Tracking (and fixing) site search

Hardware Store Aisles

Nobody goes to a hardware megastore for fun, we usually need something specific. We wander down the aisles ‘if the adhesives are here, then that must mean tile adhesives are here right?’ until we give in and ask a disinterested looking pimply youth for help. They begrudgingly motion in the general direction of where you should find your grail. Now imagine there’s no pimply youth to help you. That’s what most e-commerce sites are like.

Riffing off of a recent Seth Godin post where he points out “Broken search = no sale”…

Tracking site search (whether you’re running an e-commerce site or not) is an important place where you can go the extra mile to make sure that your audience/customers/readers are finding what they want.

A lot of people just browse a site, navigating to what they want by clicking through your (hopefully well thought out) site hierarchy, letting you lead them to the product you want. Some people however want to get right to it, they search. If they don’t find what they’re looking for, they’re gone.

As Avinash Kaushik points out, tracking WHAT people are searching for is super easy with Google Analytics but that doesn’t make your task – improving search results as easy as it could be.

You know how Google has a search quality team? For your site, that team is you.

Whether you track the inputs (search terms) and outcomes (number of results returned) using your own back-end systems or Google Analytics custom variables function isn’t as important as the act of paying attention to what they tell you.

Where to start?

Begin with the low hanging fruit and work your way up the tree – your biggest problems probably lie with the searches that return zero results.

No results found

This is the worst thing a site can return – it’s a door slammed in the user’s face. There are a number of actions you should take here:

Start with the simple, single search parameter queries. If you provide filtering functionality, ignore queries for ‘1TB Hard Drive’ with a price filter set to ‘below $10’, at least until that becomes a reasonable possibility.

1) Check your ‘no results found’ message.

If you’re going to slam a door in someone’s face, at least try and do it gently. Perhaps with some tips on how they might open it again ‘try widening your search, using fewer words’ (contextually of course!) or how you can help open it for them, with a contact form, with the subject filled in (‘I searched for “creme brulĂ©e torch” and didn’t find anything’), or even a phone number. Something that has the customer understanding that ‘your search is important to us’. Oh yeah, and it helps if you actually respond to those requests. I’ve seen ‘no results found’ pages that are completely blank. That’s the opposite of helpful.

2) Start with the largest volume.

Whatever reporting tool you’re using, sort by volume (by unique users – some people will search multiple times, the logic being ‘that can’t be right, they must have something for this search’ – incidentally this is the definition of madness. (GET A LINK) Examine each query – is it a misspelling? A synonym (hard disk vs hard drive)? Products you don’t sell (but do sell an alternative)? What does your site software allow you to do about this? Apply autocorrect filters? Add keywords to products? Make a search ‘fuzzy’? Show a message? Suggest alternatives? If the answer is none of those, time to hire a developer, find/buy a plugin or nag your ecommerce platform vendor. Work on the top ten once a week until the top ten is all low single digits – that’s when you’ve squeezed all the juice out of this.

Found something, exited on that page

These are the people that gave up in disgust – they didn’t find anything, they left. It could be they were just checking prices but whatever happened, they didn’t like what they saw enough to stick around any longer.

1) Filter out bots

Not an issue if you’re using Google Analytics as bots don’t load the tracking javascript but make sure you exclude them if you’re using something back-end.

2) Start at the top ten again and see what they saw

Do the same search yourself and see what you find. If could be that the search returned some results but not all, or the order of the search results meant that they didn’t see the product they were interested in on the first page. A search for ‘Photoshop’ might return an alpha-sorted list of products with photoshop in the name. If a company called Aardvark makes plugins and they show up first, the user doesn’t see the Adobe products in the first ten. Think about how you can provide more information to make it easier for the user to drill down and refine. Showing brand logos or a brand filter at the top is a simple solution and easier than writing your own ranking algorithm (Google have more PHDs working for them than you do).

3) Keep notes

There are some queries that you just can’t do anything about – make a note on them so you don’t duplicate your efforts – revisit occasionally.

Where next?

You could go crazy with this. Conversion rate is affected by a number of things: site design, purchase cycle duration, pricing, availability, visit intent (order tracking for an outstanding order isn’t much of a conversion opportunity). The best approach is to implement a form of user testing. There are free and paid for options. I’ve been using 4qsurvey’s free service. Whilst it has some aspects that aren’t quite optimal, you can’t argue with the price. And whilst you’ll only get a fraction of a percent of your traffic to fill in the survey, and some of the responses will be positive (pats on the back are lovely but don’t help you to improve), you will find the odd gem amongst the responses – the annoyances, frustrations and disappointments are what will drive you forward.

Image credit: D’arcy Norman via Creative Commons on Flickr.

Leave a Reply

Your email address will not be published. Required fields are marked *