Why I Deleted the Most Impressive Numbers From My Homepage

There was a section on the FastFound homepage that looked reassuringly like a proper technology company.

It had numbers.

Specific numbers.

Numbers are comforting.

They make a product feel measured.

“Better” is marketing.

“62%” feels like evidence.

There was only one problem.

They weren’t FastFound’s numbers.

FastFound is still pre-beta.

There are no six-month customer cohorts. There isn’t a large book of client outcome data. There isn’t enough history to publish meaningful retention performance.

The figures had come from research used while designing the business.

And we had done something that, in hindsight, looks obviously wrong.

We put them on the homepage.

We even knew they weren’t ours

This wasn’t a case of forgetting where a figure came from.

The page included an internal-style disclaimer making clear that the statistics were illustrative and should be replaced with verified client data before launch.

At first glance that sounds responsible.

It wasn’t.

If anything, it exposed the problem more clearly.

A disclaimer saying “this isn’t our evidence” does not magically make the number evidence about our product.

We were effectively saying:

“Here is an impressive result. Please notice the impressive result. Also, technically, it isn’t our result.”

That is not the standard I want FastFound to operate to.

So the numbers came out.

Removing data felt wrong

This is the strange part.

Replacing impressive statistics with nothing makes a homepage look less mature.

There is a strong instinct when building a new product to fill every gap.

Empty spaces look unfinished.

“No customer results yet” feels like weakness.

But FastFound genuinely has no client results yet.

It is pre-beta.

Why would the website imply otherwise?

One of the figures related to recovering missed opportunities.

Another related to repeat revenue.

A third compared customer churn under two different service models.

All three were useful research when designing the business.

None of them described FastFound’s actual performance.

So I introduced a much stricter rule.

No performance number is published as a FastFound result until I can write down where our number came from and what period it covers.

If I can’t write the source line, the statistic isn’t ready.

One number had an even deeper problem

The churn figure revealed something more interesting.

The business model had evolved toward selling bundles rather than a menu of individual services.

But the research statistic compared bundled customers with single-service customers.

That sounds like something we could eventually validate.

Except we can’t.

If FastFound doesn’t sell a meaningful cohort of single-service accounts, there will never be a FastFound single-service cohort to compare against.

The original statistic could therefore never become “our” statistic.

We could leave it there for five years and it would still belong to somebody else’s research.

That distinction matters.

We can eventually measure FastFound’s own retention.

We can ask:

“What percentage of customers were still with us after six months?”

Later:

“Twelve months?”

Those are real questions our business can answer.

But we cannot pretend to have run an experiment our business model does not actually run.

So that comparison wasn’t merely hidden pending better data.

It was deleted.

Evidence has to match the claim

This sounds obvious when written down.

During product development, it often isn’t.

Research numbers migrate.

They start in a market-analysis document.

Then they appear in a service specification.

Then somebody uses one in pricing rationale.

Then it becomes marketing copy.

After a few moves, a number can lose its provenance while retaining its authority.

This is especially easy when using AI.

Give an AI a large project context containing industry statistics and company information and it can produce a beautifully written paragraph in which the boundary between the two becomes fuzzy.

The prose is not malicious.

The source distinction simply wasn’t encoded strongly enough.

So I’ve become much more interested in provenance.

Not merely:

“What is the number?”

But:

“Whose number is it?”

“What population was measured?”

“Over what period?”

“Does that population resemble the thing we’re claiming it describes?”

“And can FastFound ever generate an equivalent number of its own?”

The empty state is now deliberate

The homepage now has a much less impressive position.

Essentially:

We don’t have client results to show yet.

That space is empty on purpose.

Oddly, I prefer it.

It says something I want the product itself to say later:

uncertainty should remain visible until evidence removes it.

There will be a strong temptation after beta begins to fill the section as quickly as possible.

One customer gets a good result.

Great. Put it on the website.

Except one customer’s result is a case study, not a performance distribution.

Ten customers?

Better, but now we have to ask whether the measure is comparable.

A few months of retention?

Useful, but not annual churn.

The new rule creates friction.

That’s intentional.

Marketing should not be able to outrun evidence simply because the page looks better with a number on it.

There is a commercial cost to this

I don’t assume radical factual caution is always rewarded.

A competitor may show bigger numbers.

Their homepage may feel more established.

A buyer may never inspect the methodology.

Removing an impressive statistic might reduce conversion.

But there is another cost that is harder to measure.

The moment a prospective customer discovers that a supposedly precise claim is actually borrowed, estimated or creatively framed, every other precise claim becomes suspect too.

Trust compounds.

So does distrust.

I’d rather have an empty proof section in pre-beta than spend the next year explaining why the proof wasn’t really ours.

Eventually I want that section full of numbers.

But when it is, they will have been annoyingly difficult to earn.

That’s the point.

FastFound is currently pre-beta. It does not yet have a body of client performance data, and this post describes why I intend to keep it that way on the website until genuine results exist.