A better product does not get found

Improving something nobody has seen is the most common way a small company quietly disappears, and the public data says so plainly.

Opsbench started in 2026. In the first weeks, the work that felt most like progress was making the offer sharper. Tighter terms. Clearer pages. A better answer to the question a founder would ask in the first email.

None of that was wasted. All of it was also invisible, because in that period the number of people who had seen any of it was approximately zero.

This is the distribution problem. It is worth naming early, because it is not a marketing problem that arrives later. It is a structural fact about the state anything new starts in: unseen is the default, and improving something unseen does not change its state.

The number that makes it concrete

Ahrefs studied roughly 14 billion pages in its Content Explorer index and found that 96.55% of them get no search traffic from Google at all. A further 1.94% get between one and ten visits a month (Ahrefs, 2023).

Read that as a base rate, not as a marketing statistic. If you publish something and do nothing else, the most likely outcome by a very wide margin is that it is read by nobody. Not read a little. Nobody.

And that base rate does not care about quality. It is not a filter that admits good pages and rejects weak ones. It is mostly a filter on links, on audience you already have, and on whether anyone was already searching for the thing you made.

A better page enters the same pool as every other page and inherits the same odds.

Why the assumption survives anyway

Because building is legible and distribution is not.

You can watch a product improve. You can list what changed this week and feel the week was used. Distribution work in the first months produces almost nothing you can point at. You send fifty messages and get one reply that goes nowhere.

There is no view of that which looks like progress, so the mind quietly reallocates the hours to the thing that does.

CB Insights analysed 431 companies that shut down since 2023 and found that running out of capital appears in 70% of the failures, but says plainly that this is the final cause of death rather than the root problem. Poor fit between product and market shows up in 43% (CB Insights, March 2026).

That figure is worth reading carefully, because "poor fit between product and market" is often distribution wearing a different coat. A product can fit a real need perfectly and still never meet the people who have it. The post mortem records it the same way either way.

The data on outreach is not usable, and that is worth saying

The obvious answer to "nobody has seen it" is to go and tell people directly. The next question is what a realistic reply rate looks like, so the effort can be sized before weeks go into it.

There is no trustworthy public number.

What exists is benchmark data published by the companies that sell outreach software, drawn from campaigns run on their own platforms. The published figures do not agree with each other in any useful way. Instantly reports an average reply rate around 3.4%.

Belkins has reported 5.8% from an analysis of 16.5 million emails, and a very different figure in a later report. Backlinko's outreach study landed near 8.5%.

A spread from under one percent to over eight percent is not a benchmark. It is a range so wide that it carries no information about what your specific messages to your specific list will do. Each of those numbers is also produced by a party with an interest in the number being encouraging.

So there is no reliable public reply rate to plan against. The only honest position is that you find out by running it, with a sample size large enough that the result is not noise, and you treat the first two hundred messages as measurement rather than as sales.

Three doors, and what each one actually charges

Search. Slow, compounds over time, and the 96.55% above is the price of admission. It works when you write about something people already type into a box, and when someone eventually links to you. Realistic horizon: quarters, not weeks.

Direct outreach. Fast to start, brutal to sustain, and the numbers are unknown until you generate them yourself. Its real cost is attention: it is the only one of the three that consumes the founder's best hours every single day it runs.

Borrowed audience. Someone else's newsletter, podcast, community, or client list. Fastest of the three by a distance. The constraint is that you have to be worth introducing, and in the first months you usually are not, because there is nothing yet to introduce except a claim.

Notice that none of the three is a channel you own on day one. That is the actual shape of the problem. Every early distribution route runs through something built by somebody else.

Choosing a door, and what each one costs

Search is slow and compounding. Its cost is months of writing before anything ranks, and its advantage is that the traffic keeps arriving after the work stops. Free tools belong to the same door: a thing that solves one narrow problem completely gets linked to and passed on, and a sales page does not.

Borrowed audience is the fastest door and the one nobody controls. Someone with an audience has to decide the offer is worth their credibility.

Direct outreach is the only door that opens on demand, and the only one where volume is entirely in the sender's hands. It is also the one most founders avoid, which is exactly why it stays available.

The trap in the three doors

The fastest door is gated by the slowest. Almost everything that earns borrowed credibility in a first year is either a public track record that does not exist yet or a body of work large enough to be evidence in itself, and both take months. So the fast route is only fast once the slow time has already been spent.

Paid acquisition looks like the way around that and mostly is not, early on. It tells you what a click costs but not whether the offer converts, and before a first customer those two failures are indistinguishable. Paid works as an amplifier for an offer that has already converted somebody, not as the test of whether it can.

The certain part is the negative claim, and it is the one worth carrying: no amount of further work on the product changes any of this. The product being better is not a distribution event.

What to do this week

1. Write down where your last ten conversations came from. Not leads, conversations. If more than seven came from one source, you have one channel and no redundancy. If you cannot reconstruct the list, that is the first finding.

2. Pick one door and commit four weeks to it. Not three doors at a third of the effort each. Four weeks is roughly the minimum before any of them produces a signal you can read.

3. Set the sample size before you start. If it is outreach, decide the number of messages up front, two hundred is a reasonable floor, and agree with yourself that you will not judge the result before you hit it. Otherwise you will quit at message forty on a bad Tuesday.

4. Count the hours the product got this week against the hours distribution got. Just the two numbers. Most founders in month three find a ratio they did not expect and would not defend out loud.

5. Stop improving anything nobody has seen. Freeze the product at good enough for one week and spend the whole week on getting it in front of people. If that week feels wasteful, the ratio in step four is the reason.