Public recipe

Methodology

The public dashboard is an educational composite. It is not a national statistic, not a live tick, and not a promise that gloom forecasts equity returns. This page is the recipe we will put in public without pretending otherwise.

1. What the number is

Each regional and sector print is scaled to a 0–100-ish mood score whose quiet weather sits between 40 and 60. Fifty is neither boom nor bust. The daily value is deterministic for a UTC date so two readers, two continents, see the same tape. Charts on the markets pages show that history in 1M, 3M, 6M, and 1Y windows, with the date on the x-axis and the print on the y-axis.

2. What we sit near

The U.S. line is calibrated in the neighborhood of widely watched household surveys such as the University of Michigan Surveys of Consumers and the Conference Board Consumer Confidence Index. Europe sits near European Commission consumer opinion and national statistical-office confidence series. Korea sits near Bank of Korea / Statistics Korea consumer-survey families. Japan and China lines are deliberately wider-uncertainty neighbors of official confidence prints. We do not claim identity with any of them. When they diverge from each other, our tape will not magically pick a winner.

3. Sector mapping

Sectors are not copies of the global average with new labels. Retail overweight purchase-intent style questions. Housing splits, in licensed files, into purchase conditions versus owner shelter comfort. Autos follow vehicle-buying conditions. Travel follows vacation intent and is noisier. Staples move slowly. Consumer tech follows durables-and-gadgets intent. Household energy is an affordability instrument. Labor follows job-plentiful / job-security items. Licensed CSV files expose those legs. The public HTML table does not, on purpose.

4. Lags we will not hide

A survey print dated Friday may describe interviews that ended earlier in the week. Official agencies revise. Our public site does not intra-day revise; it prints once. Licensed feeds timestamp vintage (preliminary vs final) when a source has one. If a vendor cannot say which vintage you stored, you cannot backtest.

5. Alternative data

We do not wash social posts into an official-looking CCI. Public language is a nowcast feature for categories where people complain in public, aggregated, and dropped when a platform’s terms change. No private messages. No lists of people. See the insights essay on social listening.

6. What this site will not do

The public pages will not score a live social firehose as if it were a statistical office. That is a different product, and a different contract.

7. Change log

2026-09-14 — Public educational tape launched. Daily UTC print, eight sectors, six regional lines, bilingual site.

2026-09-14 — Markets charts show dated windows (1M / 3M / 6M / 1Y) with labelled axes.