The zero-result searches your dashboard doesn't show you

If you open your store's search report and the "no results" counter comes out low, the natural conclusion is that your search is working well.

It's almost always wrong. And the worse part is that the better the search engine, the more wrong that number becomes.

This post explains why, which signal actually works, and why plenty of people see an empty search report while their customers search every day.

The zero-results counter lies by design

Modern search engines, ours included, don't leave the screen blank. If what you typed matches nothing, they fill in: they show you the closest thing, the best seller, whatever was in the category that seemed right. It's the correct decision for the shopper; an empty screen is a customer walking out.

But it has a consequence almost nobody warns you about: if the screen is never empty, the "zero results" counter almost never goes up. Your report says 0.3% and you relax.

Meanwhile, the customer who searched for something you don't have saw a screen full of things they didn't ask for and left with a worse impression than if you'd told them you didn't stock it. That case appears in none of your metrics.

The signal that does work: the fill

What has to be counted isn't "how many searches returned zero products", it's how many searches were answered by filling in. That is: how many times your search engine failed to find what was asked for and papered over it.

That's the number that hurts and the one you can work with. In Findalo it's the one shown as genuinely no results, and it's built that way on purpose.

One detail matters more than it looks: that same number is the one we send to your analytics. If your dashboard says one thing and your Google Analytics says another for the same event, you end up trusting neither. They have to agree, and they do.

The second blind spot: the search your analytics never sees

This one changes entire decisions.

Search engines that open results in a layer over the page, without reloading or changing the address, are convenient and fast. They're also invisible to classic analytics, which learned to count searches by reading a URL parameter.

The result is that you open your internal search report and it's empty. From there to "nobody uses my search" is one short step, and a very expensive one: people deciding not to invest in their search based on a number that doesn't exist.

The check takes ten seconds: search for something in your store and watch the address bar. If it doesn't change, your analytics isn't finding out unless someone wired it up by hand.

We send search, result-click and add-to-cart-from-search events precisely so this doesn't happen. But if you use something else, check: it's the most widespread measurement failure we've come across.

Why the industry figures contradict each other

You'll find advice saying 10-30% of searches with no results is normal, and other advice saying anything above 5% is a serious problem. Both can be right, because they aren't counting the same thing.

• By unique query, the percentage comes out high: the long tail is full of odd queries searched once.
• By session or by volume, it comes out much lower: frequent searches nearly always find something.

Before comparing yourself to a figure from the internet, check which of the two your tool is using. And if you're going to set a target, set it against yourself last month, which is the only comparison that means anything.

What to do with the list

The list of what gets searched and not found is, by some distance, the most useful report any search engine produces. You work it by sorting into four piles:

1. You have it, under another name. The most frequent case and the most profitable. It's fixed with a synonym and the effect is immediate. "Cordless vacuum" when you call it "stick vacuum cleaner".

2. You have it, but the product won't let itself be found. A product page with a name and not much else, or a product marked as not visible in search. That's fixed in the catalogue.

3. You don't have it. This isn't a search failure: it's your buying department getting a free report of unmet demand, sorted by how many people asked. We've seen it justify bringing in whole ranges.

4. Noise. Your own tests, crawlers, people typing anything. You ignore it, but it's worth a look once: sometimes the "noise" is a pattern of someone trying something.

A fifteen-minute routine, once a month

1. Open the list sorted by volume and keep the top twenty.
2. Sort them into the four piles above.
3. The synonyms from pile 1, do them now. They take minutes and apply immediately.
4. Pile 3, send it to whoever buys. No dressing it up: the list, as it is.
5. Next month, check whether last month's twenty have dropped.

That's it. You don't need a project or a dashboard: you need to look at it once a month and do something about it.

And a warning about the trial month

This list starts being worth something when there's volume. With the 1,000 searches of the trial month you'll see that it exists and how it's used, but you won't draw business conclusions from it.

If that's exactly what you're after — knowing what people are asking for and you don't have — the honest answer is that you need a few weeks of real traffic. Start the trial and give it time.