What "unfiltered" means here
Most public economic series have been transformed before anyone reads them. Seasonal adjustment, base-year indexing, definitional revisions, survey re-weighting, exchange-specific clearing — all happen upstream of the number that ends up in a chart or a headline. Naming those transformations is useful, but words alone don't make the cost of filtering visible.
So this site does one thing: for each series, it fetches two views and puts them next to each other on the same date. One column is the least-transformed version available; the other is the most commonly cited version. The gap between them — published as a Δ column on every page — is what filtering does, in the units of the series itself.
How the comparisons are constructed
- Macro series (CPI, M2, Unemployment). Both columns come from the same primary issuer, fetched via FRED. The "raw" column is the not-seasonally-adjusted variant; the "filtered" column is the seasonally adjusted variant produced by the issuer's X-13ARIMA-SEATS pipeline. Neither column is touched after fetch.
- Bitcoin. No statistical filter is applied at the source — but no exchange-neutral price exists either. The two columns are daily UTC closes from Coinbase and Kraken. Both are "raw" within their venue and "filtered" by the venue itself: each exchange's order book, fee structure, and user base shape what trades clear and at what price.
What this site is not
Not a research product. Not a forecasting product. Not a chart-of-the-day site. There is no commentary attached to the values, because the point is to publish numbers without one. The Δ column is the editorial.
Provenance
Each series page lists both sources, their issuers, and series IDs. The raw JSON file behind every page is downloadable; URLs are stable. Refresh runs nightly.
Who runs it
UnfilteredData.com is an independent project by George Clay, part of the Clay Indices family.