Do You Know What Users are Searching for on Your Web Site?

Do you know what people are searching for on your site? Got anything interesting to say about it? If so, Lou Rosenfield and Rich Wiggins are conducting a survey – What are users searching for on your site? – for an upcoming book on search analytics for web sites and intranets.

It’s funny that this should come up now as I’ve been meaning to jot down a few thoughts from analyzing search behavior on the Children’s web site which I manage. So here goes.

We actually use a great hosted search solution. I went down this route for a number of reasons, the main two being that it is cheap ($249 per year) and that it is very easy to set up and customize.

Although this is a topic for another day, I am a big fan of hosted web services like this – it’s so much easier than having to install and maintain software. Especially if you have to get this done by your IT department.

And if you change you mind or if it doesn’t work out – you just (okay, it’s a little more than ‘just’) cancel your subscription.

Some Search Stats

Anyway, back to search. Here are some recent stats for our site:
In October 2005 we had about 17,000 searches on our site. Of these, 72% produced results and 28% produced no results.

What’s interesting here is that although our web site traffic and therefore our total number of searches has been increasing every month, the split between searches that produced results and those that did not has remained remarkably consistent at around 72/28.

That doesn’t hide the fact that a quarter of people who search for something on our site don’t find anything!

Does this mean that our site doesn’t have the content that it should? Well, fortunately no. When you start looking at the phrases people are searching for, you soon see why they are not coming up with any results.

Dear user, why are you searching for Google and Yahoo! on my web site? What is that crazy misspelled phrase you are typing in there? Did you honestly expect me to understand that?

There is a really long tail when it comes to searches that produce no results. Of those 5,000 or so searches, a good 90% are one-offs making it pretty much impossible to do anything about them.

This is not to say that this information isn’t valuable. It was only by looking through our “no results” data that I noticed that people were regularly searching for “internships”, but that we didn’t have an internships web page. Needless to say, that omission was quickly rectified.

I also use this data to add additional keyword meta data to pages. Often people will type in a layman’s version or a misspelling of a medical term.

If I see it occur more than a few times in a month, I’ll add that misspelling to the appropriate web page’s meta data.

Want Help? No Thank You

So, we had 17,000 searches and about 5,000 of them produced no results. You’d think that people might check out the help link on our search results page to see if they could improve their success rate.

Think again. The help page was viewed a paltry 35 times (and that was high compared to previous months).

The lesson here is that if you’re writing help documentation for your web site search (or for any other application on your site, for that matter), stop right now and spend that time doing something that will actually add value for your visitors. No one reads help. That’s all there is to say.

More Results? One Page is Plenty Thanks

Each search on our site returns an average of 87 documents. This has also remained remarkably consistent over time. Naturally, I’d love for this number to be lower in order to increase the likelihood that the results being returned are the right ones.

However, given how poorly people use search (43% of searches were one word only) and how consistent it’s been, this seems like a pretty hard number to move.

The default setting for our site search is 10 results per page, so that’s…9 pages of results for the average search. Ack! Especially as only 5% of searchers go beyond the first page of results. The other 8 pages might as well not be there.

What to do about this? Not sure other than to work on making sure that the right pages are on that crucial first page of results. I call it the “golden page”.

Fortunately, our search results are generally pretty good. In user testing, we typically find that if someone searches for some fairly typical information, the page they need is returned on the first page of results.

Of course, whether thay actually see that and click on the result in question is another matter – they don’t always, no matter how hard I silently will them to as I’m watching over their shoulder.

Speaking of using search poorly, only 1% of searchers use phrase searches – that is, put their search terms in quotes in order to reduce the number of results. Scary. It’s a wonder people find anything!

Searching Within Results

I was surprised, however, by the number of people who go on to search within their initial search results. Of 17,000 total searches, 64% were first searches, 20% were second searches and 16% were third or more searches.

It would be nice to presume that these searchers were trying to reduce the number of results by entering a more focused search phrase the second or third time.

However, from watching user behavior, they could just as likely be hitting the search button again (and again) in the vain hope of bringing up something different. Honestly, I see it happen.

Lessons Learned

So, what’s to be taken away from all this (arguably somewhat depressing) data? Well, to paraphrase Jared Spool,

Search is the option of last resort.

People use search badly. Don’t rely on it as a safety net if your visitors can’t find what they need by navigating your site.

On the other hand, on larger sites search does get used a lot and some people even like to search as soon as they arrive on a web site.

So, monitor your search logs on a regular basis – monthly is about as often as I can manage. Get to know what people are searching for and are finding and not finding. Which pages get returned the most often – should they be?

Tweak your meta data and use the tools your search engine provides to customize your results and thus ensure that your most important pages are ranked first.

For example, on our home page we have a “Careers” tab, a big button to search job listings and a link in the footer, and still, “jobs” and related terms are our most popular search phrase. So, I made darned sure the jobs page gets returned first for those searches.

Got any tips from your own experience? Feel free to share them.