The credentials deck template we started from had three figures on its second slide. Fifty-plus projects. Thirty-plus clients. Ninety-eight per cent satisfaction. We deleted all three, because not one of them was true, and that deletion turned out to be the useful part of building the deck.
What it left was a harder question than the one we had been answering. Not how do we look credible, but what can we actually put on the page that a stranger is able to check.
The number nobody can check
Most studio case studies carry one figure per project. Forty per cent faster. Three times the conversion. Eight hours saved a week. It sits in large type near the top, unattributed, and it is the only quantity anywhere on the page.
The reader cannot test it. They cannot see what it was measured against, over what period or whether it was measured at all. So they discount it — not to zero, but to whatever they discount every agency claim to. The figure was meant to be the strongest thing on the page and it lands as the weakest, because it is the part a careful reader already knows to distrust.
It may be exactly true. That is not the issue. An unverifiable claim reads as marketing whether or not it happens to be accurate, and there is nothing you can add to it that fixes that.
What replaced it
One of our case studies looks at a business banking screen. It shows a company $84,500 and calls it the balance. The figure we cared about was $29,600.
As a claim that is "65% of the balance was already spoken for" — a sentence nobody can do anything with. As an arithmetic it is four rows:
- Reserved, $21,000 — tax and buffers, owed on a date that has not arrived
- Committed, $24,700 — payroll and suppliers, scheduled, not movable without breaking something
- At risk, $9,200 — one invoice, twelve days late and formally disputed
- Available, $29,600 — what survives all three
Now the reader can argue. They can say the tax buffer is too conservative. They can say a disputed invoice should not be written down at full value. They can say their own business treats committed payroll differently and would land somewhere else entirely. Every one of those is somebody engaging with the work, and none of them is available against "65% was already spoken for".
The components have to be uneven for this to work. $21,000, $24,700, $9,200 and $29,600 read as things that were counted. Four round numbers would read as things that were chosen.
Why a buyer reads it differently
Breaking the figure up changes what the page is claiming. One number claims a result. Four components claim a method: here is what we counted, here is what we excluded, here is what is left over.
That matters because a buyer is not really assessing whether your previous client got 40% faster. They are working out whether the same thinking, pointed at their problem, would produce something they could defend to their own people. They can borrow a method. They cannot borrow a result.
Say what each number rests on
The other half of this is admitting what kind of claim you are making at all. Every outcome in our case studies carries a status and the page prints it: measured, observed, client-reported, target, hypothesis or concept.
Two of the three case studies on our site are self-initiated concepts. Nothing in them has been measured, because there was no client system to measure. So those outcomes say concept, the sections on what we would test say hypothesis, and the only things labelled client-reported are the three places a client actually said something.
This felt like it was giving something away. It is the opposite. A reader who finds one honest label extends the benefit of the doubt to everything else on the site. A reader who finds a result where a concept should be stops believing any of it, and they are right to.
Which is roughly where the deleted slide ended up. Three impressive numbers removed, and what went in their place was smaller, duller and checkable — and it is the version we would send to someone whose opinion we cared about.