How it works

The data source

Every startup on the map comes from TrustMRR, a directory of bootstrapped founders who chose to be transparent about their revenue. The MRR numbers are self-reported and verified.

This means the map only shows founders who opted in to transparency. A sparse area does not necessarily mean an untapped opportunity. It could just mean fewer transparent founders built there.

How the clusters are built

Each startup description is run through a language model that converts text into a list of numbers representing its meaning. Two descriptions about similar products will produce similar numbers.

Those numbers are then compressed into two dimensions using a technique called UMAP, which tries to keep similar products close together on a flat plane.

Finally, an algorithm called HDBSCAN identifies dense regions in the resulting cloud of dots and labels them as clusters. The cluster names are generated by a second language model that reads the startups inside each cluster and picks a descriptive label.

How to read the map

• Dot position: dots that are close together represent products with similar descriptions. Clusters are pockets of the market where many founders are building similar things.
• Dot size: larger dots mean higher MRR. A big dot in an uncrowded area can hint at an underserved niche with real willingness to pay.
• Cluster outlines: the faint shapes around clusters are the convex hull of all startups in the group. They disappear when you zoom in.
• Uncategorized dots: some startups are too different from any cluster to fit neatly. They appear as dim dots scattered around the map.

Limitations

• The map reflects TrustMRR, not the full market. Consumer apps, VC-backed startups, and founders who prefer privacy are not here.
• Cluster labels are generated by a language model and can be imprecise. The label is a best guess, not a category definition.
• Revenue data is updated periodically. Some numbers may be weeks old.