Changing search engine without switching off the one you have

Changing the search engine on a store that's taking money is frightening, and rightly so. It's a component that touches the home page, the categories, the product pages and the phone, and if it breaks it breaks in front of customers.

The good news is that you don't have to switch anything off to try the new one. You can have both running at once, compare with your own data, and only turn off the old one once you're convinced. This post is that method, in seven days.

It works whether you're coming from another provider or from your platform's own search.

Day 0: before touching anything, write down where you're starting

Everybody skips this step, and it's the one that makes the comparison mean something. Note down, for the last four weeks:

• How many sessions use search. If you don't know, check whether your current search changes the URL: if it doesn't, your analytics is probably recording no searches at all.
• What proportion of searches find nothing. And by what definition your tool counts it, because they don't all count the same.
• Your twenty most frequent searches. That list is your test bench: they're the ones you can't afford to make worse.
• What happens after searching. How many people who search end up buying, compared with those who don't.

Keep it anywhere. Without it, in a week's time you'll be deciding on gut feeling.

Day 1: install without publishing

You install the module and connect the store. The catalogue syncs and the new search engine starts to exist without any customer seeing it: the widget isn't active until you put it there, so the usual one carries on serving.

There's only one thing to check here, and it's the one that most often goes wrong: that the number of indexed products matches your catalogue. If you have 4,000 SKUs and 3,100 show up, something is being left out — products not visible in search, a language that didn't come through, an excluded category — and it's worth sorting out before comparing anything.

Days 2 and 3: the blind test

Take your twenty most frequent searches and run them through both engines. Note which one wins each. And add five more that you know fail today, because that's where it gets decided.

Three rules so the comparison doesn't get contaminated:

• Do it on a phone as well. That's where most people search.
• Have someone other than you type them. You know the exact names; your customers don't. Ask whoever answers the phone at your company.
• Look at the first result, not the list. Hardly anyone goes past the fifth.

If the new one doesn't win clearly at this point, stop: you have your answer and you've saved yourself the entire migration.

Days 4 to 6: live, in front of everyone

Now for real: you switch the new search engine on in the store and leave the old one installed but out of the way. Three days is enough to see real behaviour.

What to look at each morning:

• That the share of searches finding nothing hasn't gone up. By the new definition, not the old one: make sure you're comparing the same thing.
• The searches that found nothing, one by one. For the first 48 hours this list is gold: nearly everything on it is fixed with a synonym in two minutes.
• That search looks right in your templates, on a phone, and on the odd pages everyone forgets: the cart, an empty category page, the result of a filter.

Day 7: you decide

You compare against what you wrote down on day 0 and you decide. If you stay, you uninstall the old one. If not, you uninstall ours and nothing has happened: your store goes back to how it was, because it never stopped being there.

What you always lose when you switch

Frankly, because nobody usually tells you this.

The analytics history doesn't travel. Not ours when you leave, and not the previous one's when you arrive. You start counting from zero. If your current search engine has two years of data, export what you can before switching it off, because afterwards you won't be able to.

The fine configuration has to be redone. Synonyms, rules, featured products, relevance weights. The concepts are similar across providers, but there's no automatic import that works well; what we do is look at your current setup together and rebuild whatever makes sense. Synonyms are usually the only thing that transfers as-is.

And the look will change. It can be made very similar, but not identical. If your store has a carefully crafted design, allow some time for fine-tuning.

Questions that always come up

Can you really have both at once? Yes, they're independent: one lives in your platform's module and the other in ours. The only thing you can't have is two widgets fighting over the same search box, and that's controlled by deciding which one is active.

What if I stop halfway through the month? There's no minimum term and you cancel from the dashboard. Nothing is charged during the trial month, so you can do the seven days without having paid anything.

Will you help? Yes, and we don't charge for it. We do the installation and the tuning with you, whatever your size.

Does it affect SEO? Internal search result pages aren't usually indexed, so the direct impact is minimal. What does change is behaviour in your store, and Google notices that from another angle.

Where to start

With day 0. Seriously: write down your four numbers before installing anything. It's half an hour, and it's what separates a decision from a hunch.

When you have them, start the trial or write to us and we'll set it up with you.