Methodology

How a verdict is reached, in enough detail that you could disagree with it on the evidence.

What we record

Every day, for every tracked listing, we record the selling price, the “original” price advertised beside it if there is one, whether the item is in stock, the URL we read it from, and the time we read it.

We take data from the retailer’s own structured sources — the public product JSON that shop platforms publish, or the schema.org product data retailers provide for search engines. Both are maintained by the retailer for their own purposes, which makes them far more reliable than reading prices out of page layout.

Each observation keeps its provenance: which source it came from, the final URL after any redirects, and a checksum of the exact data we were served.

When a verdict appears

A verdict is a claim about a listing’s own past, so it cannot exist before that past has been recorded. Until a listing has fourteen recorded days within the last ninety, no verdict is produced for it at all — that floor is enforced in the scoring code, not left to editorial judgement. Those listings are still published; what they carry instead of a verdict is the length of our record.

Above the floor, what a listing can be measured against grows with its record. A claim is not set against a published standard until our record covers that standard’s window: thirty days for the EU test, three months for California’s. Every published verdict carries the number of days behind it, so you can weigh it yourself.

Where that leaves us today: recording began 10 August 2026, and we have 2 days. A listing reaches a verdict when its own record clears the floor, and shows a day count until it does.

The four measures

Once a listing clears that floor, four measures are computed from its own history and combined into a score out of 100. The weights are fixed, and shown beside each measure.

1. Reference-price residency (weight 40)

On how many of the tracked days was the claimed “original” price actually being charged? This is the measure that matters most, and it is not our invention — it is close to the statutory test. California requires a former price to have been the prevailing price within three months; the EU requires a reduction to be measured against the lowest price of the previous thirty days. We work in 30-, 90- and 365-day windows to match those tests, and a statute is named on a listing only once our record covers the window that statute asks about.

The tolerance runs in the retailer’s favour: a claimed original of $2,499 counts as “charged” at $2,486.51 or above, half a percent below the figure claimed.

2. Discount against reality (weight 30)

Today’s price against the median of every price we have recorded for that item in the last twelve months — not against whatever figure the retailer prints beside it. Early in a listing’s record that median is drawn from weeks rather than months, which is why the day count is published beside the verdict.

3. Distance from the floor (weight 20)

Where today sits between the item’s typical price and the lowest price we have recorded for it so far, and how long ago that floor was last touched.

4. Retailer record (weight 10)

How often this retailer’s past claims have held up against our archive. Every retailer starts neutral here and stays neutral until we hold ten assessable claims for it: below that the sample is too small to say anything.

What the badges mean

The top-level verdict is decided by residency alone, not by the combined score. That is deliberate: whether a claim is honest and whether today’s price is good are two different questions, and blending them lets a genuinely low price launder a fictional reference price into a clean bill of health.

Fictional

We never observed the claimed reference price being charged during the window. This says nothing about whether today’s price is good: a product can carry a fictional reference price and still be at its lowest price in a year, and where that is true we say so on the same page.

Small saving

The reference price is real — we saw it charged — but the saving against the item’s own typical price is small. This is the one decision the combined score makes: among claims that have survived the residency test, a score below 65 out of 100 is graded a small saving rather than a genuine one.

Genuine

The reference price is real and the saving is real — 65 or above on that score.

N days recorded

No verdict yet. In its place the badge reports how long we have recorded that listing; the number climbs by one every day we record it. What a verdict needs is set out above.

No claim

The retailer is not advertising a discount on this listing today, so there is no reference price to check.

Language

We publish observations against a published standard. We do not publish accusations.

Known limitations

• We record advertised prices, not checkout totals. Cart-level automatic discounts and code-gated offers are not visible in the data we read. What we audit is the discount claim a retailer publishes, which is the right target — but it is not necessarily what you finally pay.
• Not every retailer publishes a reference price in a form we can read. Some run sitewide percentage promotions that never appear as a strikethrough figure in structured data. Those products still build price history, but they cannot carry a residency verdict.
• Some retailers block automated access. We respect that and record them as out of scope rather than working around it, which means our coverage is not the whole market.
• Our sample is not evenly weighted. Some retailers list far more products than others, so a figure aggregated across the whole archive can be dominated by a single company. We report findings per retailer for this reason, and we will say so plainly wherever an aggregate is unavoidable.
• A day missed is a day lost. Price history cannot be reconstructed after the fact, and we will not estimate one. Where our record has a gap, we show the gap.

Corrections

If you are a retailer and believe a figure we have published is wrong, write to contact@discountverdict.com with the product and the date. We will re-examine the underlying record, correct anything we got wrong, and say publicly that we did.

Last updated 10 August 2026.