Receipts
Every number links to a source. This page explains how we check those sources, build the totals and decide when a comparison is too weak to show.
What the data includes
A dated collection of franchise lifetime-revenue estimates, adapted from Wikipedia's list of highest-grossing media franchises and the sources it cites, under CC BY-SA 4.0. These are estimates of money taken in before costs, not profit, company value or audited accounts. Each release is dated; the numbers are not live.
- snapshot
- 2026-09-03
- schema
- v1
- pipeline
- 1.0.0
- franchises
- 99
- revenue records
- 360
How each number is stored
Franchises never store totals. Every published figure is an append-only revenue record — value, currency, channel, measurement basis, period, geography, source citation and a five-axis confidence score — and every display value is a deterministic rollup over the active records. Corrections are logged override entries against a record id; published files are never hand-edited. Each franchise page's provenance table reads straight from these records.
How we rate the sources
Every record scores five axes, 0–4. The composite is a weighted average; the weights below are the pipeline's own.
| Axis | What it measures | Weight |
|---|---|---|
| Source quality | Who published the figure — an industry report scores higher than a passing mention. | 0.3 |
| Category clarity | How cleanly the figure maps to one revenue channel, without guessing. | 0.25 |
| Date freshness | How recent the figure's as-of date is. | 0.15 |
| Breakdown completeness | How much of the franchise total is actually broken into channels. | 0.15 |
| Overlap risk | How likely the figure double-counts another franchise. Higher score = safer. | 0.15 |
- Solid — composite ≥ 3 and every axis ≥ 2.
- Reasonable — composite ≥ 2 and every axis ≥ 1.
- Rough — everything else. Any axis at zero forces Rough, whatever the average says. Rough segments render hatched in the barcode itself; uncertainty is not hidden in a footnote.
How we choose each business pattern
Each label follows a fixed rule based on the revenue we could break down. We check the rules from top to bottom and use the first match.
| Business pattern | Rule | The sentence it must survive |
|---|---|---|
| Shelf Empire | share(retail_merch) >= 0.60 | Retail and merchandise makes up {share}% of the revenue we could break down. We use this label at 60% or more. |
| Game Economy | share(games) >= 0.50 | Games make up {share}% of the revenue we could break down. We use this label at 50% or more. |
| Screen Empire | share(box_office) + share(home_video_streaming) >= 0.50 | Cinema, home video and streaming make up {share}% of the revenue we could break down. We use this label at 50% or more. |
| Experience Engine | share(experiences) >= 0.30 | Live experiences make up {share}% of the revenue we could break down. We use this label at 30% or more. |
| Expanded Universe | fallback | The money is spread across {n} types of revenue. No single one dominates. |
How we calculate the comparison numbers
We calculate five extra comparisons. They stay separate from reported figures, and we leave them out when the source data is not strong enough.
- Share from merchandise
- retail_merch ÷ disclosed channel sum. Shown plain at breakdown completeness ≥ 3, as “approximately” at 2, hidden below that.
- How varied the revenue is
- Normalised Shannon entropy across the 8 channels, 0–100. Requires breakdown completeness ≥ 3.
- Revenue beyond where it started
- Share of the revenue we could break down that came from outside the original medium's channel. Requires a known original medium and breakdown completeness ≥ 3.
- Average revenue per year
- gross ÷ franchise age, minimum age 3 years, in nominal and CPI-adjusted flavours. Always labelled “lifetime average — not current run rate”.
- Online interest compared with revenue
- Five-year search interest compared with estimated lifetime revenue, scored from −100 to +100. This is a rough comparison of current online interest, not an economic measure. Search data older than 90 days carries a stale badge; older than 180 days it is hidden.
Overlaps — three treatments, never a fourth
Franchise nesting is an explicit relation graph (MCU contains Avengers; Spider-Man shares characters with both). Any aggregate over a set — owner rollups, portfolio totals, filters — applies exactly one treatment per relation:
- Fully nested, source-confirmed — the child is excluded from the sum and chipped “included in the parent total”.
- Partially overlapping, sourced amount — the sourced adjustment is applied and cited.
- Partially overlapping, no sourced amount — the unadjusted total is shown with a visible warning. In that case the methodology permits the honest answer: an exact deduplicated total is unavailable. Overlap estimates are never fabricated to make an aggregate look precise.
How ownership is worded
- Owner rollups say “franchises with rights-level records linked to …” — never “owned by”, because rights are narrower than ownership.
- Aggregates carry the caveat: “lifetime revenue of those franchises — much of it earned before or outside …'s ownership”.
- Stewards are computed only for franchises whose current rights are verified. Association-status franchises render the connected company with “Current rights structure is incomplete.” and are excluded from ownership counts; unresolved franchises render no owner at all.
The game shows presentation, never different data
Barcode's daily puzzle recycles clue configurations — 3 barcode presentations, 3 era-clue types, 3 ownership-clue types. Hard rule, stated here because it is enforced in config: variants change presentation only — channel ordering, label visibility, absolute or proportional display. The underlying revenue distribution is never altered to make a repeat feel fresh.
When a change becomes a Market Move
A revision is material at a 2% move or $500M; additions, removals and ownership changes are always material. Items auto-publish to the ticker only at ≥ 0.85 classifier confidence, in auto-publishable classes, after a deterministic check that every numeral in the draft appears verbatim in the source diff. The newsletter is always human-sent.
Licences
- Revenue figures adapted from Wikipedia's list of highest-grossing media franchises and its cited sources, CC BY-SA 4.0. This site's own data files ship under the same licence — share-alike is inherited, not optional.
- Film metadata, where shown, uses the TMDB API. Its free tier is non-commercial: Free tier is non-commercial — written approval or swap before any commercial toggle flips.
- Search-interest data (for the Attention Gap) comes from Google Trends via serpapi, reported only as normalised percentiles with query, geography, window and retrieval date stored per value.
Known gaps
- Attention Gap is currently gated on 99 of 99 franchises in this snapshot — no Trends retrieval has passed the freshness gate yet, so the metric shows its gate reason instead of a number.
- 6 franchise relationships have no reliable public overlap amount. Totals that include them show the warning above and are marked as approximate.
- Some franchises publish only a lifetime total with no channel breakdown. Their barcodes carry an unattributed segment rather than an invented split.
Corrections
Numbers get better when people argue with them. If a figure is wrong, stale, or missing a better source, open a public correction issue with the source attached. Accepted corrections land as logged override entries with author and reason — the audit trail is part of the dataset. See also Market Moves for how routine revisions flow in.