03Working tool · runs entirely in your browser
Does this offer resolve at my store — and to what?
Search 28 published offers as a customer would. The checker resolves each one against the product sample, states the effective unit price, and names what it cannot know. Every result shows its reasoning.
How to read this workspace — why it exists · what data it uses · what the controls do
Why it exists. A Puff deal card carries a title and a terms sentence. It does not carry the products it covers, the stores that honour it, or a date. This resolves that card into the things a shopper actually needs: eligible products, an effective unit price, and an honest list of what is still unknown.
What it searches. 28 unique offers, every one of them read off a public Puff page on 2026-09-28. The products they resolve against are a modelled sample of 63 records built from publicly named brands and categories — they are not Puff's inventory, and no real SKU is claimed. Exactly one price in the sample is real: the $25.00 published for a Mitten Extracts 2-in-1 Flip on the store-scoped Monroe product page. It is marked wherever it appears.
What the controls do. The store selector is not decoration — a price is only shown for a store the sample has a price for, so changing the store genuinely changes the arithmetic. The category chips filter which offers appear. The toggle hides offers that resolve to nothing.
What the output is not. A verdict here is a statement about whether a publication is complete and self-consistent. It is not a statement that a store is doing anything wrong, and it is not a claim about sales.
0 offers match ·
The weights, published.
So you can predict what the tool will do, and disagree with it precisely. Filters decide whether a record qualifies. These points decide whether qualifying is certain or needs a person. There is no confidence score here, because there is no probability being estimated — it is a reconciliation total, and calling it anything else would be dressing up arithmetic as statistics.
- Brand Exact
- The record's brand is one of the brands the deal names. +300
- Brand Alias
- The record's raw brand string matches a known alias rather than a mapped brand id. Qualifies, but a human should confirm it. +240
- Category Match
- The record sits in the category the deal implies. +160
- Form Match
- The deal's product wording appears in the record's name or form. +120
- Size Match
- The record's net weight equals the size stated in the deal. +90
- Potency Match
- The record's potency equals the potency stated in the deal. +60
- Hint Token
- Per matching token of the deal's product wording. Capped. +50
- Hint Token Max
- The cap on product-wording points, so a long name cannot dominate. +150
- Size Mismatch
- The record qualifies on brand and category but its size disagrees with the deal. -80
- Form Mismatch
- The record qualifies on brand and category but its form does not match the deal's wording. -70
- Potency Mismatch
- The record qualifies but its potency disagrees with the deal. -60
- Unknown Brand String
- The record's raw brand string maps to no brand at all, so brand is treated as unknown rather than as a mismatch. 0
What a verdict meansREADY — every check passes on what is published. REVIEW — something contradicts itself; a person should look. BLOCK — a blocking check fails. UNKNOWN — nothing contradicts itself, but the publication does not carry enough to confirm it either.
The honest limit42 of 63 sample records carry no price at all, and the modelled ones exist only for three stores each. A saving is hidden rather than estimated whenever the comparison would be invented. That is the whole discipline of this page in one sentence.