The Tropos Index measures the cultural standing of brands. We publish what we measure and how. We hold the exact weights privately so the score can’t be gamed. This is what we measure, how we measure it, and where the limits of measurement lie.
A benchmark earns authority through scrutiny, not secrecy. Every category that contributes to a Tropos Index score is published here in plain language, so that any analyst, journalist, or board member can understand precisely what the number represents.
Unauditable numbers are not a hypothetical problem. Two ISO-compliant valuation firms published the same brand’s value for 2025, $103.6B apart. Neither number can be audited. A figure that cannot be checked cannot be argued with — and a benchmark no one can argue with is not a benchmark.
What we hold proprietary is the exact weighting of each category and the signal-processing that produces a score — the same posture adopted by every credible institutional benchmark. You will always know what we measure. The precise recipe is the trade secret that protects the integrity of the index from being gamed.
Age, gender, and income brackets have stopped predicting cultural behavior. Communities now form around constituencies — coherent groups defined by shared aesthetics, vocabularies, and authority structures. The constituency map is our observational layer: a taxonomy of contemporary subcultures, mapped with the rigor of a benchmark provider.
Each constituency is documented to a consistent standard — its definition, the signal patterns by which it communicates, its key voices, the verticals where it holds influence, and its cultural trajectory: rising, stable, or in decline. Each brand’s coverage is mapped to the constituencies it reaches.
We launch with a starter taxonomy across Fashion, Beauty, Entertainment, CPG, and Technology, expanding continuously as new communities cohere and others dissolve.
PROPHET is our predictive layer. When a pattern in the coverage looks like it’s building, we write it down as a dated, yes-or-no forecast with a probability, and publish it. Nothing is sent to a client without an editor’s review.
An alert is drafted only when a detected pattern crosses a client-defined threshold — and every alert is reviewed by an editor before it is sent. PROPHET monitors seven categories of cultural risk:
Every brand we cover receives a Tropos Index score from 0 to 100 that tracks its cultural standing in real time. The score is a weighted composite of three disclosed pillars. The pillars are public; their relative weighting is proprietary.
The scale and quality of the coverage a brand actually receives — how much is published about it over 30 days, and the standing of the outlets publishing it. These are the only two measures in this pillar.
The direction and speed of cultural momentum — week-over-week change in coverage volume, and the trajectory of search interest in the brand. These are the only two measures in this pillar.
Durability of standing — how long the brand has held cultural presence, and how widely that presence is spread across the constituencies we track rather than concentrated in one. These are the only two measures in this pillar.
Six components make up the score, two in each pillar, and every one of them is measured on the brand itself. Four components that were previously carried at a neutral baseline — organic-versus-paid presence, insider engagement, constituency-velocity correlation, and recovery — were removed from the composite on 24 September 2026 rather than left standing in it, because a value we substitute is not a measurement and should not be weighted as one. The Index can return any score from 0 to 100. Where a brand’s coverage is too thin to measure a component, we return no score for that brand rather than a substituted one.
The Index measures the magnitude of a brand’s cultural standing — how much of it a brand holds. It is not a risk rating: a brand with modest measured presence is small, not in trouble. Cultural risk is a separate reading, produced by a separate instrument.
| Score | Tier | Reading |
|---|---|---|
| 85–100 | Elite | Cultural icon standing — sustained, broad and durable |
| 75–84 | Strong | Broad cultural standing across multiple audiences |
| 65–74 | Solid | Substantial standing; established presence and breadth |
| 50–64 | Present | Real cultural presence — concentrated, or still building |
| 0–49 | Peripheral | Limited measured cultural presence |
Every tier is reachable. The Index can return any score from 0 to 100, and Strong and Elite are held by the same arithmetic as every other tier — no component of the score is substituted, so nothing caps the top of the scale.
Which brands occupy which tier is a reading of a particular day, not a property of the instrument, and it changes as coverage changes. The current board states the occupancy as of its own scoring date.
One structural limit is worth publishing anyway: two of the six components read change rather than level, so a brand already at the top of its category — sustaining a position rather than gaining one — scores lower on those two than its standing warrants. That is a property of the specification, and we state it rather than wait to be asked.
Partly, and deliberately: cultural standing is scale-inflected, and a brand nobody is discussing does not hold much of it. Coverage volume is the single largest input to the Index, and we do not present that as incidental.
The test that matters is whether the Index separates brands of the same size. Banding the board into quartiles by 30-day coverage volume, measured on the 24 August 2026 scoring corpus, the Index still spreads brands within every band — by 13.2 to 20.8 points, with a within-band spread that is 58% of the total spread across the board.
Volume tells you which band a brand is in; the other components decide where in it.
Sentiment-derived readings were re-based on 7 September 2026; percentages are withheld from client surfaces until the re-based classifier has 30 days of data.
Several components of the Resilience and Velocity pillars are computed from a brand’s constituency alignment. Those components are computed across the entire constituency taxonomy — including constituencies whose tagging accuracy we have measured as poor, and constituencies we have not yet measured at all. The constituency map, by contrast, states the measured accuracy of every constituency it shows. So the map is more forthcoming about measurement quality than the score currently is. We regard this as a genuine inconsistency rather than a nuance; it will be resolved when those components are next re-specified, and we are stating it here until then rather than waiting to be asked.
Cultural standing cannot be read from social media alone. We collect published coverage every day from four source categories, and we weight each item by the authority of its source — an argument advanced in a trade publication carries more weight than an anonymous post.
Source-quality hierarchy, highest to lowest: Trade & specialist press → YouTube and Hacker News → press-release wires.
Our infrastructure ingests and structures cultural signal at a scale no individual could read. But a number is not an insight. Every score, forecast, and briefing passes through human editorial review before it reaches a client.
AI augments the analysis; it never replaces the judgment applied to it.
We do not claim to predict the future. We forecast probabilities and attach a stated confidence to every forecast — low, medium, or high. A brand’s score is a measurement of standing, not a guarantee of outcome.
Where signal is thin, we say so. Where a category is too new to score with confidence, we flag it rather than manufacture precision. Initial weightings are deliberately conservative and are calibrated against observed outcomes over time. Honesty about the limits of measurement is itself a measure of the institution’s credibility.
Assigning a signal to a constituency is a judgment, and judgments have an error rate. We measure ours per constituency rather than as a single headline figure, by drawing a sample of that constituency’s tagged signals and re-judging each one against an independent reference model that sees the constituency’s definition but not the patterns our tagger matches on. Every constituency on the constituency map carries the result on its card: the point estimate, the 95% confidence interval, and the number of items audited.
We publish the interval and the sample size every time, and we do not publish the percentage without them. A bare accuracy figure cannot be checked, cannot be argued with, and conceals exactly the thing a reader needs — how much of it is measurement and how much is small-sample noise. An accuracy figure of 44% measured on 90 items tells a client more than a 67% measured on 15, and we would rather publish the first.
We apply no accuracy floor. Constituencies are not hidden from the map for measuring poorly; they render at the figure they earned, and the map encodes how well each one is measured — solid points are constituencies with narrow intervals, fainter points wider ones, and a hollow point is a constituency we have not yet measured. The one thing we withhold is a percentage we cannot support: where too few items have been audited to state a figure, the card says “not yet measured” rather than inventing one. Where a constituency has so few tagged signals that we have judged all of them, its interval is as narrow as that constituency can ever be measured, and more auditing cannot improve it.
Tag precision refreshes only when the audit re-runs. It is a periodic labelling pass, not a continuous measurement: the figures on the constituency map are the results of the most recent audit, dated on every card, and they do not move as new signals arrive, as scores recompute, or overnight. A constituency’s precision figure can therefore be older than the coverage shown beside it. The confidence bands the map draws — the boundaries separating a solid point from a faint one — are set from the observed spread of intervals across the taxonomy, so they are re-tuned when the audit re-runs and the taxonomy grows, rather than fixed constants.
The Tropos Index methodology is governed, not improvised. Category weightings are reviewed on a quarterly cadence and recalibrated against three tests: whether brands earning genuine cultural recognition score highly, whether brands that later suffered cultural failures showed deterioration beforehand, and the judgment of our founder.
Material changes to the framework are versioned and disclosed to Founding Partners, who hold standing input into the methodology’s evolution. This page reflects Version 1.0.
Constituencies are occasionally renamed, where a name has drifted from the definition it labels. A rename changes what a constituency is called and nothing else: the same signals tag to it, its definition is untouched, and no score moves. We record every one here with the date it took effect and the reason.
Because our measurements are stored against a constituency’s identity rather than its name, historical views keep the name that was in force when they were measured. A report delivered in July carries July’s name permanently — we do not rewrite documents that have already been sent. So if a brief in your archive names a constituency differently from the app, both are correct as of their own dates, and the row below is the explanation.
| Effective | Previous name | Current name | Reason |
|---|---|---|---|
| 22 August 2026 | Prestige-TV Tastemakers | Prestige Film & TV Tastemakers | The name was narrower than the definition, which has always covered prestige film as well as television. Definition, signal patterns and tagging are unchanged and no score moves. Verified before it was applied: across 400 signals the rename moved fewer primary-tribe assignments (46) than re-running the old name against itself (48), i.e. within the tagger’s own noise floor. One genuine recall loss is recorded rather than buried — a Letterboxd item went untagged under the new name, and Letterboxd is one of this tribe’s five signal patterns. One item in 400, against a 12% noise floor. |
Analysts, academics, and press evaluating our methodology are invited to write to chris@troposindex.com. We welcome scrutiny.