How SaaS Ideas Are Scored: Our Methodology
Every number on Top SaaS Ideas can be traced back to something measurable. This page explains where ideas come from, what AI is and is not allowed to do, and the fixed formulas behind each score, so you can decide how much weight to give them.
Current database: 2199 ideas across 23 industries, grouped from 2345 real conversations, signals last checked 2026-09-15.
1. Where the conversations come from
We collect public posts and comments where people describe a problem in their own words. The current sources are:
- Reddit: founder communities and industry communities where operators talk shop, such as property management, accounting, restaurants, recruiting and logistics.
- Hacker News: Ask HN threads and comments where people ask for a tool, compare workarounds, or say what they would pay for.
- Software Recommendations Stack Exchange: questions from people looking for software that does a specific job.
Collection targets phrasing that signals an unmet need, such as "is there a tool", "looking for software", "wish there was" or "how do you manage", rather than general chatter. We store the title, the link, the date and the public engagement numbers. We do not republish anyone's post; every idea is written in our own words and links back to the source for Pro members.
2. From conversation to idea
Conversations are labelled in batches by an AI model. For each one it decides whether the post describes a concrete, recurring problem that software could solve, and if so restates the problem, names the buyer, assigns an industry, suggests a descriptive product idea and the phrase a buyer would search for, and flags the product type (B2B, B2C, AI, micro SaaS, API, extension, no-code friendly) and a build complexity from 1 to 5.
The model is not allowed to produce scores, counts or evidence. Conversations describing the same problem in the same industry are then merged into one idea, so an idea with many conversations behind it is an idea many people raised independently.
3. The signals we measure
| Signal | What it is | Why it matters |
|---|---|---|
| Mentions | Separate conversations describing the same problem | One post is an anecdote; several independent posts are a pattern |
| Engagement | Upvotes and comments on those conversations | Other people agreeing is a cheap, honest demand signal |
| Search depth | How many Google autocomplete suggestions exist around the idea's keyword (0 to 10) | Autocomplete only suggests what people actually type |
| Launches | Show HN launches matching the keyword | A proxy for how many builders already tried this |
| Category depth | Autocomplete suggestions for "best" plus the keyword | An established category produces many "best ..." searches |
| Recency | Mentions in the last 90 days versus the 270 days before | Separates rising problems from old ones |
4. The formulas
All scores run from 0 to 100 and are clamped to that range.
| Score | Formula |
|---|---|
| Demand | 25 × log10(1 + upvotes + 2 × comments) + 4 × search depth + 6 × min(mentions, 5) |
| Competition | 35 × log10(1 + launches) + 5 × category depth |
| Opportunity | Demand × (1 − Competition ÷ 140) |
| Rising | At least 2 recent mentions, and the recent monthly rate is at least 1.5 times the earlier rate |
Logarithms stop one viral thread from swamping everything else. Comments count double because writing a reply is a stronger signal than clicking an arrow. Opportunity never drops to zero on competition alone, because a crowded market with strong demand can still be worth entering with a sharper angle.
5. What we do not claim
- No search volume figures. We do not publish monthly search volume estimates. If a licensed data source is added later, its numbers will be labelled as third-party estimates.
- No revenue predictions. A high opportunity score means evidence of demand and room to compete, not a forecast of income.
- No invented examples. Every idea shown on this site exists in the database with real sources behind it.
6. Updates and monitoring
New conversations are collected and every idea's signals are re-checked weekly. Ideas that are new that week, or that have just turned rising, feed the Monday alert email that Pro members receive for the industries they choose.
7. Corrections
If an idea is misclassified, a source link is broken, or a score looks wrong, email hello@topsaasideas.com. Corrections are applied in the next weekly run.
Want to see the method in action? Pull a free idea with the SaaS idea generator, or unlock the full database.