On this page
"SALE" is a design element, not a fact. Anyone can print it in red. This post explains what FlyerHawk actually does before it's willing to call a price a deal — partly for transparency, partly because the method is a useful way to think about grocery prices even without us.
The data
Our database currently holds over 84,000 flyer prices covering more than 45,000 distinct products across 16 grocery chains, growing every week. Every price is recorded with its store, dates, package size, and computed per-unit cost. That history — not the retailer's "was $9.99" claim — is the yardstick everything gets measured against.
The method
1. Every price is scored against its own history. For each product we build percentile benchmarks from its recorded prices — at the same store, and across the market. A price in the bottom 10% of its history rates "amazing," the bottom 25% "great," the bottom 40% "good." Everything else is just a Tuesday.
2. A percentile alone isn't enough. Some products list at nearly the same price every single week, which would make a 21-cent discount technically "bottom 10%." So top ratings also require real savings — roughly 10% below the typical price for an "amazing," 5% for a "great." No meaningful savings, no badge.
3. Sizes never cross-contaminate. A 1 L chocolate milk is only ever compared with other 1 L chocolate milks — never with 4 L jugs, multipacks, or the milk-chocolate almonds that happen to share the words. Same for a dozen eggs vs an 18-pack. Comparing across formats is how "50% off!" illusions are manufactured; we structurally can't make that mistake.
4. Per-unit price gets a veto. A smaller pack at a lower sticker price can look like a deal while costing more per gram. Wherever we know the per-unit price, a top-rated deal must also be cheap per unit, or it gets demoted.
5. Multi-product lines don't get to cheat. Flyers love lines like "Product A, Product B OR Product C — $9.99." One price across three products means the price belongs to none of them individually, and we treat it with exactly that suspicion.
What this means in practice
On a typical Thursday, the large majority of advertised specials fail these checks. The minority that pass are the ones worth driving for — and those are the only ones we'll email you about. One weekly email, deals that beat their own history, nothing else.
That's the whole editorial policy: the data decides, and the data has a long memory.
