How to find SaaS ideas, and the evidence a complaint leaves out

To find SaaS ideas, read where people describe a problem in their own words — review and alternatives sites, Hacker News, Stack Exchange, the forums your buyer uses — and keep each problem with its source. That gives you candidates, each carrying one of the 5 kinds of evidence a SaaS idea needs. The other four — people paying or signing up, money made in the space, a way to reach buyers, the trend — you gather separately, idea by idea.

The kinds of evidence, and where each one lives

A Discovery run rates every idea it evaluates on the same 5 kinds of evidence. They are also what is worth knowing before you write code: a SaaS idea is a bet that a particular buyer pays every month, and each kind tests a different part of that bet. Few places answer more than one of them.

What a found idea carries

That is what an idea found from a complaint looks like, ours or yours: a description, a buyer, a problem, and where the problem was written down. Nothing in it shows whether anyone pays to have the problem solved, or how you would reach the next buyer. A rating for those can only be as good as what the record says about them.

The same gap across a whole run

Here are the ratings for all 26 ideas the run evaluated. The niche was “real estate agent and property management tools for solo agents”.

The problem was better than weak for 19 of 26. Revenue was better than weak for none, distribution for none, demand for 1, and trend was not verified for 26 of 26. Three things to know before reading more into that.

The problem row is high partly by design. Before evaluation, a gate drops ideas with no documented problem behind them: 21 of the run's 69 candidates went that way. The ideas that reached this chart are the ones that had one.

These are ratings, not measurements. One niche, one day, from a model that would not rate the same record identically twice. They say what the ideas' records supported, not how often a market has people paying.

The sweep's own searches do not close the gap. It searches for 3 of the 5 kinds in every market, and what those searches return becomes more candidate ideas. An idea taken from a revenue listing carries only what its description says about that listing; an idea taken from a complaint carries the complaint. Neither picks up the other's evidence.

The gap follows from what these records carry, not from the niche: start from a complaint and you have the complaint. The rest you go and get.

Why you still start from the complaint

Missing a documented problem is the most common reason our gate drops an idea. Across the 33 real Discovery runs we held on 2026-09-22, most of them our own, we found no documented problem behind 496 of the 895 ideas the gate dropped. The per-run version of that count lists every rejection with its reason.

How to keep a problem so it still holds up a month later — the link, the date, what you inferred — is its own page: where to find real customer problems.

Where to look for payment

Demand and revenue are the two kinds closest to money: people paying or signing up, and money already made in the space. For your niche specifically, they tend to show up in a handful of places, and none of them is a list of ideas.

• Complaints about price. A complaint that a tool costs too much often comes from someone who is paying for it. Read it as a lead on revenue, then check.
• Paid listings where your buyer shops for tools. A paid app in a marketplace shows the job is sold at a price, and reviews from people in your niche suggest some of them bought it. The one- and two-star reviews are often the complaint and the price on one page.
• Jobs posted for the manual version. If people in the niche hire someone to do the work by hand, the post usually states a budget, and a job that comes back every month is a cost a product could replace.
• Small software businesses in the niche, listed for sale. Marketplaces for small online businesses show the revenue the seller reports. A listing in your niche is revenue evidence for your niche — the seller's number, not an audited one.
• Someone paying you. Everything above is evidence that somebody else gets paid. The only payment evidence that is yours is a stranger paying, or committing to pay, before the product exists — will people pay for my idea covers how to ask without a sales call.

Where to look for a channel

Our finalists step may not recommend building an idea without distribution evidence. By hand, it is mostly a matter of naming a place:

• Name where your buyer already looks for tools. A marketplace category, a plugin directory, a launch site, a forum where people ask for recommendations. If you can name one and see paid competitors listed there, the channel exists. If you cannot name one, that is the finding.
• Search for buying questions, not complaints. “Best X for Y” and “X vs Y” are people choosing a tool. A complaint is the shape of a problem; a buying question is the shape of a channel.
• Check who owns the channel. A marketplace that brings you buyers can also remove you. Platform risk counts how often that dependency is named when an idea is dropped.

The short version

• A SaaS idea needs 5 kinds of evidence: the problem, people paying or signing up, money made in the space, a way to reach buyers, and the trend. Few places answer more than one.
• An idea found from a complaint carries the complaint. On our published run, the problem was better than weak for 19 of 26 evaluated ideas, revenue for none and distribution for none.
• Start from the complaint anyway: missing it is the most common reason our gate drops an idea, 496 of 895.
• Gather the rest idea by idea: price complaints, paid listings, jobs for the manual version, businesses for sale — and then a stranger paying you.
• Name the place your buyer already looks for tools before you write code.

The run behind these ratings is published unedited. Each finalist's ratings are under “How we scored this”; the other ideas show their score.