Online is not the same as launched

A feature we had built was live for weeks. In that time it was used once. There were more than four thousand people who could have used it.

The cause was mundane: it depended on a setting that had never been switched on, so for ordinary users the button simply wasn't there. The interesting part isn't the bug. It's that weeks passed before anybody noticed — us included.

Shipping is not launching

Work feels finished when the code is live. For the business it isn't finished until somebody uses it. Between those two moments there is a gap that nobody owns: an announcement nobody made, a button placed where nobody looks, a permission nobody granted.

It's the least glamorous part of any project and the one most likely to waste the whole budget.

What the number actually told us

One use out of four thousand isn't a signal about preferences. It's a signal about plumbing. Real rejection looks different: people find the thing, try it, and don't come back. When almost nobody tries at all, the problem is upstream — they never saw it.

So the rule we took from it: before concluding "users don't want this", prove that they could have found it.

The obstacle: we tested it as ourselves

As administrators, we could see everything. The feature worked perfectly in every test we ran, because the restriction that hid it never applied to us. We were the only people in the world for whom it was visible.

Now anything customer-facing gets checked from a logged-out, ordinary account first — the way an actual customer meets it. It takes two minutes and it is the single cheapest test in this whole trade.

A launch checklist that costs ten minutes

• Open it as an anonymous visitor, and again as a normal user with no special rights
• Is there a path to it from somewhere people already go, or does it require knowing the address?
• Does anyone outside the project know it exists — sales, support, the person who answers the phone?
• Is there a number that would tell you within a week whether it's being used?
• Whose job is it to look at that number?

The last question is the one that gets skipped, and it's the one that makes the other four matter.

The honest part

Here's the uncomfortable consequence: a good share of the features that "nobody wanted" were never really offered. Before you delete something because usage is low, check whether it was ever actually visible to the people you built it for.

Before concluding that nobody wants it, prove that somebody could find it.