotion pays only if the extra volume covers all of that, and the costs of running it.

The questions

Four questions every promotion must answer

What would have sold anyway?

The counterfactual. Without it, the whole of the event’s sales gets credited to the promotion, including the sales the season would have delivered on its own.

What margin did the discount give away?

The discount on sales that would have happened anyway is a straight cost. It is the largest number in most promotion P&Ls, and the one most reviews leave out.

What did it borrow from the weeks after?

Customers who would have bought in December at full price bought in November at a discount. The event looks better, and the weeks after look worse, and nobody joins the two up.

What did it clear, and what did that save?

Aged stock sold at 25% off is worth more than the same stock sold at 40% off in January. That saving is real, and it belongs in the verdict.

From the field. At one retailer I was told Black Friday was only for electronics. We ran it anyway, off site, through email and affiliates. The sales came, and Commercial claimed them. Without a counterfactual every function can claim the uplift, and usually does. The argument ends when there is an agreed answer to the first question.

Before the event

The uplift a discount needs just to stand still

Break-even uplift is the discount divided by the contribution margin left after the discount. Contribution margin here is gross margin less the variable cost of serving the order, before the discount. The answer is the extra volume needed to make the same contribution as not running the promotion at all, before a penny of media.

Discount	40% contribution margin	50% contribution margin	60% contribution margin
10% off	33%	25%	20%
20% off	100%	67%	50%
25% off	167%	100%	71%
30% off	300%	150%	100%
40% off	Never pays	400%	200%

At 25% off a line with a 50% contribution margin, volume has to double to stand still. At 40% off a line with a 40% margin, no volume is enough. The workbook carries the full grid, from 10% to 50% off and from 30% to 70% margin.

The grid is the floor. The Pre-Event Plan adds media, creative, supplier funding, the share of the extra volume borrowed from the weeks after and the sales lost on lines not in the promotion, and gives the uplift needed with every cost counted. In the sample, the grid says 109%. With every cost counted, it is 225%. The plan assumes 110%, which clears the grid and loses £18,306.

The plan also checks the stock. Divide the units on the promoted lines by the planned daily rate: in the sample, the stock runs out on day four of four. The last day sells what is left, not what customers wanted, and the media keeps paying for traffic to broken lines.

Stop conditions are agreed before the event starts, with a threshold and an owner each: best-seller availability, contribution per order, checkout completion, the delivery promise and early returns reasons. A promotion that is going wrong is rarely stopped by the person who planned it, unless the rule was written down first.

From the field. One launch offer added a free service to every order. The service had no virtual stock set up in the ERP, the order management system or the platform, so checkout could not complete an order that contained it. It took three hours to fix, on launch day. Test the mechanic end to end before the media goes live, not after.

After the event

What would have happened anyway

Enter totals for three periods, before, during and after the event, for the promoted lines and for everything else: orders, units, full-price value, discount, returns, cost of goods and new customers. The workbook builds the counterfactual from the daily rate on the promoted lines before the event, scaled by the season.

The season is measured one of two ways. Control lines uses the lines not in the promotion: if they traded 15% above their earlier rate on the event days, the promoted lines would have too. Manual factor takes a lift you enter, from a year without the event or from your own forecast, and lets the workbook measure the halo or cannibalisation on the rest of the range. A holdout, where a region, a segment or a set of lines is deliberately left out, is better evidence than either. Where you have one, enter it as the control lines.

The Promotion P&L then runs expected against actual, line by line: orders, full-price value, discount, returns, net sales, cost of goods, variable costs, contribution, then the promotion’s own costs net of supplier funding.

The weeks after

Volume up 236%. It still lost money.

From the workbook’s fictional retailer, a four-day Black Friday at 25% off:

+236%

Volume against the counterfactual

£7,020

Made in the event

£11,817

Given back in the fortnight after

A sales report would call this the best event of the year. It needed 194% more volume to cover the discount, and got 236%. After media and creative, the event days made £7,020. The following fortnight then ran £11,817 below what the season would have delivered, because customers had already bought. Net, the promotion lost £4,797, and gave away £20,240 of margin on sales that would have happened anyway.

The total also hides the mix. During one peak event I ran, a morning read showed sales up 6%, orders up 46% and average order value down 27%: customers were buying a different, cheaper mix from the one the stock had been bought for. The Golden Quarter trading kit covers the event days themselves. This workbook covers the reckoning afterwards.

The verdict

Five verdicts, one of them uncomfortable

Verdict	When	Now what?
Scale or repeat	It made money, and gave back less than half of it afterwards	Repeat it, and test whether a little more depth or range pays as well
Repeat, smaller	It made money, but more than half the gain was borrowed from the weeks after	Run it shorter, shallower or narrower
Fix	It made money, but best-seller availability fell below the threshold, or checkout failed	Repeat it only once the leak is fixed
Clearance only	It lost money, but the markdown it saved on aged stock more than covers the loss	Run it again as a clearance tool on aged lines, not as a traffic event
Stop	It lost money, and cleared nothing that saved more	Stop it, or redesign it: it moved sales without making money

The sample lands on Clearance only. Seven hundred aged units sold at 25% off would otherwise have needed 40% off in January, which saved £8,400 of markdown and more than covers the £4,797 lost. That is a defensible reason to run the event. It is a different reason from the one most Black Friday plans give, which is traffic.

Each verdict comes with What? So what? Now what?, written out in sentences on the Overview, with an owner. The workbook recommends; the business decides.

From the field. At one business, a dash for cash ended in a promotion of 20% off everything. Sales jumped. Availability broke, because the products were built from components: one sold-out component took several products offline while the rest of their stock sat in the warehouse. That is the Fix verdict. The money was there, and the business could not take it.

Controls

Checks and testing

Ten checks run on the data: every period has days, the promoted lines traded before the event, the control lines traded in every period, discounts and returns no larger than the sales they belong to, manual factors set when the manual method is chosen, shares between 0 and 100%, a planned discount that can pay at all, planned stock that covers the event, and event availability recorded. The workbook has been tested in 20 scenarios, from a discount deeper than the margin to a baseline with no sales, and through every verdict. Every cell you should not change is locked.

The series

With the other tools

The Ecommerce Budget Builder carries event weeks on their history, and says plainly that it cannot tell whether a promotion created sales or merely moved them. This workbook is where that question gets answered. The Ecommerce KPI Framework puts discount depth on the SLT dashboard as a decision metric; this is where the depth of each event is decided before it runs and judged after it. Availability and Stock Intelligence tells you which lines need clearing and what the markdown would cost, which is the number the clearance verdict depends on.

Limits

What it does not do

The counterfactual is an estimate. The control-lines method assumes the promotion did not move the other lines. If it did, use a manual factor, or better, a holdout.
It does not measure what customers learn. A business that promotes often trains customers to wait. The weeks after capture some of that, not all of it.
It works in totals, not lines. Run it once for the whole event, or once per category if the mechanics differed.
The second order is optional. New customers from an event often buy again less often than customers acquired at full price. Enter the 90-day repeat rate when you have it; the verdict does not wait for it.

From the field. One business bought another retailer’s range as a job lot at distressed prices and traded it for three years. It sold through. Customers learned to expect the prices, and when the stock ran out, the commercial model was broken. No promotion P&L captures that on the week it happens. It shows up in the years after.

Sources

Where this comes from

The method and the episodes come from planning and reviewing promotions across fashion, furniture, home and electronics since 1999. The figures in the workbook belong to a fictional retailer and describe no real business or benchmark.

If last year’s Black Friday review reported only sales, run it through the workbook before you sign off this year’s depth. If the answer is Stop, that is worth a conversation before the media is booked.

Back to Thinking