Leadmore AI reaches $30K/month. Talked to 50-100 people before building, and says the MVP should have shipped with one feature instead of three

Dollar figures below are each accompanied by an approximate yen conversion at ¥150 to the dollar.

The Indie Hackers interview with Richard Wang opens with a result: monthly revenue over $30,000 (about ¥4.5 million). The product is Leadmore AI, a tool for B2B marketers that automates the process, targeting Reddit, of discovering relevant subreddits, posting content, and tracking leads. Wang has over five years of experience in the internet industry and is currently working on three products.

Upfront disclosure: this article contains no monthly revenue trajectory, no customer count, no unit price, and no launch date. As material for a numbers-driven verification piece, it’s thin. Still, it’s worth covering because the figures that are disclosed skew toward “hours invested before building” rather than revenue, and that granularity is specific.

What’s disclosed, and what isn’t

Revenue breakdown is blank, yet pre-build conversation count is disclosed with a specific range: “50-100.” This asymmetry itself characterizes the nature of this case.

The credit system, and a balance that can be refunded

The billing model is credit-based. Users purchase credits, consumed per action — posting, commenting, exploring subreddits. In Wang’s own explanation: “unused credits can be refunded at any time. That makes the model very user-friendly.”

Compared to a flat monthly subscription, this design has two effects. From the buyer’s side, since unused credit comes back, the psychological barrier to an initial purchase is lowered. From the seller’s side, the amount of credit consumed becomes a direct measurement of whether the product is actually being used. A contract where unused balance just piles up is revenue on paper, but it isn’t a signal of continuation.

The stack is Next.js on the front end, Go and Gin on the back end, MongoDB for business data, ClickHouse for analytics, and Function Compute for background processing. The whole thing is built serverless, a choice explained as being for iteration speed.

No dramatic turning point. What worked was reordering the sequence

There’s no event in this case that can be pointed to as the moment the trajectory changed. No viral moment, no funding round, no major partnership. With launch date itself undisclosed, there’s no axis on which to even compare before and after.

Instead, what Wang repeats is a statement about the order of operations: “spend a significant amount of time on user research and direct conversations with users before you start building anything. It might be one month, or two, or three. Don’t start building until you truly understand the demand.”

And validation happened through content and operation. Before writing code, share industry knowledge, talk directly with anyone who responds, and confirm willingness to pay early. Only then does the MVP ship, in 1-2 weeks. As a time allocation, that’s 1-3 months for validation against 1-2 weeks for implementation. The ratio is inverted compared to typical solo development.

The one clear “thing I’d redo” also connects back to this sequencing. “If I could do it again, I’d cut the first MVP from 3 features to 1.” Even after three months of understanding demand, there were still two features too many, by his own reckoning.

Why does putting validation first work?

“Talk to your customers” is well-worn advice that, by itself, explains nothing. In this case, the structure that worked sits one level deeper.

Content-based validation serves both demand confirmation and customer acquisition at once. Wang’s acquisition is “largely operations- and content-driven growth”, sharing industry insight and practical knowledge on platforms like Reddit, explaining the value of what he’s building. This is a direct extension of what he was doing during the validation period. By the time validation ends, a prospect list and an early content asset have already formed simultaneously. The reason a 1-3 month validation period pays off is that it was designed as “acquisition done early,” not as “research cost paid before building.”

The product’s channel and the acquisition channel, meanwhile, coincide. Leadmore AI is a tool that automates Reddit marketing, and its own customer acquisition happens on Reddit. If the product works, the founders’ own acquisition should work too. If it doesn’t work, the product’s value is itself questionable. Having acquisition and product validation be the same experiment isn’t a generalizable pattern, but it’s a powerful one.

Underneath both sits an explicitly stated priority order of metrics. “Revenue is determined by new acquisition × conversion rate × retention. Of these, what we prioritize most is retention. If retention is weak, that usually means the product isn’t delivering enough value. Only once retention is healthy do we lean hard into acquisition efficiency and conversion.” Stepping on the acquisition pedal first risks papering over a value defect with ad spend while the numbers alone climb. Fixing this order in place is itself the mechanism.

What didn’t work

Wang names four failures: over-loading the MVP with features, starting to build before validating demand, expanding the product’s scope too early under competitive pressure. And taking a broad positioning instead of a niche one.

The third, “fighting the urge to broaden scope in response to competitors”, is treated as its own section in the interview. Following every feature a competitor ships pulls a once-narrow focus back to three features in no time. It’s in this context that Wang says: “don’t blindly chase trends. What matters is understanding your own domain-specific strength clearly, and iterating around that.”

He also says: “today’s indie developers need to build up their operational and growth capability. In many cases, operational ability can matter more than pure development ability.” It’s a hard thing for developers to hear, but the time allocation of 1-3 months validation to 1-2 weeks implementation embodies exactly that, in action.

Conditions for replication, and what this article can’t tell you

What’s replicable is the methodology: having 10+ deep conversations before building, using content to serve both validation and acquisition, cutting the MVP to a single feature, and not spending on acquisition until retention is healthy. None of it requires capital or connections.

What’s hard to replicate is whether you can land on a subject where the product and the acquisition channel coincide. Building a Reddit marketing tool and having it sell on Reddit is close to a lucky alignment. Try the same thing with a generic business tool, and the venue for validation and the venue for sales split apart, the 1-3 month validation period reverts to pure cost.

And the biggest limitation is that this article doesn’t disclose the breakdown of the outcome. How many customers make up $30,000/month, what the average credit purchase is, how much refunding occurs. None of it is known. Connecting the input of “talked to 50-100 people before building” to the output of “$30,000/month” as causation isn’t possible from this primary source alone. What can be taken away is simply the fact that, in at least one success story, most of the time went to validation, not implementation.

Related reading

• Buying a GitHub repo for $20K and growing it to $1.5M ARR: Damon Chen’s “buy, don’t build” solo development
• Zero ad spend, four people, $1M ARR. How Plausible proved a “sell on philosophy” SaaS against Google Analytics