The number in your copy that nothing can check

We hit this on our own site this morning. A page of ours opened a section with the heading “Four failure modes, four different fixes”, and then described four of them, accurately. That week the product had grown a fifth. Nothing on the page had changed, nothing on it was misspelled, and every sentence under that heading was still true. The heading had become wrong anyway, because it was not describing four things. It was claiming there were only four.

Why this one survives every check you run

A wrong price is caught because a price has a counterpart. Somewhere a field holds the real figure, and comparing the two is mechanical. The same is true of a dispatch window, a stock level, a weight, a size. This is the ordinary, tractable kind of wrong, and most of the tooling a shop ever buys is built to find it.

A count of a set has no such field. There is no column in your catalogue called “number of ways to wear it”. The three in “three ways to wear it” was arrived at once, by a person, by looking at the set and counting it, and then it was written into a sentence and detached from the thing it counted. From that moment the set and the sentence have no connection at all. The set can grow, shrink or be reorganised and the sentence will not move, because nothing is holding it to anything.

That is what makes it behave like a promise rather than a number. It has the shape of a fact — a digit, or a number word, sitting in a heading, looking checkable. It reads as the most verifiable sentence on the page. It is one of the least.

Where they are in your shop

They cluster in the places where someone once wanted to sound organised, which is why they are usually in headings and opening lines rather than in body copy.

• Counts of your own range: “our four core fabrics”, “the three finishes”, “all six colourways”. These go stale upward every time you add a line, and downward every time you quietly stop stocking one.
• Counts of what is included: “five things in every box”, “a two-part set”, “three care cards”. These are the expensive ones, because the customer counts them on arrival.
• Counts of steps: “in three steps”, “our four-stage finishing process”. A step gets added to the process and nobody thinks of the page that describes it.
• Counts in alt text and image captions, which no one re-reads: “the four sizes side by side”, under a photograph that now shows five.

Finding them costs one pass with the find box in your admin. Search your product descriptions and page copy for the number words — two, three, four, five, six — and for the bare digits beside a plural noun. Most hits will be harmless (“two weeks”, “100% cotton”). The ones that matter are the hits where the number is counting something you own and could add to.

The fix is not to remember

The obvious response is to resolve to update the count whenever the set changes. That is the one fix that does not work, and it is worth being blunt about why: the moment you add the fifth fabric you are thinking about the fifth fabric, its photographs, its stock and its price. You are not thinking about a heading written two years ago on a different page. Nobody is. That is not a discipline problem, it is the reason the sentence went stale in the first place.

Two things do work, and they are both duller than remembering.

• Stop stating the count. “Ways to wear it” says everything “three ways to wear it” says, minus the claim that goes false. You lose almost nothing: the reader can see how many there are, because they are listed underneath. The number in the heading was never informing them, it was reassuring you.
• Or bind it to the set. If the page is generated from your catalogue, have the heading count the list it is standing in front of rather than carry a number typed by hand. Then the only way the sentence can be wrong is if the list is wrong, which is a problem you would have anyway.

We took the second option where we could and the first where we could not, and we added a build check that reads any stated count out of the finished page and fails if the set behind it has grown. That is more machinery than a shop needs. The judgement underneath it is not: a number that counts something you control is a promise about a closed set, and you should only make it where something other than your memory is keeping it closed.

This is the awkward edge of something we have written before. A promise in your copy is harder to check than a number in it — that is the general rule, and a count of a set is the exception that proves it, because it is a number that behaves exactly like a promise. It is also another way a list of yours goes stale without anyone touching it.