Black Friday 2026: were the discounts real? — the methodology, committed in advance

The question

For each tracked product advertised at a discount during the event: was the sale price actually below what the retailer charged in the 90 days before the sale — and where a “was” price is claimed, did we ever observe the product selling at it?

Data source

WEM price_history — scheduled samples of each retailer's own displayed price. Prices are sampled, not watched — the study can prove a price was charged, never that a price was not. Every claim will respect that asymmetry.

The evidence bar — per product, no exceptions

• At least 20 recorded observations in the 90-day pre-sale window.
• Observations on at least 50% of the window’s days — volume alone can lie; twenty samples in one week say nothing about the other eleven.
• A product below either threshold is scored “insufficient history” and excluded from every headline figure — never counted as a fake, never counted as genuine.

For comparison, WEM’s own product pages act on 3 observations. The press bar is 20 because the cost of error here is a newspaper telling readers a named retailer faked a discount: the bar is what survives that retailer’s press office reading it.

Verdicts

Genuine reduction

The sale price is below every price we recorded in the window.

Matches a prior price

We recorded the same price or lower before the sale began — reported with the date and price observed (“it was already £X on 4 October”).

Price never moved

One distinct price across a properly covered window — a deal in name only.

Insufficient history

Excluded from the denominator. The honest default.

Where a retailer displays a “was” price, we will additionally report on how many observed days the product sold at or above it. UK guidance expects a reference price to have been the genuine selling price for a meaningful period; a single day is the shape that fails.

What we commit NOT to do

• No invented or extrapolated percentages: every headline number is computed from recorded observations that we can produce on request.
• No naming a retailer on evidence below the bar above.
• No counting “insufficient history” toward either side of the story.
• If the corpus does not clear the bar for a publishable sample by the publish window, we publish nothing and say why. Silence beats a wrong number.
• Right of reply: any retailer named in the study is contacted before publication.

Verify this yourself

The scoring rules live in WEM’s codebase (src/lib/press/price-reality.ts), and per-product evidence is visible on the public history pages (e.g. a live example). Journalists can request the underlying observations for any product in the study: [email protected].