We’re proposing a measurable definition of brand health.
Proposing, not proclaiming. Brand health has never had an agreed, testable definition, so we wrote one down as seven axioms, built a measurement system that follows from them, and committed to grading it against outcomes it does not get to choose. If you reject an axiom, you should reject the score. If you accept them, everything on this page is the evidence that the definition holds.
Read it at the depth
you work at.
The same methodology, written three times for three readers: a plain-language whitepaper for the people who decide budgets, a technical paper for the people who check mathematics, and an engine specification for the people who build on the score. All three are print-friendly, cross-linked, and versioned against IMPRS V32.
The formula has been rewritten, recalibrated and re-validated continuously since its first version. This archive is reconstructed from the actual engine source at each stage. Versions whose code predates this repository are labeled; their notes come from references preserved in later changelogs, not from memory.
V32
IMPRSALGO.TS · JULY 2026CURRENT PRODUCTION ENGINESingle-channel sentiment, the version scoring every audit today. Sentiment influences the score through exactly one auditable channel.
V26 carried no standalone changelog entry in any recovered source; its work is folded into the V27 record. Full validation methodology, including the mandatory balanced-accuracy framing for every directional claim, is documented in the sections above and in the Technical Paper.
We would rather tell you
exactly where we stand.
IMPRS has completed retrospective validation on a sealed 122-company panel across six years of revenue data. That is real evidence, and it is the first rigorous rung on the evidence ladder: a well-documented backtest. It establishes that the methodology is promising. It does not prove it.
122 companies is respectable for a pre-launch company and small by scientific standards. The formula was iterated across twelve versions on this same panel, with hold-out testing each time. Even so, repeated iteration on one dataset can produce optimistic results before external data arrives. We think that is the honest thing to say, so we are saying it.
We believe brand measurement should earn trust the way scientific claims do: by surviving future tests, not only by explaining past data. So every production forecast is now logged before its outcome is known, and prospective performance will be published without retrospective editing as results come in.
The score is the product. Forecasting is an application.
A credit bureau does not advertise that it predicts default. It publishes a measure of creditworthiness, and lenders use it to make their own decisions. IMPRS is the same. It is a measure of brand health. The forecasts, early warnings, and directional signals elsewhere on this page are applications built on top of that measure, not the measure itself.
Brand health has no direct observable measurement, so we validate it against subsequent business outcomes: realized brand equity, measured as revenue growth. In retrospective validation, when the score clearly separated two accounts of the same type, the higher score identified the faster-growing one up to 86.6% of the time for B2B brands and 81.7% for Fortune 500 brands. The top-scoring half of a tier outgrew the bottom half in 80% of panel years for four of six tiers.
Every one of those measures is size-independent, so a score that merely tracked company size would fail them. Across every metric and every segment we tested, IMPRS outperformed a benchmark index rebuilt from the industry standard, on the identical panel, even after we handed the benchmark trailing revenue as an input.
Why growth ordering is the right test: a brand health measure that merely tracked size would rank Walmart above every challenger forever. Every measure here is size-independent, so the only way to score well is to read the underlying equity, the audience behavior that showed up in revenue later.
We do not validate against today. We validate against what happened next.
Descriptive analytics tell you a score matches this quarter’s numbers. That proves nothing. The test that matters is whether the score computed in one year lines up with the revenue that arrives the year after. Every association on this page is measured that way: signal in year N, outcome in year N plus one.
Movement carried the most information. On the same panel, changes in the score were associated with next-year revenue changes at rho +0.395 pooled, against +0.160 for the score’s level, and +0.563 for consumer brands. This is why the product is built around deltas: the morning briefing, the agency delta report, the forecast all track how the score is moving, because in the backtest that is where the signal lived.
Six tiers. Six different truths.
A score that claims one accuracy number for every kind of account is hiding something. Accuracy differed by segment, so we publish it by segment: what held, what did not, and what we refuse to claim yet. Each panel leads with the measurement result and marks its forecasting numbers as the application layer.
The cleanest measurement result in the panel.
Agencies are the launch segment because the numbers needed no qualifying. On all four measurement metrics, IMPRS outperformed the rebuilt industry-style index, and the index itself went negative: an agency's size, spend, and reach were anti-correlated with its subsequent growth.
The panel is 15 firms, so the intervals are wide, and single-year direction calls carry modest skill. What agencies can stand behind is ordering: who is stronger, and which way they are heading. Read the wide interval on the gated-call row as exactly what it is. This tier will move most once the live panel grows.
The score you can
defend in a client room.
Agencies sell outcomes, so the score they report needs receipts. We launch with agencies because agencies posted the cleanest measurement result in the backtest, on 15 real holding groups and challengers across six years, tested against a brand strength index rebuilt from the industry standard’s published component categories: marketing investment, stakeholder equity, business performance.
The reconstruction sees agency size, spend, follower base, and even trailing revenue. It still ordered future growth at 50.5%, no better than random, and its level correlation with agency growth came out negative. IMPRS came in higher on all four measured metrics. When two agencies’ scores were clearly separated, the higher IMPRS identified the faster-growing firm 64.7% of the time [56.9%, 71.8%].
Read it straight: the agency panel is 15 firms and single-year direction calls carry modest skill. The claim agencies can stand behind is ordering, who is stronger and which way they are heading. That is what a retainer conversation runs on, and it is the claim we are now testing prospectively.
A month-over-month IMPRS delta you show a client is backed by a movement association of +0.410 between score changes and the revenue changes that followed, and a 64.7% ordering result when the score confidently separates two brands. The industry-style alternative offered chance.
Higher ordering,
every segment.
Brand Finance’s Brand Strength Index is the most cited brand health measure in the industry. It is published annually as a ranking, without forward accuracy statistics, confidence intervals, or a public validation panel. We cannot compare against their proprietary model, and we do not claim to. So we did the next best thing: we rebuilt an index from its published component categories, marketing investment (30%), stakeholder equity (35%), business performance (35%), and ran it on our identical 122-company panel.
We handed the reconstruction trailing revenue as an input, which should help it. Across every metric and every segment we tested, IMPRS still ordered next-year growth more accurately, 56.5% versus 42.3% pooled. Below 50% means no ordering skill at all: knowing a company’s size, spend, and follower count tells you who is big, not who is about to grow.
Two honest caveats. First, part of the reconstruction’s weakness is structural: any size-dominated index anti-correlates with future growth, because smaller companies grow faster. Second, where the reconstruction does inherit signal from autocorrelated trailing revenue, in B2B, it comes out ahead on one metric, and we report that in the B2B panel above. The fair summary is narrower than a headline: IMPRS out-ranked a spend, audience, and revenue composite at ordering future growth. We invite anyone to rebuild both sides independently.
When the engine speaks, it earns the right to be heard.
Forecasting is an application of the measurement, not the product. The forecast engine derives a direction from changes in the underlying score, and it stays quiet unless the signal is strong. It issues a call only when the move is material for that tier and agrees with where the account sits. Everything else gets an honest flat.
In the backtest that discipline was the difference between negative skill and real skill. Across brands and agencies the engine called direction on about a third of company-years and was right 84.4% of the time [77.1%, 89.7%], which is +17.2 points over the base rate. Raw direction-match, by contrast, loses to a parrot that always says up, so we removed it from the page entirely.
This gate ships in the production engine, tier by tier. For creator accounts it goes further and suppresses down-calls outright, because on the panel they inverted. A creator client on an agency roster gets the creator guard, not the agency configuration. These are the calls now being logged prospectively.
A forecast that always has an opinion is a forecast you cannot act on. A flat call means the signal did not clear the validated bar. The same gate that stays quiet on weak signals is what makes the issued calls worth taking to a client.
The backtest is the hypothesis.
The live data is the experiment.
Everything above is retrospective, measured on a sealed panel the formula was iterated on. Honest measurement can still produce optimistic results before external data arrives. The only way to settle that is to issue forecasts in advance and let reality grade them.
So every direction call the production engine issues after July 2026 is timestamped and stored before the outcome is known. It is never edited later. As outcomes resolve, we will publish prospective performance alongside the original claims, including the misses, in the same place with the same weight.
Our goal is not to defend a statistic. It is to measure reality accurately. If the prospective numbers come in materially below the backtest, that is what this page will say.
Calls logged is a live count from the production forecast engine, climbing as accounts are scored. Resolved against revenue stays at zero until the first fiscal-year outcomes land. For context, the retrospective backtest issued 128 gated calls across the sealed panel at 84.4% precision. The counter above starts from zero on purpose. We would rather you watch it fill than take the backtest on faith.
Credit bureaus turned thousands of loan signals into one number banks could trust. IMPRS does the same for brand health. Every score from 350 to 900 was calibrated against one outcome: did stronger brand health show up in stronger subsequent revenue?
Five pillars, each measured against revenue.
Impact.
The depth of the impression. Saves, bookmarks, screenshots, and DM shares: signals that something was worth keeping, not just scrolling past.
Impact is the most stable single signal of long-term brand equity in the panel. A brand with consistently high Impact but low Reach is a diamond waiting to be found. The reverse is a liability. For agencies, Impact carries the heaviest weight of any segment at 40%.
Brand Health Index (BHI)
68 metrics measuring how audiences respond to your brand. Pillar-weighted composite of Presence, Reach, and Impact signals across 5 platforms, blended with revealed-preference Sentiment when revenue is connected. This is what your audience thinks of you, even if they have not told you.
Growth Momentum Score (GMS)
49 metrics measuring how your brand is actively growing its audience. This composite is Momentum-dominant, which is why Momentum is the highest-weighted pillar across nearly every segment. Growth velocity is what most strongly moves the audience numbers six months out.
The score is deterministic. The same inputs always produce the same number, and no language model touches the calculation. AI is used only to explain the score in plain language and to recommend actions. The 60/40 ceiling between the two parts is fixed regardless of data richness, so platforms with more API access cannot inflate a score. Every brand is measured on the same structural basis.
Evidence needs
a definition.
Most of the brand measurement industry publishes rankings without publishing how often those rankings are right. A number without a standard behind it is an opinion with a font. These are the seven rules every figure on this page has to pass before we print it.
The rules exist because accuracy claims are easy to inflate, sometimes without meaning to. Revenue rises in most company-years, so a predictor that always says up looks 71% accurate on our panel. Under this standard that predictor scores zero skill, and a small panel is labeled as one.
Everything on this page today is retrospective: measured on a sealed historical panel. Prospective results, measured on live accounts before outcomes are known, are labeled separately and begin July 2026. We never blur the two.
Every association measures a signal computed in year N against revenue realized in year N+1. Same-year fit is never presented as if it were forward evidence.
Revenue rises in most company-years, so raw direction-match flatters everyone. We report skill over the base rate, where a predictor that always says up scores zero. Raw accuracy alone is never the claim.
No point estimates without sample sizes and confidence intervals. When an interval is wide, you see that it is wide. When a panel is small, we say the number could move.
Results are computed on a frozen, deterministic formula with hold-out validation on each iteration. The formula is never tuned on the data it is graded on. AI explains the score. It never generates it.
When we compare against another methodology, we rebuild it from its published components and run it on our identical dataset. The reconstruction is about forty lines of committed code. We invite anyone to rebuild it, and both sides, independently.
Every change to the scoring formula creates a new version number. Validation results stay permanently attached to the version that produced them. Future versions are validated independently and are never retroactively merged into earlier evidence. V32 is always V32. V33 earns its own.
Common questions, direct answers.
What is the IMPRS score?
IMPRS is a mathematically derived measure of brand health, a deterministic score from 350 to 900 computed from 117 signals across five pillars: Impact, Momentum, Presence, Reach, and Sentiment. AI explains the score and recommends actions. It never generates the score, so two people looking at the same account get the same number.
Is IMPRS proven?
Not yet, and we will not say otherwise. IMPRS has completed retrospective validation on a sealed 122-company panel across six years of revenue. That establishes the methodology is promising, not proven. It is a well-documented backtest, the first rigorous rung on the evidence ladder. Prospective validation on live accounts began in July 2026 and is the test that will settle it.
Does IMPRS predict revenue?
IMPRS is a measure of brand health, not a revenue predictor. In retrospective validation, stronger IMPRS scores and rising IMPRS scores were associated with stronger subsequent revenue growth. The forecast module is an application built on those associations. It derives a direction from changes in the score, and it abstains when the signal is weak.
How was IMPRS validated?
Against what happened next: signal in year N, revenue in year N plus one, on a frozen formula with hold-out validation on each iteration. The primary result is measurement accuracy. When the score clearly separated two same-type accounts, the higher score identified the faster grower up to 86.6% of the time for B2B brands. Every figure ships with its sample size and a 95% confidence interval.
How does IMPRS compare to Brand Finance?
We cannot compare against their proprietary model and do not claim to. Brand Finance publishes its Brand Strength Index as a ranking without forward accuracy statistics. We rebuilt an index from its published component categories and ran it on our identical panel, even handing it trailing revenue. IMPRS out-ranked that reconstruction on next-year growth ordering in every segment. Part of the gap is structural, since any size-dominated index anti-correlates with growth, and we invite an independent party to rebuild both sides.
How will you know if you are wrong?
Every production forecast issued after July 2026 is logged before its outcome is known and is never edited afterward. As outcomes resolve, we publish the prospective record, hits and misses alike, next to the original claims. If the live numbers come in materially below the backtest, this page will say so. Our goal is to measure reality accurately, not to defend a statistic.