I’ve recently looked at site search box design, and best practices for results pages. Today, we look at how to use site search data. 

The terms that customers type into your site search box represent a wealth of valuable data that can be used to learn about your users’ behaviour. 

This data can be used to improve the site search functionality, to optimise results pages for common searches, improve merchandising and more.

I’ve been asking several e-commerce experts about how site search data can be used most effectively…

Why is site search data useful? 

According to Osric Powell from SLI Systems

E-commerce is increasingly dependent on data driven analysis for decision making. Site Search Analytics provides a unique and direct insight into the shopping behaviour of your visitor.

Analysis of the language used by your visitors is one of the most powerful and freely available sources of information you can glean from the tools you use to measure site search performance. Furthermore, the findings from Site Search reports is easier to communicate to internal departments and managers for maximum overall benefit to the business.

How you can make the most of site search data

Here are 19 useful tips, which provide ideas on what to look for from site search data, and the lessons that can be learned. 

Level of usage

Understanding how many people are using search and how that is performing over specific periods of time is useful to know. Also how may repeat visitors are using search is good to know and can be used for more time sensitive marketing.

Review the top search terms on your site

The most popular search terms can tell you which products being are looking for most, so you can then optimise your site search to make it easier for users. For example, you could use auto-complete to suggest popular products as people type. 

Also, if people are searching for products that are currently unavailable, this can help your buyers to make decisions about stock levels. 

Keyword value

Understanding the value of keywords used from the search box directly informs PPC and again gives your merchandising/email marketing teams some degree of focus when designing their campaigns.

Immediately increase your revenue

If you’re an e-commerce retailer, it’s great to have a process of working through your top 100 or so search phrases every so often, and checking what comes back when you actually search for those yourself.

One example, supplied by Dan Barker, is from Austin Reed’s site. If you search for ‘Barbour’ there, you’ll see that it returns zero results.

I’m pretty sure Austin Reed sells Barbour in some of their stores, and does sell dozens of wax jackets on the site. Therefore it’s likely to be a search term used on the site, but obviously nobody’s taken the time to take a look at that.

Even if Austin Reed doesn’t stock Barbour jackets, than providing results for wax jackets is surely better than ‘no results found’. 

Use this data for other marketing activities 

The way people search for products on your site is also an indication of the way they may search on Google. Therefore, site search data can be used to improve search advertising. 

There is a lot you can learn from the language people use when searching for products. 

Give the data to your buyers/writers/salespeople

It’s fairly easy in any analytics package to set up automated reports of top searches. These are often really useful for buyers, who need to know when demand is bumping up and down, but also need to be able to judge potential demand for products that aren’t already stocked.

According to Dan Barker:

For salespeople the data can be great. For example, if you’re selling job ads and you can tell potential advertisers things like “we had 30,000 searches for ‘digital marketing manager’ on the site last month, and only 4,000 for ‘online marketing manager’, so you should put the ad in using the word ‘digital'” that’s brilliant both from a ‘sales’ & a ‘helping the customer’ angle.

Using search for merchandising relevance

According to Oscric Powell:

Knowing the top search terms helps you to understand your most popular products, and this is obviously important. However, there is an added insight in that this helps you to understand how your visitors perceive you.

For example, one of our white goods retailers have found that visitors were entering search terms related to memory (ram, dimm, SD card etc) although they no longer stock such products. This could be an opportunity to start carrying such items or to partner with sites that do.

There is a deeper level of insight that can also be useful for merchandisers. Suppose you sell TVs and find that your most popular search terms is ‘30” LCD TV’. Which one of the tens or hundreds of TVs that you stock do you merchandise in your sort order above the fold on the first page?

Understanding your most popular terms and how that correlates with the most popular product (sku) is well worth tracking where you have many products that meet with the originally search term.

Look at ‘poor results’

What is the definition of a poor result? For some it is simply when someone enters a search terms and the site returns a ‘no results found’ page. This is important but doesn’t tell the whole story.

Another definition of a poor result is where the search term used returns hundreds (or even thousands) of results, as can happen on Amazon:

This can also have a detrimental effect on the site especially if you have poor filtered navigation options to assist the user with filtering the results.

Conversely, a poor user experience can happen if too few results are returned. A ‘poor results’ report should be able to provide this level of insight.

Timeliness of keywords

Knowing when people are searching (in terms of trends and patterns) can give your business a competitive advantage.

Analysis of the terms provides a powerful way to co-ordinate your PPC and adwords strategies ahead of time and can even provide savings by bidding for keywords before they rise in cost.

This is not the same as asking how many times visitors searched for a particular keyword during July. Rather we are asking what keywords are being used and how it trends out.  

For example, if you sell flowers, fruit seeds etc it can be useful to know that searches for ‘poppy seeds’ started appearing in March, peaked mid-July and began tailing off during August. 

Click depth

An indicator of search relevance and also how deeply visitors are traversing the site. Are users clicking through to pages five and six? Could we get them to that product quicker by cross linking results with suggested keywords and popular search terms?

Click rank

Where are users clicking on your search results? According to Osric, you tend to see the highest density of clicks on page one and in the top five to ten (above the fold) products.

If that is not the case then something is wrong with search results relevancy.

Pages from where search is conducted

Depesh Mandalia, Head of Conversion & Product at ticket.com:

It’s important to know which pages (aside from the homepage) customers are searching from since it may indicate a poor UX or that we’re not catering for customers as well as we should.

For example, some of our PPC landing pages are high on the ‘visited site search from’ measure which means we need to better serve our target keywords or create new target pages and split those keywords out which are driving site search hits (hence drilling back to KWs from those start pages is important).

Exit rate

This is important since it means users are exiting from site search most likely after not finding what they’re looking for.

Are visitors finding what they want? 

Looking at data on the percentage of search exits (people who abandon the site after viewing site search results) can tell you whether or not people are happy with what they find. 

It’s also a good idea to review search results for the most popular and most profitable lines, to ensure that the process works well. 

Trawl for other elements and info that are missing from your site

E-business consultant Dan Barker

On e-commerce sites I often see people searching for ‘size guide’, ‘store finder’, ‘delivery’, ‘return’ etc. Trawling through on-site search terms often gives you clues for that. It can also tell you when navigational options or links are missing.

For example, on holiday sites, if you find people are searching by ‘country’ & you only offer the ability to navigate by ‘city’, you can find clues as to where there’s enough demand to make the site change.

Looking at the page people were on when using terms like that often gives clues as to where information is missing too. So, if everyone’s searching for the word ‘delivery’ from product pages, it may be worth adding delivery info to your product pages.

Average order value (AOV)

Average order value and how that pertains to search visitors helps with the ROI argument for Site Search. AOV should consistently outperform the metrics you see for visitors who do not use the search box.

Look at searches that return no results

This may be an indication of people searching for products you don’t have, which may tell you what to stock in future, or that they are using different terms to describe your products. 

Device usage

What search terms, types and usage are we seeing across different channels, and are there differences in search behaviour between users of different devices. 

This information can be used to inform future design updates, or to optimise certain pages for different devices. 

Look at which filters people use

Knowing what facets users are selecting (and the order by which they are doing) is a very important part of assisting the user experience and how much content to surface. Too many retailers do not filter the amount of garbage they surface and thus confuse the visitor. 

For example, filtering by particular sizes may be more likely to lead to sales than filtering by others, and therefore you can spot inventory gaps. Or perhaps particular colours are more popular than others.

This can be interesting for travel sites too (where you get an idea of things like the lead-times people are searching for, whether star ratings/price are more popular for hotels in particular destinations, etc.).