Check the starting price behind a discount

We kept the 3 currency groups separate. In the EUR group, 9 supplied references matched the earlier amount, 5 were higher & 4 were lower. In the NOK group, 47 matched, 5 were higher & 1 was lower. In the SEK group, all 41 matched. These are counts within the selected observations, not estimates for the places where those currencies are used.^[1]

When the supplied reference & earlier amount were equal, substituting one for the other left the percentage calculation unchanged. That explains why a reference-based reduction can also describe the change from that particular earlier observation in those cases.

The median gap was zero in all 3 groups. A median describes the middle of the observed results; it does not mean every comparison agreed. The EUR group had an equal split between matching & different reference amounts, while NOK had 6 differences & SEK had none. Keeping those counts beside the median makes the result easier to interpret.

We also retained differences in both directions. A supplied reference below the earlier amount produces a smaller percentage reduction than using that earlier amount. The task is to identify the comparison that a particular line of copy is making.

The 5 NOK cases had the same simple relationship: a higher supplied reference, with the current amount equal to the earlier observed amount. For each, the reference-based percentage was positive & the change from the earlier observation was zero. These were observed pairs in the selected records.

They occurred at 2 locations, with 3 comparisons at one & 2 at the other. This matters because a single location could otherwise account for the entire observation. It still does not establish separate ownership, independent pricing policies or a common practice beyond these cases. The source identifies physical locations, not the business decisions behind their prices.

The earlier observation was a dated point, at most 28 days before the selected SALE-coded record. Seeing the same amount at both points does not show that the price stayed unchanged between them. It leaves room for intervening prices that were not observed. Nor did the first SALE-coded record tell us when a promotion began.

For copy review, the distinction is narrower & more useful than a judgment about the offer. A reference-based percentage cannot, by itself, establish a particular change since the earlier observation. A merchant may have evidence for another comparison; the creator needs that evidence & meaning rather than an assumption drawn from the percentage alone.

A reference-based reduction asks how far the current amount sits below the supplied reference. Subtract the current amount from the reference, then divide by the reference. A change from an earlier observation asks a different question: subtract the current amount from that earlier amount, then divide by the earlier amount. Multiplying by 100 expresses either answer as a percentage.

The calculation is familiar; the choice of denominator carries the meaning. Calling a supplied reference an earlier price adds a claim about the product’s history. A date attached to one earlier observation adds useful context, but it still does not make that amount a verified regular price or prove that it was available continuously.

For a hypothetical example, imagine a product with an earlier observed amount of 80 units. A later SALE-coded record also shows 80 units, alongside a supplied reference of 100 units. The reduction against the reference is (100 − 80) ÷ 100 = 20%. The change from the earlier observation is (80 − 80) ÷ 80 = 0%. The 2 answers are 20 percentage points apart.

The invented 80- & 100-unit amounts explain the calculation; they are not prices from the dataset. In the measured NOK records, the gap between the 2 calculations ranged from −1.6 to 37.6 percentage points, with a median of zero. The largest gap is a case in this selected set, not a usual discount or an estimate for Norwegian shops.

A creator does not need to publish the formula in every caption. The person approving the claim does need to distinguish the reference comparison from a dated change. The final wording should preserve whichever meaning the merchant can support instead of silently converting one into the other.

Start the handoff with the product & offer, then attach the comparison. The merchant supplies the basis for the price statement; the creator carries that meaning into the words, visuals & delivery format. This is a practical recommendation from the observed distinction, not a tested claim about approval speed or campaign results.

The product description needs enough detail to identify the item being discussed, including the relevant variant or pack & the location or channel covered by the offer. A barcode match was our research convention; it did not prove a product’s historical pack identity. In an actual brief, the merchant can supply product details that this dataset could not establish.

Keep the current amount beside its currency & the date it was checked. Put the comparison amount in a separate field, with a plain explanation of what it represents. If it is a dated observation, retain that date. If it is another kind of merchant-supplied reference, name that basis rather than describing it as a recent observed amount.

Give the dates separate jobs. The check date identifies when the current amount was confirmed. An observation date identifies one earlier point in the evidence. An offer period describes when the merchant says the offer applies. Copying one date into another field changes what the brief says. In particular, the first sale-labelled record in a dataset does not supply a promotion’s start date. Keeping the dates distinct makes it easier to ask the merchant for the one fact that is missing.

Use the following handoff as a copyable starting point. It keeps the price statement, supporting material & approved wording together without requiring the creator to reconstruct the comparison from a graphic.

Product & offer: [product, variant or pack, location or channel, currency].

Current amount: [amount], checked on [date], for [offer period or availability confirmed by the merchant].

Comparison: [amount] means [plain description of the benchmark]. If it is a dated observation, the observation date is [date].

Evidence: [attached source or named merchant confirmation] supports [the specific comparison being claimed].

Offer conditions: [eligibility, quantity, membership or other applicable terms supplied by the merchant].

Wording to approve: [exact percentage statement & its named comparison]. Merchant approver: [name & approval date].

A blank comparison field is useful information. Return the percentage line for clarification rather than filling the gap with “previous price” or a date that the supplied material does not establish. An earlier screenshot can support what was observed on that date; it cannot supply every missing detail about an offer.

Watch the verbs around the percentage. “Compared with” identifies a comparison; “has dropped since” asserts a change over time. A merchant’s evidence may support the first statement without supporting the second. A creator can preserve the confirmed comparison without adding a story about the intervening period. If recent price movement is essential to the campaign’s message, ask for evidence that addresses that movement rather than substituting a reference field or a single earlier observation.

Read the approved sentence once without its image. Then read it beside the price graphic. Check that both make the same comparison, use the same amounts & preserve the relevant conditions. If the graphic uses one benchmark & the caption implies another, revise the handoff before treating the wording as settled.

This check is especially useful when one person prepares the offer & another adapts it for several posts. The percentage can remain numerically identical while its surrounding words change the claim. Carrying the named comparison through each version gives the creator a concrete fact to preserve.

The EUR comparisons came from 3 locations, but 16 of the 18 belonged to one. All 5 cases with a supplied reference above the earlier amount came from that location. Removing it left 2 comparisons & no positive gaps. The EUR result therefore cannot establish that this relationship appeared across several locations.

The NOK comparisons came from 5 locations. Removing any one still left at least 2 of the 5 cases in which the current amount equaled the earlier observation while the reference was higher. Removing the largest location left 38 comparisons across the other 4. These checks preserve a finite counterexample beyond one location; they do not turn the records into a representative sample.

The SEK comparisons came from one location. All 41 references matched the earlier amount, & removing that location left no comparisons. This is a matching result within one observed setting. Its concentration deserves the same visibility as the differences in the other groups.

We also repeated the summaries using only the earliest eligible comparison for each barcode–location sequence. EUR stayed at 18 comparisons & NOK at 53. SEK fell from 41 to 39, with all remaining references still matching. This check reduces the weight of repeated sequences; it is not another independent study.

The dates were concentrated too. The usable EUR observations were in April; NOK observations were in June, August & September; SEK observations were in April, June, August & September. The April–September eligibility window did not produce a continuous six-month record in each group.

We used a pinned Open Prices snapshot maintained by Open Food Facts, downloaded on October 23, 2026. Open Prices collects product-price observations; it is not a panel designed to represent every shop or promotion. We selected product records backed by price-tag proof metadata & an explicit physical location. The outcome comparison required a SALE classification; loyalty, quantity & other discount types were not silently relabelled as SALE.^[1][2]

Within each exact barcode, location & currency, we linked a SALE-coded date to the nearest strictly earlier date carrying an explicit “not discounted” flag, with a maximum gap of 28 calendar days. That flag identified the earlier observation for this study. It did not certify a regular or unconditional price. The eligibility period was April 1–September 30, with March observations included to support the first April comparisons.

We selected the first SALE-coded date attached to each earlier observation before checking whether its reference amount was complete. A later, more complete record could not replace it. Same-day competing amounts or classifications remained unresolved, & an ambiguous nearer observation could not be skipped in favor of an older one. Identical analytical price states were collapsed while their proof evidence was retained.

This produced 128 selected comparisons. Fourteen lacked the required reference information & 2 remained unresolved, leaving 112 for the percentage comparison. All 112 had a supplied reference above the current amount; none had a zero current amount. The calculations used exact values, so a matching amount or zero gap was not inferred from rounded display. We reported the currencies separately & made no exchange-rate conversion.

The proof checks required concordant dates & currencies in the metadata; we did not reproduce proof images. The analysis also checked location consistency, missing records, repeated sequences & the effect of removing each location. These rules support the stated finite comparisons. They do not reveal an uninterrupted price history, offer eligibility or the merchant’s reason for choosing a reference.

The article contains information from Open Prices, developed by Open Food Facts, under the Open Database License. Location metadata uses OpenStreetMap.^[5] No proof images or individual source records are reproduced.^[3][4]