TL;DR: "What should I build?" is the wrong first question. Of 1,055,233 scored discussions in our corpus, only 3.55% carry strong evidence that anyone would pay to fix the problem — and those 37,495 are not spread evenly. They concentrate in payments, security operations and store operations at five to eight times the baseline rate, and they thin out in exactly the categories that feel most exciting to build in. The practical move for one person is to pick a category by its paying rate, then find the narrowest missing operation inside it.
The baseline you are working against
Every discussion our pipeline processes gets two independent scores: how badly the described problem hurts, and whether there is evidence someone would spend money on it. The second one is brutal.
| Willingness to pay | Share of 1,055,233 scored discussions |
|---|---|
| 8 or above | 3.55% (37,495) |
| 7 or above | 13.56% (143,092) |
| 1 or 2 | 44.4% (468,268) |
Nearly half of everything people complain about online scores in the bottom two bands. That is not a failure of the complaints — they are real problems — it is a reminder that annoyance is free and budgets are not. If you are picking a market by reading threads, you are sampling from a pool where roughly one in twenty-eight items has money behind it.
Which raises the useful question: is that one-in-twenty-eight evenly distributed?
It is not. Paying intent clusters.
Grouping the same scored discussions by the product category our pipeline assigns, and keeping only categories with at least 500 scored discussions so the rates are stable:
| Category | Scored discussions | Share scoring 8+ on willingness to pay |
|---|---|---|
| Payment Infrastructure | 824 | 27.3% |
| DevSecOps | 719 | 25.9% |
| Payment Processing | 1,261 | 24.7% |
| Sales Enablement | 804 | 18.3% |
| DevOps | 562 | 17.4% |
| E-commerce Operations | 747 | 15.8% |
| E-commerce Platform | 519 | 13.9% |
| Identity and Access Management | 541 | 12.2% |
| Password Management | 1,293 | 11.6% |
| Subscription Management | 1,075 | 10.5% |
| Legal Tech | 650 | 10.0% |
| DevTools / AI Coding Assistants | 790 | 9.6% |
| AI Cost Management | 541 | 9.4% |
| AI Writing Assistant | 1,092 | 8.7% |
| — corpus baseline — | 1,055,233 | 3.55% |
Two things stand out. The top of the table is infrastructure that touches money or breaks deploys. The bottom of the table is where most indie products are currently being built.
Note also that even the worst category here is more than twice the corpus baseline. That is a selection effect worth naming: any category that accumulates 500+ scored discussions is already a category people care about. The interesting comparison is between the rows, not against the 3.55%.
A caveat about these labels, because it changes how you should read the table
The category label is generated per discussion by the extraction model, not chosen from a fixed taxonomy. Across the corpus there are 220,073 distinct labels. That means the table above is not a clean market map — it is a set of the largest islands in a very fragmented naming space, and the same real category shows up under several names.
That fragmentation is usually a problem. Here it accidentally produces a stronger result: three separately generated payment-flavoured labels — Payment Infrastructure (27.3%), Payment Processing (24.7%), and Payment Processing Infrastructure (a further 183 high-scoring discussions) — independently land in the same band. Nobody reconciled them. They agree because the underlying signal is real, not because they share a bucket.
One row we removed, and why
An earlier version of this table included AI Automation Templates at 17.9%. Sampling five of its high-scoring discussions returned five near-identical complaints about the same two open-source tools being hard to set up without coding skills. That is one conversation reflected many times, not a category with 17.9% paying intent, so it is not in the table.
This is the failure mode to watch for whenever a small category posts a surprising rate. Check whether the number is made of independent complaints before you build on it.
What "buildable by one person" actually looks like
A high paying rate does not make a category solo-friendly. Payment infrastructure as a category is not a one-person business. The wedge inside it can be.
Reading the actual high-scoring discussions in the top categories, the buildable ones share a shape: a specific operation that a platform someone already pays for does not support. A few representative shapes, paraphrased from the scored material:
- A store platform creates gift cards one at a time in the admin, with no bulk import, for merchants who need thousands.
- Automated dependency scanning flags reachable vulnerabilities but cannot open the remediation pull request, so a human does it by hand for every finding.
- A payment processor is unavailable or restricted in a region, and the merchant needs a fallback path rather than a new processor.
- Attribution numbers disagree across a store admin and three ad platforms, so someone reconciles them manually every week.
- A container image ships a base OS with critical CVEs, and the deploy is blocked until someone rebuilds it.
None of these is a platform. Each is one missing operation, adjacent to a budget that already exists, with a user who can describe the problem in a sentence. That is the shape that fits one developer.
The corresponding anti-pattern is a category-sized ambition — "a better CRM", "an AI writing tool" — where you compete on breadth against teams and the paying rate is below average anyway.
How to run this yourself
- Start from the category, not the idea. Pick from the top of a paying-rate table rather than from what you find interesting. Interest is a tiebreaker, not a filter.
- Check the platform the complaint sits on. A complaint in a product's own support forum comes from someone who already installed it. We break this down in how to tell if a pain point is worth building for.
- Find the missing operation, not the missing product. Search the category for the words "no way to", "one at a time", "have to manually". The narrower the gap, the better it fits one person.
- Verify someone already pays for the surrounding thing. The wedge inherits the budget. If nothing in the category is paid for, the wedge has no budget to inherit.
- Check it is not already built. Most of the time it is, and that is good news — see how to check if your SaaS idea already exists.
You can run steps 1 and 3 against this dataset directly in the validator, or browse the scored material in the dashboard.
Honest caveats
- These are model-assigned scores, not revenue. Willingness to pay is inferred from the text of a discussion. It is a filter for where to look, not proof that a market exists.
- The categories are labels, not a taxonomy. See the 220,073-label caveat above. Treat the table as directional.
- A high rate in a category says nothing about competition. Payments scores well partly because payment problems are expensive, and expensive problems attract funded companies. The wedge has to be narrow enough that being small is an advantage.
- Our corpus is where our crawlers go. It covers developer forums, app stores, social platforms and job boards. Categories whose users do not post in those places are under-represented here regardless of how much money is in them.
Where the data comes from
PainHunt processes discussions from 22 active sources and scores each one for pain intensity, commercial intent and willingness to pay. The figures in this article come from 1,064,011 processed discussions, of which 1,055,233 carry a willingness-to-pay score, as of 2026-09-20. We link back to every original discussion and never republish its text.