How we score opportunities
Published in full, because a score you cannot inspect is a score you have to take on faith.
The seven dimensions
| Dimension | What it measures | Weight |
|---|---|---|
| Pain severity | How badly the problem hurts, judged from the language people use and the cost of the workaround they tolerate. | 0.18 |
| Competition gap | 10 means nobody serves this well. 0 means the space is crowded and genuinely well solved. | 0.18 |
| Market size | How many businesses or people have this exact problem — not the size of the broad category it sits in. | 0.15 |
| Monetization clarity | How obvious it is who pays, how much, and why they would sign it off. | 0.14 |
| Build feasibility | 10 means a solo developer ships a credible v1 in a month. 0 means a funded team and years. | 0.13 |
| Evidence strength | Computed, not judged. Derived from how many independent pain points back the cluster and how many distinct platforms they came from. | 0.12 |
| Distribution ease | How reachable these customers are without a sales team. | 0.10 |
Weights sum to 1.00. Each dimension is scored 0–10, and the weighted total is scaled to 0–100.
What a score is for
The score ranks opportunities relative to each other given the evidence we could find. It is a triage device — a way to decide what to read first out of hundreds of candidates.
It is not a prediction. Two opportunities scoring 80 are not equally likely to succeed; they are equally worth investigating. Treating a ranking as a probability is the main way to misuse it.
How evidence strength is computed
Six of the seven dimensions are judged by a model reading the cluster. Evidence strength is not, because a model asked to rate its own evidence has an obvious incentive problem.
Instead it is arithmetic on two measured quantities: the number of independent pain points in the cluster, and the number of distinct source platforms they came from. Volume contributes logarithmically — the tenth complaint from the same forum adds less than the first from a new one. Cross-source breadth is weighted more heavily, because single-community volume is the easiest signal to mistake for a market.
Known biases
- People who complain in public skew technical, English-speaking and concentrated in industries with active online communities
- Absence of complaint is evidence of under-sampling, not of an absent problem
- Review sites over-represent buyers with budget; forums over-represent individual practitioners
- Job boards over-represent work that is easy to specify, which is not always the work most worth automating
What we do about them
Weighting breadth over volume is the main correction. An opportunity backed by four platforms scores materially higher on evidence than one backed by forty posts on a single site.
The second correction is transparency: every piece of evidence is listed with its source and date, so you can see the shape of the sample rather than inheriting our summary of it.
Questions
Can I recompute the score with my own weights?+
Yes. Every dimension and its weight is shown on each opportunity page, so you can apply your own weighting to the parts. If you care far more about build feasibility than market size, the raw dimensions support that.
Why is competition gap weighted as highly as pain severity?+
Because a severe problem that three funded companies already solve well is not an opportunity for a new entrant. Severity tells you the problem is worth solving; the gap tells you whether there is room for you to solve it.
What happens when a cluster has only two pain points?+
It still gets synthesised, but evidence strength stays low and drags the composite down accordingly. The write-up is also instructed to scope the idea narrowly and say plainly that the evidence is thin.