The revenue distribution for small software businesses has an extremely long floor. Across a dataset of 3,787 micro-SaaS products that report public billing figures, the median product collects about $145 a month, the average sits near $4,298, and only about 6% clear $10,000 in monthly recurring revenue. That spread explains more about the category than any trend list. A thin top slice pulls the mean upward while the majority of products sit close enough to zero that their founders are effectively paying for a hobby.
The binding constraint moved this year. Building a working product no longer consumes a solo founder's first two quarters, because Lovable, Cursor, Replit and v0 absorbed most of that work. What's left is distribution, inference cost and support, three expenses that scale with usage and none of which gets cheaper because the code was easy to write. More founders now ship into the same narrow niches, so the fight is over attention and price rather than features.
Buyers hardened at the same time. Roughly $285 billion was wiped from software company valuations in February 2026, according to investment bank analyses of the sector, and Atlassian reported its first decline in the number of people using its software. Procurement teams started asking what a tool delivers per dollar instead of how many seats it covers. For anyone selling small software, that question now arrives in the first email.
What a Median Product Has to Sell
Take the floor seriously. At $29 a month, $145 is five paying accounts. Reaching $10,000 at that price takes roughly 345 accounts, and at the 8% median free-to-paid conversion that ChartMogul measured across 200 software products in January 2026, it takes about 4,300 trials to get there. A founder closing four trials a week is looking at two decades of work before the business pays a salary.
That math is the argument for higher prices, which is why the November cohort clearing the threshold looks nothing like the indie-hacker template. A $199 plan reaches $10,000 with 50 accounts, and 50 accounts is a list one person can hold in their head, call quarterly, and keep alive on a Saturday. The product can stay narrow as long as the buyer is a business with a budget line rather than an individual with a credit card and a short memory.
Churn punishes everything else. A product losing 4% of its accounts a month is down about 39% of its base after a year, so $10,000 in MRR demands close to $3,900 in new revenue annually just to stand still, before growth. Onboarding effort and retention became the real product work in 2026, the same conclusion behind the shift away from growth at any cost that reshaped the bootstrapped end of the market this year.
Seats Stopped Tracking Value
Pricing followed the workload. About 60% of software companies now run some form of usage-based pricing, with roughly 46% on a hybrid base fee plus meter, according to Causo's H1 2026 pricing report. The shift isn't ideological. Once software performs work instead of hosting a place to type, the number of humans holding a login says nothing about what the customer consumed.
An agentic tool that drafts 400 invoices or reconciles a month of bank lines can triple its output inside one account without adding a user. Charging per seat there is a donation of margin, because cost of goods climbs while the bill stays flat. Founders who sold $29 all-you-can-use plans in early 2026 are the ones quietly introducing caps and overage rates this quarter.
The counterweight is administration. A meter needs reconciliation, dunning, dispute handling and a customer finance team that can forecast a bill, and that machinery costs more attention than it returns below roughly $10,000 in MRR. Under that line, a flat price with a hard usage cap and an honest upgrade path beats a clever meter, which is also where the interests of a bootstrapped operator and a funded one split, a divergence that runs through the unspoken terms of venture money. Above the line, the meter is how you hold 60% margins on a product whose costs grow with every outcome it delivers.
The Cost Lines That Stay Out of the Price
A $29 plan that runs 900 model calls a month per customer carries real cost of goods. At current commercial rates for a mid-sized model, those calls cost roughly $12.60 in inference, which is 43% of the sticker price before a single retry. Add a 20% overhead from failed generations and the same account is at 52% of revenue in compute alone, before the payment processor takes its cut. Here is how four price structures compare on a modelled 400-account product base:
| Price structure | Revenue per account | Direct cost per account | Gross margin |
|---|---|---|---|
| Flat seat at $29 | $29.00 | $18.29 | 37% |
| $19 base plus $0.02 per action | $37.00 | $19.40 | 48% |
| 1.5% of approved transaction value | $46.50 | $19.70 | 58% |
| Flat $199 with a 4,500-action cap | $199.00 | $86.17 | 57% |
Direct cost here covers inference, a 20% retry allowance, loaded support time at $5.75 an hour-equivalent per ticket, and a 3% payment fee. Four costs get left out of most pricing sheets anyway:
- Inference burned on failed generations and retries, which rarely appears on a founder's dashboard.
- Support time per account, because onboarding a $29 customer and a $199 customer consumes the same hour.
- Refunds and chargebacks, which run near 3% plus a fixed fee before a single dispute is opened.
- Engineering hours maintaining model integrations, which break more often than a stable API contract.
The pattern in that table matters more than the headline price. The same consumption costs the same $15.12 in compute whether the customer sits on the flat plan or the outcome plan, yet gross margin moves from 37% to 58% on the billing mechanism alone. Raising the flat price to $39 recovers some of the gap and costs conversions. Changing the meter costs nothing except one uncomfortable conversation with an existing customer.
Where Price Holds Beside Money and Rules
The niches that held price through 2026 share one trait. The buyer's alternative is a penalty, a lost deal, or a manual hour someone already resents. Vertical compliance tools qualify, and so do transaction-adjacent products taking a slice of recovered revenue, and the unglamorous software that keeps a property portfolio's maintenance and filings in order. None of these compete against another app for the buyer's attention. They compete against a fine or a spreadsheet.
Agentic workflows with an approval gate are the most interesting of the group. When an agent proposes a refund, a purchase order or a contract change, a human has to sign it, and the record of who signed what becomes the product. Audit trails, permissions and version history are dull features a finance director will pay for monthly, because the alternative is explaining an unlogged decision to an auditor in a room with no coffee.
Geography shapes the price as much as the category does. Invoice and tax data now has to arrive in structured form for authorities across Europe, Kenya and Nigeria, on deadline, with a receipt attached. That turns a chore into a workflow with a legal deadline, which is the closest thing to guaranteed demand in small software. It also means a dollar-denominated $29 price is the wrong instrument in most of those markets, where a local-currency rate pitched against local wages converts far better than a global price list built in San Francisco.
The $145 median isn't a market failure. It's a verdict on products that solved a preference rather than an obligation, priced for a buyer who could cancel on a Tuesday and feel nothing. The 6% that clear $10,000 did not outbuild anyone in November 2026, and most of them shipped less software than their competitors. They picked a meter the customer could watch moving in their own favour, and then charged enough that a failed month didn't put them out of business.
Strategic Takeaways
- ✓Only about 6% of billing-tracked micro-SaaS products clear $10,000 in monthly recurring revenue, and the products that do charge for completed work instead of logins.
- ✓A usage meter improves margin only once inference and support costs are large enough to eat a flat subscription, which in practice means above roughly $10,000 in MRR.
- ✓The niches holding price in November 2026 sit beside a payment, a compliance deadline, or a workflow where a wrong answer costs the buyer money.

Maleek Joshua
Business & Work Editor at Largern, writing about markets, entrepreneurship, careers, and the scenes behind how companies and people make money.
Comments (0)
No comments yet
Be the first reader to share your perspective on this story.



