Three mistakes we make on product pages

The incomplete value, the vague value and the false value. All three keep you out of searches you are already paying for with ads.

Last week, close to midnight, I bought a small Playmobil figure on eBay. The listing title read ZULU MASSAI AFRIKANISCH INDIGEN.

Zulu is South Africa. Maasai is Kenya and Tanzania. Both are in Africa, and that is where the resemblance ends. Between them sit three thousand kilometres and two languages that share nothing.

The seller had stacked both words into one line and added “African indigenous” at the end, in case neither of them worked.

He was not being careless. He was being findable.

I searched for those figures and got back exactly what somebody had written in. Words that could describe half a continent.

So I asked an AI assistant a different question. I wanted figures that clearly represent one country, by culture and by origin. People, not animals. I said all of it out loud, including the part I did not want.

The answer came back as set numbers. References, catalogue entries, a series and a year. The first question returned a continent. The second returned a record.

Nothing clever happened in that second search. I filtered. Category, attribute and exclusion, because saying what you do not want is also a filter. The same thing I do on Ricardo or Galaxus, the Swiss marketplaces I buy from without ever thinking about it, except that there the filters are already built.

What changed is not the searching

Every marketplace trained us the same way. The shelf has no end, so you narrow before you look. The closer you get to what you actually want, the better what comes back.

Google always let us do this. Quotation marks, a plus, a minus. The filters were there the whole time and nobody taught them, so people spent twenty years typing two vague words into a box and scrolling instead.

What changed is not the filtering. It is that you can now do it out loud. That is the whole shift. Not intelligence. Filters.

We are using two things at once. A search engine looks everywhere and hands you a list. A shop looks only at its own shelves and hands you filters. The assistant does both.

Your website is now a product listing inside a shop you do not own, beside competitors you never agreed to share a shelf with, sorted by details you may never have given anyone.

A filter needs something to filter on

When somebody tells me Google is not indexing their products, the first thing I do is open their product pages. Almost always I find the same three mistakes.

1. The incomplete value: 50x49x61

On one page the whole specification was this line:

DIMENSIONS: 50x49x61

Height, width and depth in an order nobody wrote down, in a unit nobody named. In millimetres it is a fitting. In centimetres a box. In metres a room.

One letter decides which search you turn up in, and that letter is not on the page.

This is not a rare case. It happens with clothing sizes, with shoe sizes, with street addresses. The number on its own says nothing.

2. The vague value: “African”

If your catalogue describes a product as “African”, there is nothing to narrow. If it says Zulu, and shield, and spear, and beadwork, and the year, every one of those words becomes a handle. You are not writing description. You are building the drop-downs your buyer is already using in their head.

And they have to be on the page. Not three clicks away, not implied because everybody already knows Zulu is African. The assistant reads what it landed on. Whatever you left for later, it cannot say about you.

3. The false value: a medieval Maori

This is the worst of the three, because the field is filled in and the work looks done.

In another listing from the same search, a Maori warrior was filed under motif Medieval, inside the Pirates category, under accessories. Country of manufacture Germany according to one field, country of origin Malta according to the field beside it. The title was written properly. The values that get filtered were not.

Somebody once told me that mixing categories helps, that it lets the customer compare and discover other options. That is half right, but only in the recommendations block, where the customer does not know yet what they want.

Not on the product page. The product page is where you say what this is. Putting it on the shelf next door does not create discovery. It pollutes the filter for everyone who actually wanted a pirate accessory.

Underneath that idea sits an assumption: that the customer has the patience to look around. They used to, because they walked down the aisle. Now the first one looking is a machine, and a machine does not wander. It compares.

I had no patience at midnight either.

A missing value leaves you out. A false value puts you in somebody else’s drawer.

Twenty years in, we still do not say what we are selling.

Where this shows up in the budget

Most of the near-empty product pages I have analysed, I reached through an ad. The company had paid to put that page in front of me, because the page could not manage it on its own. Organic takes longer, so it kept getting postponed.

That spend is rent. You pay it every month for the sentence you never wrote.

One company, millions of pages, one description

I worked once at a company with millions of products in its catalogue. They all carried the same description. Picture somebody who receives a million letters with identical text and has to decide which one to answer first. That is what we were handing the search engines.

I flagged it early, before the site moved to its new platform. What I asked for was not copywriting. It was an automation that would put each product’s own terms on the page as text. Material, size, thread, tolerance. Nothing extraordinary. The words a customer filters by are the same words that customer types. Not a coincidence. It is the same person doing the same thing in two places.

It did not make the first release. The migration pulled the traffic down, and the millions of identical pages were one of the heavier reasons why: Search Console counted them for me week after week while Analytics showed what they cost. Opening both of them scared me.

I got it approved nine months later.

Indexing started moving within weeks. Within a month the channel was back where it had been before the migration.

Nobody had decided to write them all the same way. It was easier to repeat the previous one than to write the next one.

Roughly is not a fact

I spent sixteen years in newsrooms before I touched a keyword. A story with a named source and a document went on air. A story with three people saying roughly the same thing did not. Not because the three were lying. Because roughly is not a fact.

Vague words sell once. I have been on the selling end of that. Someone from the place I had just named picked the thing up and told me, politely, that it was not from there. The sale was real. The trust was not.

What you can look at today

Open your catalogue and read one entry as if you were a stranger with a filter in your hand.

Count how many things you could narrow that product by. Then open the site where your competitors sell, and count the filters running down the left side of the page.