Why We Deleted the Numbers From Our Homepage

Why did we delete the numbers from our own homepage?

Because we could not prove them. Our homepage used to carry counters and a response time promise. When we repositioned the studio in September 2026, we went through every claim and removed the ones we could not substantiate. That felt bad for about a week and right ever since.

It is an awkward thing to write about, because those counters were doing a job. They filled space above the fold and they made the studio look established.

They were also the least defensible thing on the page, and we were asking clients to be more rigorous than we were being.

What is wrong with a stat counter?

Usually nothing at all, except that it is unverifiable, which turns out to be everything. A row of numbers saying projects delivered, average score and happy clients tells a visitor nothing they can go and check for themselves. It is decoration shaped like evidence, and most buyers read it as decoration.

The deeper problem is what it signals to a careful buyer. Someone senior enough to sign a contract has seen a thousand of these, knows they are unaudited, and discounts them completely. So the counter persuades the people who were never going to scrutinise you and loses credibility with the people who were.

And there is drift. A number that was true when someone typed it stays on the page for two years without anyone updating it. At that point it is not a claim, it is an artefact.

What does the law actually expect of a claim?

More than most marketing sites assume, and the principle is worth knowing even outside the regulated categories. The United States Federal Trade Commission states in its compliance guidance that "Before disseminating an ad, advertisers must have adequate substantiation for all objective product claims conveyed, expressly or by implication, to consumers acting reasonably."

Two parts of that sentence do the work. Before disseminating, which means you need the evidence in hand at publish time rather than when challenged. And by implication, which means the impression the page creates counts, not only the literal words.

The FTC also states plainly that "Advertisers must have a reasonable basis for their product claims before disseminating an ad." We should be honest that this guidance is written for consumer advertising, and much of it addresses health products specifically. We are not claiming it binds a B2B software company the same way. We are saying it is a good standard to hold yourself to.

Does results not typical fix an outlier?

No, and this is the most useful thing in the FTC material. Its guidance says directly that "Testimonials that report results more dramatic than users can generally expect are likely to be deceptive," and adds that "attempts to disclaim dramatic results with statements like 'Results not typical' don't cure the deception."

The expectation is stronger than a disclaimer. The FTC says such testimonials "should be accompanied by a clear and conspicuous disclosure of the results a typical consumer can actually expect." Its endorsement guidance puts the same idea another way, saying an ad featuring an exceptional result "must make clear to the audience what the generally expected results of following that same regimen are."

Translate that into B2B and it is a sharp test. If your best case study result is on your homepage without context, you are implying it is normal. If it is not normal, the fine print does not save you, and more importantly it does not save you with the buyer.

Which numbers are worth showing?

The ones anyone can check. Performance metrics are the clearest example, because a visitor can verify them in a browser in thirty seconds. Publishing a load time or a Core Web Vitals result is a claim with a receipt attached.

The thresholds are public, which makes the claim legible. web.dev states that "LCP should occur within 2.5 seconds of when the page first starts loading," that pages "should have a INP of 200 milliseconds or less," and that they should "maintain a CLS of 0.1. or less."

Accessibility results work the same way. A stated score against WCAG criteria can be tested by anyone with Lighthouse or a screen reader. So can anything visible in the product itself. These are the numbers we kept, because keeping them costs us nothing if someone checks.

How should you present a client result honestly?

With the scope attached to it. A percentage on its own is a rumour with a decimal point. The same percentage becomes evidence once you say what was measured, over what period, against what baseline, and what else was changing at the same time. That context is the whole claim.

Say what you did and did not do. On one project we built to a design provided by the client, so we describe it as development work and never imply we designed it. That distinction costs us a little credit on that case study and buys credibility on all the others.

Name the things you cannot attribute. If traffic rose while the client also started a paid campaign, say so. A buyer who reads a caveat you volunteered will trust the number more than one you presented as clean.

And get permission before publishing anything specific. A result you can name a client for is worth several you cannot. We touched on the page level version of this in our notes on designing a case study page.

What do you put there instead of a counter?

Something specific enough that no competitor could have written the same sentence. A line describing exactly who you work with and what you do for them. One real result with its full context attached. A screenshot of the actual work rather than an illustration standing in for it.

Specificity replaces volume. Three lines about a genuine project convince more than four counters, because the details are the proof. Nobody invents the fact that a job portal integration linked open roles as CMS entries. That kind of detail only comes from having done it.

The space can also just be smaller. A hero without a stat row is shorter, faster and easier to read. We wrote about the general version of that in our piece on how social proof works in design, and about the wider question in our notes on designing a site that reads as authentic.

Does removing numbers cost you conversions?

We would not claim to know, and we are not going to invent a figure to make this piece land better, because that is precisely the behaviour the rest of the article argues against. What we can do is set out the trade we were actually weighing when we made the call.

What we can say is what we were weighing. An unverifiable counter might lift conversion slightly among visitors who skim. A claim we cannot defend costs us with the buyers who read carefully and with anyone who asks us to back it up in a sales conversation.

The asymmetry made the decision. The downside of a soft claim is small and constant. The downside of a claim you cannot support is rare and severe, and it arrives at the worst possible moment, which is usually during procurement.

How do you keep the numbers you do publish true?

Write down where each one came from, next to the number. We keep a list of the studio claims we can support and what backs each one, and nothing goes on the site unless it is on that list. It is unglamorous and it has caught several things.

Date the measurements. A PageSpeed result is true for a moment, not forever, because the site changes and so do the scoring tools. A number with a date on it is honest. The same number presented as permanent slowly becomes false.

Then set a review. Once or twice a year, read every number on your site and ask whether you could still prove it today. Most teams have never done this once.

What would we ask before adding a number?

Three questions, in this order. Can we show where this number came from. Would we be comfortable if a prospective client asked us to prove it live in a meeting. And is it typical of our work, or is it our single best result being quietly presented as the normal one.

If any answer is uncomfortable, the number comes off the page. That rule cost us our counters and it made the rest of the site easier to defend, which is a trade we would make again.

If you are looking at your own site wondering which claims would survive that test, we are happy to go through it with you. You can find us at phoenix.studio.