BLOG · MARKET ANALYSIS
A sample of 312 businesses over two years. Median monthly revenue per founder ranges from $4,000 to $11,400 depending on model. Reaching $1,000 MRR within two years happens for one in six. The model you pick decides your ceiling more than your execution does.
Discussions about AI business models are usually comparisons of anecdotes. Somebody knows a person who built a directory and does well; somebody else knows a person who built an agency and burned out. Both stories are true and neither is data.
This is data. A sample of 312 businesses across Q1 2024 to Q1 2026, reported as median monthly revenue per founder by model. Medians, not means, because the mean in every one of these categories is dominated by a handful of outliers that describe nothing about typical experience.
| Model | Median MRR per founder | Character |
|---|---|---|
| Vertical one-person SaaS | $11,400 | Highest median, slowest start, no client delivery load. |
| Productized AI agency | $9,200 | Strong median, delivery-bound, does not compound without hiring. |
| Programmatic SEO directory | $6,800 | Lowest operating cost, most exposed to platform changes. |
| Newsletter / membership | $4,000 to $7,000 | Lowest median, most consistent, most dependent on audience. |
The SEO directory figure deserves a note on why its economics look better than its median suggests: operating cost is around $80 per month. A directory at $6,800 per founder is keeping most of it. An agency at $9,200 is paying delivery salaries out of it. The medians are not directly comparable as take-home.
One in six. That is the share of micro-SaaS businesses that reach $1,000 MRR within two years.
Read it as a base rate rather than a discouragement. It means the modal outcome for a micro-SaaS attempt is not $1,000 a month — it is under that, and in the first year the most likely monthly revenue is $0 to $100. Five of six attempts do not clear a threshold that most people would describe as a rounding error in their household budget, within two years.
This is not an argument against the model. The vertical SaaS median of $11,400 is the highest in the sample, and the businesses that clear the base rate do well. It is an argument for not confusing the median of survivors with the distribution of attempts.
A $11,400 median and a 1-in-6 chance of reaching $1,000 are describing the same dataset. Both are true. Only one of them is usually quoted.
The medians hide very different cost structures, and the honest comparison is per founder-hour rather than per dollar.
Vertical SaaS has the highest median and the slowest path, because a vertical product requires domain depth before it can be sold. Nothing compounds until the product is right, and there is no client to bill in the meantime. The failure mode is running out of runway before the product earns.
Productized agency has a strong median and a hard ceiling. Delivery is linear in clients, so revenue stops compounding when the founder's calendar fills. The escape is productising further or hiring, and both change the business into something else.
Programmatic SEO has the lowest operating cost and the least control. The asset is traffic, and traffic is rented from platforms that change terms without notice. A directory that works for two years can lose most of its value in an update it did not choose.
Newsletter and membership has the lowest median and the most predictable revenue, because it is the only model here where the customer relationship is direct and recurring by construction.
One category sits outside the table because its distribution is not comparable: marketplaces have the highest ceiling and the worst failure rate. A marketplace that works produces revenue no other model here approaches. A marketplace that does not work produces nothing, and the two outcomes are separated by a liquidity problem that is very hard to solve alone.
Included for completeness rather than as a recommendation. The median for marketplaces is not a useful number because the distribution is bimodal.
The model decision sets the shape of the next two years, and it is usually made quickly and revisited rarely. Three questions make it more deliberate.
One: do you need revenue in ninety days or can you wait eighteen months? The answers point at opposite rows of the table. Two: do you want to sell work or maintain a product? Agencies require continuous selling; SaaS requires continuous maintenance. Most people are suited to one and describe themselves as suited to both. Three: what happens to you in the outcome the base rate actually predicts? If the answer to a $0 to $100 first year is "I would have to stop," the model with the highest median is not the right choice.
Because the audience for it is deciding how to build, not deciding whom to buy from. Attaching a deployment check to a model-selection analysis would be a category error, and it would make the numbers look like a funnel.
The numbers are published so the decision can be made on evidence. If they change which row you are aiming at, they have done their work.
There is no offer attached to this piece. The numbers below are published so that model selection can be made on evidence rather than on which model is being talked about most.
Run the deployment checkNo form. No call. Read the distribution and pick with your eyes open.