How predictable promotions teach customers to wait
2026-10-07
Any promotion that is predictable comes at a cost which is not shown in the promotion report. Namely, the losses from full-price sales during the weeks when customers have learned to wait.
- Ecommerce
- Pricing
- Promotions
- Retail Strategy
The promotion report is the most reliable document in the field of ecommerce; it displays the relevant figures such as sessions, conversions, units sold, revenue, and the increase compared to the corresponding days of the previous year. However, it does not include the two-week period prior to the promotion, during which some of those customers had already been on the site, seen the price and then decided to return later.
That fortnight is getting longer. Adobe's US holiday forecast, published on 28 September, expects $95.8 billion to be spent online in October alone, up 8%, with the October Prime Day event pulling discounting forward across the market. It also sets out the expected discount curve for the season: up to 14% off in early November, up to 21% just before Thanksgiving and up to 30% in Cyber Week. Black Friday is forecast to grow 9.2%, faster than Cyber Monday's 6.2%, and Adobe puts part of that down to discounts starting earlier. Customers can read that curve as easily as we can.
Below is what the evidence says about customers learning to wait, how large the effect is, and which promotional decisions make it bigger or smaller. Much of the research is older than I'd like, and I'll say where that matters.
What a discount leaves behind
Reference price is the price a customer carries in their head and compares against what they see. The research summarised by Mazumdar, Raj and Sinha in the Journal of Marketing in 2005 shows that people build it from prices they have seen before, discounted ones included. A product that has been 30% off twice this year has a lower reference price than its ticket suggests, and paying full price starts to feel like overpaying.
Frequency matters as much as depth. Kalwani and Yim's experiment in the Journal of Marketing Research in 1992 found that more frequent promotions raised people's expectations of future promotions and lowered the price they expected to pay. That paper is 34 years old and I've only been able to check it through how later studies cite it, so I'm leaning on it lightly. The more recent evidence below carries more weight.
The behavioural effect is simple to describe: if a customer expects the price to fall and has a rough idea of when this will happen, then they will delay their purchase. As a result, the commercial effect is clear in that during the weeks between the events the number of full-price conversions declines and the level of discount required to prompt a purchase increases from one season to the next.
Strategic waiting has been measured
Soysal and Krishnamurthi, in Marketing Science in 2012, built a demand model on sales data from a national specialty apparel retailer's coat range. Their estimate was that customers who time purchases around expected markdowns left the retailer with revenue about 9% lower than it would have earned from customers who buy as soon as they want the product. They found something useful alongside it: the risk that a coat might sell out before the markdown reduced the waiting. Scarcity, when it is real, pulls purchases forward.
The largest piece of evidence comes from Alibaba. Zhang, Dai, Dong and colleagues, in Management Science in 2020, analysed a randomised field experiment covering more than 100 million customers and 11,000 sellers. Customers in the treatment group were offered discounts on products sitting in their shopping basket. On the day, sales of those products doubled. Over the following month, the same customers browsed more and bought more often, which is the good news. They also added a higher share of the products they viewed to their basket and paid lower prices for what they bought, which the authors read as customers angling for more basket promotions and searching harder for low prices. The effect spilt over to sellers who had never offered the promotion.
Both studies are older than four years, so I looked for anything more recent that contradicts them and didn't find a field study that does. The most recent evidence is survey-based. In BCG's annual Black Friday study, fielded in September 2025 with more than 10,000 consumers in 10 markets, 77% said they sometimes or often delay purchases earlier in the year to take advantage of year-end sales. Stated intentions aren't behaviour, but it is hard to read that figure as anything other than a market that has learned the calendar.
How the cost shows up in a trading calendar
Imagine a retailer holding a 20 per cent sale across the whole website every five or six weeks, together with one on Black Friday. Each of these sales performs well since each one draws in demand that had been waiting for it. The periods between the sales are then quiet, which is why the following sale is scheduled to address those quiet periods, and as a result the cycle becomes shorter. Towards the end of the season, a greater proportion of the revenue is generated through promotions, the full-price sales of the current range have declined, and the clearance required to get rid of the remaining stock is greater than what the purchase plan had anticipated. There is nothing wrong with any of the individual event reports. The loss is instead contained between the reports.
Nike's results on 1 October are a useful, if uncomfortable, illustration of the reverse. NIKE Brand Digital revenue fell 13% in the quarter to August while gross margin rose 60 basis points to 42.8%. On the earnings call, as reported in published transcripts, management described its Greater China reset as removing unprofitable digital distribution and cutting promotional activity, with Greater China revenue down 26%. So far the numbers show the cost of stepping off the promotional treadmill: revenue goes first. The margin and brand benefit isn't in the data yet, and anyone quoting Nike as proof should say so.
Promotions that teach less
For nearly everybody, Black Friday remains on the calendar, and the important point is how much each event teaches, since three factors affect this.
The first method is range exclusion; in this case, if the new season ranges and the core lines never take part in sitewide events, then their reference price remains unchanged and customers discover that it is worthwhile to wait and buy the older stock. This ensures that full-price sales are achieved on the lines which have a margin. On the day of the event, the price appears as a smaller headline, which constitutes a more difficult discussion with the board than is appropriate.
The second point is predictability. Strategic delaying relies on customers being certain that a discount is coming and having an idea of when it will happen. If events are liable to change their date, alter their format, or appear without any regular pattern, then waiting becomes a risk. Soysal and Krishnamurthi's finding regarding stockout risk also shows the same thing: when customers think that the product might be out of stock before the price drops, some of them will buy at that moment. This only works if the scarcity is genuine. Countdowns that are invented and false messages indicating low stock constitute a regulatory issue under the Digital Markets, Competition and Consumers Act 2024, as well as a trust issue.
The third one is targeting, and in this case the evidence is less convincing. With targeted offers, the public price is left unchanged, so customers who see the full price on the site continue to have their reference price. However, the Alibaba experiment was a targeted offer and yet it still made the customers who received it more strategic for a month afterwards. Since targeting applies only to a smaller number of people, those people still picked up the lesson.
Measuring the waiting in your own data
The data necessary to draw this conclusion is already held by most teams. To see this, take the two weeks prior to each of the major events last year and compare the full-price conversion rates and full-price revenue per visitor with the corresponding weeks that didn't have an event scheduled. The more pronounced the drop as the event approaches, the waiting effect among your own customers is. Also examine the proportion of returning customers who only purchase on promotion within your active customer base and how that proportion has changed over the three-year period; if it continues to increase, then the calendar is conditioning your customers regardless of what the event reports indicate.
Both of these measures are not precise. Since seasonality and marketing spending occur during the same weeks, the comparison has to be handled with care. Although it provides a figure for the cost that the promotion report omits, this figure should be included in the room when this year's Black Friday sales figures are being decided, together with the number of events scheduled around it.