HOW THE NUMBERS ARE MADE

Receipts All Day is run by one person — a food-service operator with a decade behind the pass and a data habit. Press: press@receiptsallday.com. Spotted an error: heard@receiptsallday.com.

ON THIS TICKET

• where the data comes from — and what it can't say
• the 3-source rule
• the seasonality check
• what counts as "smash burger"
• the blocking verifier
• known limitations

WHERE THE DATA COMES FROM

• Wikipedia pageviews (Wikimedia API): monthly views of a food's article. Measures curiosity — people looking something up. Does NOT measure sales, orders, or how much anyone ate.
• Google Trends: relative search interest, 0–100 against its own peak. A shape, not a count. Never anchors a claim alone.
• Reddit (a fixed panel of food subreddits): what people argue about. Discourse, not consumption.
• Food-media RSS: what the press is writing about. Coverage, not popularity.
• New sources are announced here as they're added, with what they measure and what they don't.

THE 3-SOURCE RULE

No quantitative claim publishes on one signal. Relative tools (Google Trends, Pinterest) must be corroborated by raw-value signals (Wikipedia pageviews are raw counts). Where signals disagree, we investigate or stay quiet.

THE SEASONALITY CHECK

Before any "growing" or "dying" claim, the pipeline compares the same window year over year. Fur coats spike every winter; pumpkin spice dies every July. If the clock explains the move, the claim is blocked — mechanically, not editorially.

WHAT COUNTS AS "SMASH BURGER"

Which searches, articles, and phrases count as a given food is a versioned list, and every change is reviewed before it touches the data. Naming things carefully is most of the job.

THE BLOCKING VERIFIER

Nothing publishes with unverified numbers. Before a post ships, an automated gate proves every number in the chart and caption against the source data, checks the data is fresh, and a language model flags overclaims ("dead" versus "at 11% of peak") for human review. The gate blocks; a human can never overrule a block, only a warning.

FACTS VS INTERPRETATION

Numbers print as line items. Takes are labeled as takes. When we grade the industry's trend predictions, we grade institutions in aggregate — never individual writers.

KNOWN LIMITATIONS

• Lookups measure curiosity, not consumption. A viral moment and a grocery habit look identical in pageviews.
• English Wikipedia skews American and online. So do we, for now.
• Relative indexes (Trends, Pinterest) can't say how big — only how big compared to their own peak.
• Exact-phrase matching misses slang we haven't added yet, and new foods take a few months to accumulate enough history to say anything honest about.
• Corrections are public, within 24 hours, on the corrections page — same check number, reprinted ticket.