The Score in Misano is one number, 0 to 100. It folds five dimensions of a company into a single figure that should tell you, at a glance, how strong the business is. But one number won't tell you how it got there. So I'm opening it up — what's computed by hard math, what the AI judges, and the gates that can pull it down.
I'll walk through it in full detail and finish by breaking down a real company — Barrandov Studio a.s. — dimension by dimension. Every number in this post is either from my own Misano instance or from the app's code.
Why one number — and why you can take it apart
An investor doesn't need ten metrics. They need one that ranks a hundred companies against each other. That's the Score. But collapsing everything into a single figure has one flaw: it hides what's inside. I don't want that.
So the Score is built from dimensions you can expand, and you set the weights yourself in your strategy. If risk matters to you more than growth, you pull up the weight on Risk. It's not a black box you have to take as-is.
Five dimensions, one number
The Score is a weighted average of five dimensions. The default weights:
- Finance — 37.5%
- Market — 18.75%
- Competitive advantage — 18.75%
- Risk — 15%
- Management — 10%
Finance carries the most weight on purpose — for mid-market companies, the numbers in the filings hold more than qualitative impressions. The weights are defaults; every strategy can reset them.
There's a sixth axis too — Strategy fit — but it doesn't go into the Score. I'll come back to it below, because it measures something other than company quality.
What math computes, what the AI judges
Two of the five dimensions are pure math over the figures in the filings — Finance and Risk. The same formula gives the same result every time; no model has a say. The other three — Market, Competitive advantage, Management — are qualitative, judged by the AI (Claude Sonnet) against rules I'll show you too.
The split is deliberate. Anything I can compute from data, I keep away from the model — it's cheaper, faster, and above all it returns the same answer every time. I bring the model in where no formula will do: judging whether a market is healthy, or whether a company has a defensible advantage.
Finance — computed by math
Finance is built from four sub-scores, each 0–100. They're computed from the filing figures, then averaged with these weights: growth 30%, margin 25%, efficiency 20%, stability 25%.
Each sub-score comes from linear interpolation between anchor points. Here are all of them:
- Growth (3-year revenue CAGR): −10% → 10 points · 0% → 30 · 10% → 50 · 20% → 70 · 30% → 90 · 50%+ → 100
- Margin (EBITDA margin): −10% → 5 · 0% → 15 · 5% → 30 · 15% → 55 · 25% → 85 · 40% → 95 · 60%+ → 100
- Efficiency (revenue per employee): CZK 0.5M → 15 · 1.25M → 30 · 2.5M → 55 · 5M → 85 · 10M → 95 · 20M+ → 100
- Stability (revenue-growth consistency): every year up → 90 · one decline → 60 · several declines → 30. Without multi-year data, margin stands in (≥25% → 65, ≥15% → 55, ≥5% → 45, else 35).
An example, to be concrete: a company at 25% CAGR sits midway between the 20% → 70 and 30% → 90 anchors, so growth scores 80. No magic — just a straight line between two points.
Risk — computed by math
Risk is deterministic too, and reads inverted: a higher score means a safer company. It starts at 55, then points are added or subtracted from the balance sheet and cyclicality:
- In liquidation → straight to 10.
- Equity to assets: negative → −30 · ≥50% → +20 · ≥30% → +10.
- Average EBITDA margin: ≥15% → +10 · negative → −15.
- Revenue declines: none → +10 · two or more → −15.
- Thin data → −10.
The result is clamped to 0–100. It's a first-draft rubric and I tune it as I go — which is why I'm showing you the exact thresholds, not just the idea.
Market, competitive advantage, management — judged by the AI
These three can't be computed from a table. They're judged by a single call to Claude Sonnet, which gets facts about the company plus your strategy and returns three scores, 0–100, each with a short rationale. What I ask the model for:
- Market: objective sector health — growth, market size, structural tailwinds. 50 = a stable, mature market; growing goes up, declining goes down. Independent of whether you want that sector or not.
- Competitive advantage: defensibility — switching costs, brand, regulatory risk, network effects, proprietary tech. 0 = a pure commodity, 100 = a durable, hard-to-attack edge.
- Management: leadership track record, team depth, incentive alignment. Public data is scarce, so when it's missing the model holds a neutral 50 — it doesn't guess upward.
And the one rule I care about most: the model must not invent facts. It doesn't have a list of licences or certifications in the data, so when a licence isn't there, it treats that as unknown, not as 'doesn't have it.' A company is never penalised for an unverified, presumed absence. A score drops only when the data positively shows the absence.
Strategy fit: a sixth axis, not part of the Score
Strategy fit measures something entirely different from company quality — how well it fits your thesis. That's why it sits outside the Score, as its own axis. The AI judges it too, but strictly against your strategy: focus sectors, deal size, geography, exclusions, anti-patterns.
0 means an exclusion was hit — a fund with 'never alcohol' scores a brewer a zero. 100 is a textbook fit. And mind contrarian theses: if you deliberately seek a sector despite a weak market, fit is high — it's measured against your thesis, not the market's health.
When fit drops to 10 or below, that's an exclusion and the Score is capped at 20, however strong the numbers. An excluded company shouldn't climb just because it runs a tidy business.
The hard gates
Before any average is computed, a company passes through a few gates. These take precedence over everything else:
- Not an operating company (a holding, a dormant entity) → Score 0.
- Less than a year of revenue or insufficient data → Score 0. Better no number than an invented one.
- One or two years of data → it's scored, but flagged preliminary.
- Average EBITDA margin over the last three years in the red → Score capped at 25. If the company has just turned profitable (latest year above 15% margin and a 30-point jump), the cap lifts to 50 — half credit for the turn, not full.
The breakdown on a real company: Barrandov Studio a.s.
Now I'll put it together on a company you can look up in the register yourself. Barrandov Studio a.s. has a Score of 77 in Misano. Here's what that number is made of. The financial inputs are from the public filings for 2020–2022 on Justice.cz.
Finance: 85. Revenue grew from CZK 385M to 643M between 2020 and 2022 — a CAGR around 29% → growth 89. Latest EBITDA margin 22% → margin 77. Revenue per employee CZK 4.9M → efficiency 83. Three years with no revenue decline → stability 90. The weighted average of the four: 85.
Risk: 95. Starting at 55, equity is 83% of assets (+20), average margin above 15% (+10), no revenue decline (+10). A strong balance sheet, effectively debt-free — it lands near the top.
Market 72, Competitive advantage 65, Management 50 — these three come from the AI. The rationales it returned: for market, 'film and TV production, growing CEE studio demand, streaming tailwind'; for advantage, 'historic brand, physical infrastructure, location-locked asset'; for management, 'limited public data on leadership, no founder signals' — an honest neutral 50, not a guess.
Together, at the default weights: 85·0.375 + 72·0.1875 + 65·0.1875 + 50·0.10 + 95·0.15 = 77.
And fit with my strategy? Even though it's a quality company — 30. The AI noted: 'revenue CZK 643M exceeds the 300M cap; pure-entertainment anti-pattern risk.' That's the point, made visible: a strong company (Score 77) that doesn't fit my particular strategy (fit 30). Quality and fit aren't the same thing, and Misano keeps them deliberately apart. Fit is always relative to your thesis — under a different strategy, Barrandov would land differently.
Why I'm showing this
Because a metric you don't understand can't be used to decide on money. Once you can see that finance makes up 37.5% of a Score of 77, and that a fit of 30 is only about a deal-size cap, you can work with it — pull a weight, adjust the strategy, or override the number with your own judgement.
The thresholds and weights aren't carved in stone; I tune them by how the model holds up on real companies. That's why I show them in full — including the parts that may move tomorrow.
Try it
You'll see it best on your own company. Click Get started, enter a real business, and expand the dimensions — you'll get the same breakdown you saw for Barrandov, on your data. 14-day money-back guarantee; the risk is on me.
Thanks for reading this far. And if something's missing or off in the breakdown, write to me at hello@misano.ai — I read everything.