Strategic Fit Engine
One of the tools we have built. It answers a question that usually costs a deal team weeks: who should buy this company, or what should this buyer acquire. It gets there by deriving the answer from the counterparty's own behaviour rather than from a generic screen.
Most target lists are built backwards
The usual approach starts with the universe (everything in the sector, in the geography, in the size band) and narrows it with judgement applied at the end. The criteria that actually decide the outcome never get written down, so the list cannot be argued with and cannot be rerun.
This engine inverts that. It spends its first pass entirely on the counterparty: what they have bought, at what size, to fill which gap, and what their competitors just took off the board. The scoring rubric falls out of that reading, so every rank has a stated reason behind it, and changing the counterparty changes the criteria rather than just reshuffling the same list.
The same engine, pointed either way
Two real runs. Switch between them to see how the rubric, the universe and the output all change while the method stays identical.
A strategic buyer wants to know what to acquire. The engine reads the buyer's own deal history first, then scores the market against it.
- Buyer
- Duolingo
- Sector
- Coding education and developer learning
- Geography
- Europe: UK, Germany, France, Nordics, Netherlands
Three sub-$50m acqui-hires of creative and gaming studios: a product-first buyer, not a revenue buyer.
- Acquisitions read
- NextBeat 2025 · Hobbes 2024 · Gunner 2022
- Pattern
- Talent and IP tuck-ins under $50m
- Deal range
- $20m floor · $50–200m sweet spot
- Gap identified
- No presence in coding education
Mobile-native code execution
Codecademy remains desktop-first; a mobile code runtime would leapfrog the market leader's core weakness.
Breadth of language coverage
Codecademy's 40-language catalogue is its main defensibility; entering narrow means entering behind.
AI tutoring integration readiness
A target already running LLM feedback slots into Duolingo Max without a rebuild.
Gamification loop alignment
Streaks, XP and bite-sized lessons are the buyer's actual product architecture, not a nice-to-have.
C5–C8 are fixed: technology and IP, market position, team, and legal and regulatory risk. Only C1–C4 change with the counterparty.
30–40 primary · 22–29 monitor · below 22 deprioritise. Totals are recalculated in code, never taken from the model.
Mimo
Germany · Series B · $10m ARR · $28.6m raised
Mimo is the closest structural match to how Duolingo already builds: a mobile-first, gamified daily-lesson product with in-app code execution and 15m downloads on $28.6m raised. It scores top of the list on the criterion the buyer cannot compromise on, and lowest on the one it can build.
Every score carries a sentence of reasoning in the full output. Figures not publicly available are marked as such rather than estimated.
It augments the banker, it does not replace one
The same principle that governs everything we build applies here: an output is only useful if you know what it is worth. So the engine is explicit about the two things it genuinely compresses, and the two it does not touch.
Figures shown in the example above are drawn from real runs and are illustrative of the output format. Company data in any run is model-derived unless individually verified, and is labelled accordingly in the delivered report.
Want this pointed at your mandate?
The engine takes a counterparty, a sector and a geography. If you tell us yours, we can show you what comes back before you commit to anything.