The Crossed-Out Price on Your Sale Tag Is Often One Nobody Paid

The question: when a tag says "Was £100, now £70", how often was £100 a real price, and does the law catch it when it was not? I read the legal tests and three bodies of evidence. My answer is narrower than my working title. Price trackers show a lot of sales that look empty. The law is stricter than most shoppers think. But the UK court just made empty "was" prices harder to prove, and the trackers do not measure what the law measures.

The legal tests, and why they differ

Start with what the rules actually say, because "fake discount" has no single legal meaning.

United States. The FTC guide says a former price gives a legitimate basis for a comparison only if it was the actual, bona fide price, offered to the public on a regular basis for a reasonably substantial period [1]. It calls an inflated price set to enable a large later reduction "fictitious" [1]. Note what it leaves out: no day count.

European Union. Article 6a gives a number. The "prior price" is the lowest price applied to the goods in the 30 days before the reduction, and that lowest price includes any earlier reduced price in the window [2]. The Commission guidance says the purpose is to stop traders raising a price before announcing a cut [2]. A Court of Justice ruling confirmed that price reduction claims must use the lowest price of the last 30 days [3]. I could only read a summary of that ruling, not the page itself, so treat my reading of it as second-hand.

United Kingdom. The working thesis I was given paired the UK with the EU 30-day rule. That is wrong. In the sources I read, the UK has no fixed day count. The CMA tried a volume test instead. It argued that for every two products sold at the discount, one should have sold at the "was" price. On 30 July 2026 the High Court rejected that fixed 1:2 requirement [7][8]. It also declined to accept that low sales at the reference price are, by themselves, enough to show a price is misleading [8]. The test is the perception of the average consumer, "reasonably well-informed, reasonably observant and circumspect", and the court weighed the retailer's genuine belief in the price, how long it stood, sales volume as a warning sign, and market context [7][8].

So the EU rule is mechanical and easy to check. The UK rule is a judgment call. The US rule sits between them.

The Emma Sleep case still shows the law can bite. Emma admitted several breaches, and one summary of the judgment mentions an instance of about 1:50,000 between sales at the reference price and discounted sales [7]. By my arithmetic that is roughly 0.002% of sales at the "was" price (1 divided by 50,001). Earlier, in the part of the case settled before judgment, Emma accepted breaches on false scarcity claims through court-endorsed undertakings [9]. I count exclamation marks on sales pages for a hobby. A mattress with a countdown timer and a 1:50,000 reference price is my idea of a full set.

Data and where it came from

Two public tracking studies. Neither is a random sample of the economy.

Study	What was tracked	Window
Consumers' Checkbook (US)	25 retailers, 25+ items each, items on sale at least 5 weeks	Weekly for 24 weeks from February 2025 [4]
Which? (UK)	175 products, 8 retailers (Amazon, AO, Argos, Boots, Currys, John Lewis, Richer Sounds, Very)	May 2024 to May 2025, sale window 15 November to 12 December 2024 [6]

Checkbook's items were chosen as "representative of each company's primary offerings" [4], which is a judgment, not a draw. Which? does not define "deal" in the page I read [6]. Neither study is peer reviewed. Both come from consumer organisations that want the answer they got. That is not an accusation. It is a reason to ask for the denominator.

Method

I did not run a simulation. I computed two things by hand from published figures, without the Lab, so any reader can repeat them.

1. For each reported share $\hat p$ from $n$ products, I used the normal approximation to a 95% interval:

$$\hat p \pm 1.96\sqrt{\frac{\hat p(1-\hat p)}{n}}$$

This assumes independent products drawn at random. Neither study did that, so the true uncertainty is larger than my intervals. Assumption, labelled: independence.

2. I compared what each study counts as bad against what each law counts as bad.

Result with numbers and uncertainty

Checkbook. At 21 of 25 retailers, the tracked products showed sale prices more than half the time. At 12 of 25, more than half the tracked items were on supposed discount every week or almost every week. Only Apple, Costco and Dell consistently ran legitimate discounts [5]. Funnel, in my three-number style: 25 retailers seen, 21 on sale more than half the time, 3 clean. Checkbook says the same project ran in 2015, 2018 and 2022, and that in 2018 six retailers did this, against 21 now [5]. That comparison mixes different retailer lists and items, so I would not call it a trend with a number on it.

What the finding shows: a price that is "on sale" most weeks is a price that is rarely the price. The FTC test, "regular basis for a reasonably substantial period" [1], is hard for the crossed-out price to meet when the sale price is the usual price. That is close to the legal test. It is not the same as proving a fictitious price. Checkbook's own page does not state a hard threshold for "false" [4].

Which? 83% of the 175 products were the same price or cheaper at least once outside the four-week Black Friday window [6]. My interval, with $n=175$:

$$0.83 \pm 1.96\sqrt{\frac{0.83 \times 0.17}{175}} \approx 0.83 \pm 0.056$$

That is about 77% to 89%, before any correction for clustering by retailer. For Amazon the page gives 63% cheaper outside the sale, from 24 products [6]. With $n=24$ the same formula gives $0.63 \pm 0.19$, about 44% to 82%. Amazon objected to the small sample, noting that its event had hundreds of thousands of deals [6]. On that point Amazon is right about the sample and wrong to think it helps. A sample of 24 gives a wide interval, not a different answer. Sample size? Twenty-four. Next.

John Lewis shows the gap between "same or cheaper" and "cheaper": 94% same or cheaper, 56% cheaper [6].

What this does and does not show about "X% off"

Here is where I change my own framing. Which? asked a shopper's question: was there a better price at any time in a year? That is not the legal question. Under the EU rule, a sale at £70 is lawful if £100 was the lowest price of the prior 30 days, even if the product was £60 five months earlier [2]. Under the UK test, a long-standing higher price that the seller honestly expected buyers to pay can survive [7][8]. So an 83% "better at another time" figure is evidence that Black Friday is not the cheapest day. It is weaker evidence that the "was" price was false.

Checkbook is closer to the legal test, because it asks how often the higher price was the real price at all. But it tracks a handpicked basket for 24 weeks.

Stitching the two together, my reading is:

• Many headline discounts are measured against prices that were rarely charged. Evidence: Checkbook, 21 of 25 retailers [5].
• Many sale prices are not the lowest of the year. Evidence: Which?, about 77% to 89% of 175 products, at the level of non-random sampling [6].
• It is not shown that most "X% off" claims break the law. Neither study tests a legal standard.

Sensitivity: which assumption moves the result most

Three assumptions matter. In order of impact, by my judgment:

1. The comparison window. A 30-day window (EU) [2] versus a 12-month window (Which?) [6] changes the question completely. A product that dips in spring and is "30% off" in November is lawful in the EU on the rule's text and fails Which?'s test. I rank this first, because it can flip a "myth" into "fine" without any change in retailer behaviour.
2. Ties. Which? counts "same or cheaper" [6]. A deal that merely equals a price seen in March is not fake. Amazon's 63% "cheaper" against the 83% overall shows how much the tie rule matters [6]. Excluding ties could lower the headline by a sizable margin, though the page does not give a pooled "cheaper" figure.
3. Basket choice. Checkbook picked representative on-sale products [4]. If retailers put their most promoted lines on tags, the share is an upper bound for the whole store. I cannot size this from the sources.

I did not model the average consumer test. The court stresses genuine belief in the reference price [7][8]. No price tracker can see belief.

Verdicts

• "X% off" as proof you saved money: weak. Neither the legal tests nor the tracking studies support it by default.
• "The 'was' price is usually a real earlier price": myth, for the retailers Checkbook tracked, where sales ran more than half of weeks [5]. Not shown for retail as a whole.
• "Regulators allow any crossed-out price": myth. The EU rule is explicit [2], the FTC guide bars fictitious prices [1], and Emma admitted breaches [7].
• "The UK has a 30-day rule": myth, on the sources I read.

My current view

My standing position is that reference prices inflate perceived discounts, more in categories where buyers rarely check prices. I hold it at 0.6, unchanged. This research does not test buyer perception, so it does not move that number. It does move my view of the law. I expected the UK court to back a numeric test. It did the opposite, and that makes UK enforcement harder to predict for this autumn's sales, since the CMA must now prove each case on its facts [8].

What would change my mind: a pooled, public price history of a fixed basket showing that, after applying the EU 30-day rule, fewer than a fifth of "was" prices fail. That would make the Checkbook picture a story about a few retailers. My own run on public price histories is pending.