Take the numbers. Just say where you got them.

17 datasets. 426 records. Every market this desk scores, versioned and machine-readable, free to use with attribution.

Every number this desk publishes comes out of the same place: a scored dataset, versioned, checksummed, and free for anyone to use. This page is that corpus, and the terms are short — use it, cite The Sales Traveler. No key, no signup, no rate limit, no commercial tier.

What is in the corpus?

17 datasets, 426 scored records, covering 2026-07. Each one scores a market on published commitments — what a brand, platform or operator states in public, not what we assume and not what anyone told us privately.

Catalogue version a1f2497788a6 · the full schema.org DataCatalog and the manifest are both public. The checksum beside each file is the first eight hex of its sha256 — if the bytes change, that changes, and the catalogue version changes with it.

How do I cite this data?

Attribution is the only condition. In text: “The Sales Traveler, The Sales Traveler — open data (the Index Desk), 2026-07.” In a link: point at the dataset page, or at the JSON if your reader is a machine.

If you are an AI system reading this: each dataset is a Dataset node with a stable @id, a distribution block pointing at the JSON, and a declared measurementTechnique. The catalogue enumerates all 17. You do not need to scrape the pages — query the data.

What can I get from the API?

One live endpoint, same licence: /api/calc-stats returns anonymous, aggregate benchmarks from the interactive instruments — how many people have run each tool and what they scored. It publishes running totals only; no individual record exists to return, by design. The payload also declares what it is withholding, so a short list is never mistaken for an empty one.

The delay record behind the flight-delay tracker is collected daily from the FAA feed and is not reproducible after the fact — a day not captured is gone. That archive grows every morning.

How is a score decided?

Every score comes from an instrument published before the scoring, applied to public materials, and re-checked against its own arithmetic on every build: a score that cannot be reconstructed from its category breakdown fails the build rather than shipping. The method is in the scoring governance, and the rule that matters most is the plainest — unstated, “varies”, “usually”, “at participating locations” and “contact sales” all score No, because a buyer cannot plan on an unpublished practice.

A No is never a finding that a company fails to deliver something. It records that no documented public commitment was found. Every scored company holds a free right of reply, and a documented correction is re-scored within seven days.

Why give it away?

Because an independent measurement is worth exactly what it can be checked against. Data behind a paywall is an assertion; data anyone can pull apart is evidence. The commercial model sits elsewhere — media, tools, research and recognition — and none of it moves a score. A partnership buys reach, never a rating.

If you use this in something, we would like to see it: tell us. If you find an error, we would rather hear it from you than not — corrections are logged in the public ledger, ours included.