I ran 2,232 startups through 4 filters to find markets worth entering. 13 survived.

I have a notes app full of startup ideas and no way to tell which one is worth three months.

The usual advice - "find a problem people have" - never got me past that. Neither did competitor research, which for me meant two days of Googling and a spreadsheet that ended in a vibe: "seems crowded" or "probably nothing there." Both are just the feeling of not having numbers.

So I built the filter I wanted. Here it is, concretely enough that you can run it yourself with any revenue dataset.

The four filters

I have monthly revenue for 2,232 indie startups. I grouped them into 535 real sub-niches - not "AI," which is useless, but "AI video editing," "backlink platforms," "specialized developer APIs." Then:

535 → 13. That last step is the brutal one - 84 down to 13 - and it's why "low competition" on its own is a trap. Plenty of markets have no leader because there's nothing there to lead.

Worth staring at: 171 of the 323 markets are flat or shrinking. Over half. And the most fragmented markets in the whole dataset - leader holding 7–13% - are almost all AI, and almost all flat. Twenty companies splitting $7k/month is not an opportunity, it's twenty people who all read the same Twitter thread.

The 13

Notice these are small. $35k/month across six companies is invisible to a funded team and life-changing to one person. That's the point of the size floor being $20k and not $2M.

How to actually read one of these

Take "Organic Search, Backlink & PR" - 9 companies, $165k/month, leader holds 38%, median growth +36%. Drill into the companies and it stops being a market and starts being a situation:

The leader is losing a third of its revenue while #3 compounds at +610% and closes on #2. That's not a stable market with an incumbent - that's a market actively changing hands right now.

That's the read that took me two days of Googling before and takes 30 seconds now.

What I'd do next with that row

1. Sign up for all 5. Note pricing and where each one is clunky.

2. Find why Presscart is bleeding - churn, pricing, an outage, a Google update. That reason is either your wedge or your warning.

3. Search the pain-point feed for what their customers actually complain about, and check whether it's the same thing killing #1.

4. Only then decide if you want to spend a year on backlinks.

Market → companies → complaints. I had that backwards for years: fall in love with an idea, then hunt for a market to justify it. That's how you build something nobody was going to pay for.

Two things worth stealing if you build this yourself

Verify your data source by hand before you build on it. The revenue API I use documents a money field ambiguously - read wrong, a company doing $35/subscriber looks like it's doing 35 cents. I only caught it by checking one company I could verify manually. Two of its own doc pages also disagree on whether a growth field is a percentage or a decimal.

Put a revenue floor on any growth ranking. My first "fastest growing" list opened with a company at +1,218,899%. That's $1 → $12k. Sorted descending, a percentage column is just a list of rounding errors.

All 535 markets are live on PainBase now, filterable by every column above, with the companies inside each one. The pain-point feed that was already there is now step 3 instead of step 1.

If you're stuck choosing between ideas right now: reply with the space you're looking at and I'll pull its leader share, size and growth.