Most Discounts Lose Money, and 75% of Marketers Can't Tell Which Ones

A Kantar and Les Binet survey of 250 senior UK consumer goods leaders, presented at the IPA Effectiveness Conference on October 7, 2026, found that 75% could not estimate what share of their price promotions made money last year. Promotions now take 36% of marketing and trade spend, about double paid media's 17%, and the literature the study cites suggests roughly two-thirds of them lose money.

So the biggest line in the budget is also the one almost nobody can grade. That's the headline, and I don't think it's only a UK grocery problem. The full breakdown on PPC Land walks through all 33 slides, and if you run ecommerce or DTC with a Black Friday calendar getting locked in right now, a lot of it reads like it was written about you.

Dom Boyd, Kantar's global chief strategy officer, put it plainly in Research Live's write-up: "Many are measuring promotions on gut feel or partial analysis, not really knowing what the impact on profit is." Binet went further and called price promotions "like class A drugs: expensive, dangerous and addictive." A bit dramatic. Not wrong, though.

82% think their last promo paid off, which is the scary part

This is the bit that stuck with me. 82% of respondents said their most recent significant promotion was profitable (28% clearly, 55% marginally). And per the IPA's own summary, more than 60% of those people based that verdict on personal judgement or partial analysis, not event-level profit and loss data.

So you have a group of senior people who mostly can't say what share of promotions work, yet mostly believe the last one did. And 76% said their company repeats similar promotions often or sometimes even when profit was never assessed. The promo runs because it ran last year, which is a pretty normal way for budgets to work, honestly. It's just expensive when the line item is a third of the money.

In fairness to the respondents, the deck has holes. Several headline figures come from subsets of the 250, no margin of error is published, and the 36% spend figure blends sources from the UK and US. PPC Land flags a couple of slides where the math doesn't even add up. I'd treat the exact percentages loosely. The direction, though, matches about forty years of academic work, and that part is hard to argue with.

Where the "extra" sales actually come from

The deck includes an illustrative model built from the promotions literature (Van Heerde, Dawes, Ailawadi and others). A promotion lifts sales from 100 units to 150. Of those 50 extra units:

• 22 are genuinely incremental
• 15 are time-shifted (people who would have bought next month, buying now)
• 9 are cannibalised from your own other products
• 4 moved over from other stores or channels

That puts real incremental sales at roughly 15% of everything sold on promotion. The other 85% is mostly people who were going to buy anyway, getting a discount for it.

Now add margin. Say you sell a $50 product at 40% gross margin, so $20 profit a unit. Run 20% off and your margin drops to $10 a unit. To make the same gross profit during the promo window, you need volume to double. Not rise 50%, double. And that's before you subtract the 15 pull-forward units that will show up as a dip in the weeks after.

Think of it like a gym that offers half-price January memberships and counts every sign-up as new revenue, including the people who were already paying full price and just renewed early.

Most promotions don't double volume.

Which is roughly why the literature lands on "around two-thirds are unprofitable."

The 13-week window is too short, and most teams use shorter

When asked about their longest evaluation window, 2% only looked at the promotion period itself, 10% looked at four weeks and 27% looked at eight. Only 19% went to 26 weeks or more. The deck says 81% evaluate over 13 weeks or less.

That matters because the costs of discounting seem to show up late. 75% of respondents agreed repeated promotions raise price sensitivity, and Kantar cites Worldpanel and BrandZ data suggesting brands that promote repeatedly end up about 1.2 times more price elastic. You won't see that in an eight-week readout. You'll see it next year when the full-price weeks get softer and nobody connects it back. It's a similar trap to what Google's Gary Illyes described for SEO, where recovery takes 3 to 6 months and teams keep judging fixes at week two.

There's a related piece of research PPC Land covered from TransUnion and MMA Global, suggesting brand campaigns get undervalued by short-term measurement. Same disease, different symptom. Short windows flatter whatever produces a spike.

Google starts advertising your discounts today

This one's timely in a slightly annoying way. Starting October 12, 2026, Google Ads automatically extracts discount offers from your website and attaches them as promotion assets in Search and Performance Max campaigns that have linked location assets. Advertisers are enrolled by default. Campaigns that already have manual promotion assets are excluded.

On paper, that's helpful. More eye-catching ads, less setup. But if you can't currently say which of your promotions make money, having an algorithm pick which discounts to push to a wider audience probably isn't the fix. The opt-out is account-level, under automated assets settings, then Automated Promotions. I'd at least look at it this week, and personally I'd switch it off until you know which offers you actually want amplified.

The consumer side isn't helping either. An Adtaxi survey of more than 1,100 US adults found 40% had switched to lower-cost brands and 49% compared prices more often. Shoppers are trained to wait for the deal, and every unmeasured promo trains them a little more.

A holdout you can run before the BFCM calendar locks

You probably can't commission econometrics in the next three weeks. You can run a rough incrementality check on one promotion, which is more than 75% of the surveyed marketers can say. (PPC Land has a decent incrementality explainer if the concept is fuzzy.)

The version I'd use:

1. Pick one promotion that's already on the Q4 calendar. Ideally one that repeats, since that's where the 76% habit lives.
2. Hold some of your audience out. For email or SMS offers, withhold the code from a random 10% of the list. For paid or site-wide promos, pick two or three comparable regions or stores and don't run the offer there. Match them on last year's sales for the same weeks.
3. Measure gross profit, not revenue. Compare profit per customer (or per region) between the promo and holdout groups during the promo.
4. Keep measuring for six to eight weeks after. The pull-forward dip lands here. If the promo group buys less afterward than the holdout, subtract it.
5. Set the bar before you look. At 40% margin and 20% off, the promo group needs roughly twice the holdout's unit volume just to break even. If your result is a 30% to 50% lift, which is common, the promo probably lost money.

It won't be perfect. Regional holdouts are noisy, and a 10% email holdout on a small list might not reach significance. But a noisy measurement beats gut feel, and gut feel is apparently what most of the industry is running on. If you use promo codes to attribute sales anywhere, the same caution applies; we wrote about how promo codes miss most sales and credit some they never earned.

Promotions aren't the villain here, blind repetition is

Worth saying, because the Binet quote makes it sound like an abolition argument and the deck isn't one. It lists cases where promotions work: recruiting genuinely new buyers, clearing a defined stock problem. Stock clearance was the stated goal for 17% of respondents' last big promotion, and that's a perfectly sane reason to discount. The trouble is the 33% who said manufacturer profit and revenue was the goal, without the data to check.

And 68% said they'd run fewer promotions if they could do it without losing retailer support or position. So people mostly know. They're stuck in the habit, the targets and the "we did it last year" logic. I'm not sure a survey changes that by itself.

My guess: when the 2026 BFCM numbers come in, fewer than 1 in 5 brands will have measured their promotions over more than eight weeks, and plenty of them will plan 2027 by copying this year's calendar again. If you can bring one real holdout result to that planning meeting, you'll probably be the only person in the room with one. That seems like a decent reason to try it on just one promotion.