Why you can trust the number
CPI is a black box. This is the opposite.
You cannot see how any single CPI number was built. Here, every figure shows its source, its confidence rating, and the exact script that produced it. Three rules govern everything:
1
Real prices, not guesses
Actual transaction prices wherever they exist. Most of all, we replace CPI's survey of imagined rents with the real sale price of a home.
2
Sourced & reproducible
Every item carries a source and a confidence rating. Estimates and pre-1950 figures are flagged. The raw data is one auditable file.
3
Steelman first
We state the strongest version of the official defense before disputing it. Where a government figure is fair, we say so — that's what makes the rest land.
How the number is computed loads here…
Every figure earns a confidence rating
Two independent checks run on every item: an internal audit (outlier + sanity tests, and whether a real per-year series exists) and an external cross-source check against a fully independent second source. The six tiers, best to weakest, with live counts from the latest audit:
Cross-verifiedAn independent second source confirms today's price within 5%.
Audited · item-levelPasses every internal check and has a real, item-specific price series (the green line on its chart).
Category modelEndpoints sourced, but the year-by-year path follows the broad BLS category index (a labeled proxy) — honest, but not item-level.
PlausiblePasses the internal audit; no second source or series yet.
ReviewFlagged as a category outlier or by a sanity check; awaiting a human or second-source look.
ConflictA second source actively disagrees by more than 10%. We'd rather show the disagreement than hide it.
Scheduled jobs grow this automatically — a monthly harvester backfills real per-year series and a weekly job adds independent second prices and re-runs the audit. The cross-verified count only goes up.
Source-quality tiers
Tier 1Gold Standard
An authoritative, continuous official series — BLS Average Price Data, Census, EIA.
Tier 2Sourced
Real data with gaps, a definition shift, or a documented splice — e.g. the Shiller-derived 1913 home price.
Tier 3Reconstructed
Built from catalogs, menus, ads, or estimates. Estimates and pre-1950 figures are flagged.
Common criticisms — answered
Tap each. We don't dodge the hard ones.
"Why not just use CPI?"
We do use CPI's structure — the same eight major groups at CPI's own weights. We change only what CPI measures with a hypothetical: housing. CPI fills ~⅓ of its index with Owners' Equivalent Rent, a survey of imagined rents. We use the real sale price instead.
"Why these categories and weights?"
They aren't ours — they're BLS's official relative-importance weights. Using CPI's own basket is exactly what makes this a fair, apples-to-apples replacement rather than a cherry-pick.
"What about quality improvements (hedonics)?"
For the BLS categories we keep CPI's quality adjustments as-is — so our number is, if anything, conservative there. We only strip hedonics where it most distorts lived cost: cars, where we use raw transaction prices, because you can't actually buy the 1913 level of car anymore.
"Isn't gold cherry-picked?"
Gold lives on its own page as a sound-money lens — it does not touch the headline index, which is built from consumer prices, not gold.
"Aren't wages up too?"
Yes — and many goods are genuinely cheaper in hours of work (electronics, clothing, some appliances). We say so plainly. But housing, healthcare, and college are dramatically worse in labor terms, and those are the things that determine whether a life feels affordable.
A living prototype: the basket mirrors CPI's full structure; figures, sourcing, and the historical series keep improving. We ship the good rather than wait on the perfect.
See the headline figure → · Browse all prices →