Issue 3 – Search results are not relevant – to anyone
- Datahantering och kvalitet
The ultimate guide – how to solve 4 most common issues with site search
The site search seems to be an area in digital development that is hard to crack for the most companies. In this blog post series Tuija Riekkinen and Hans Ahlborg capture the most common issues and remedies to ensure a better performance of the most neglected sales booster of your website.
Like the collaborative writing process of this blog post series proves it, the key for tackling challenges with the site search, requires collaboration and dialogue between different digital competences – a multi-discipline team.
Issue 1 – Site search is not promoted – Design conventions and other reasons
Issue 2 – Missing content & data structures prevent engines to do their job
Issue 3 – Search results are not relevant – to anyone
Issue 4 – Too many results and no possibilities to refine search results
What does “relevance” mean?
Relevancy is a difficult topic, since it is not an easy thing to define.
When a business representative says that the site search returns irrelevant results, there is always some witty IT-person who kicks off the discussion about the semantics of the word “relevance”. The IT-person might ask that whose relevance are we talking about – yours or mine? Although the intention for challenging the statement is good, it often kills the discussion on the actual topic – site search not returning useful results.
Site search is an area that lives in the grey zone between IT and business. It is often viewed as some kind of “as-is set up”; it will work for itself. Plug and play.
But this is not true.
Search engines need quite a lot of love and attention in order to return good results.
Most of today’s search engines like Solr, Elastic Search, Epifind, Oracle, Watson and Google GSA are independent of platform and very easy to adapt for any business need. However, these search engines will need the assistance of trained developers, information architects, search editors and behavioral specialist to perform in an optimal way.
How to solve this
If site search indeed returns bad results, there is no magic the IT can do by themselves. In order to improve the search results, some detective work is needed. The good news is that the search engines are 100% logical, so it is possible find the reasons for bad results.
Like a detective, you need a case to work with to really get an understanding of what the true issues are.
In order to find a case to solve, you need a good sample of bad searches, list of most popular search phrases and web analytics. Try drilling down to the details to find out cases where consumer journey was interrupted after using site search.
A search engine collects everything it receives and discards nothing unless you clean it out, so look for search trends and all zero-returned searches (or one search phrase searches with gazillion hits) to find out what is wrong with the set-up.
The detective work cannot be done solely by business or IT. You need both.
You need a group of detectives who all have their own area of expertise. Different competences will share in their knowledge and best practices to achieve the one goal – better site search.
A page is a collection of content fragments
One important member of the group of detectives is a person who knows the content structure of the pages.
This person can answer to questions such as how the page is being composed, where does the content come from, who is the editor of the different pieces of contents and what content is unique to the page and what content is shared across many pages.
Sharing content across the site is common for e-commerce sites, for example product pages are hardly ever 100% unique content wise.
By understanding the structure of the pages, you’ll start seeing the different fragments of content. All these fragments are being indexed and weighted by the search engine.
The unique content fragment should get more weight than the ones shared across many pages. There might also be content fragments which are completely irrelevant and should not be indexed at all.
Just the fact of tuning and putting focus on what should not be indexed is quite important. In one project, we worked on, the search was returning 3217 instances of “e-mail” into the search result. This was caused by the fact that each page of the site had a content responsible with an e-mail link and the word e-mail. The machine had not been told that it should ignore this footer in each page and subsequently showed this in the search result. This is where for an example the text in script “no robot.txt” should be used.
After you’ve analysed the content fragments, the detective who understands how the search engine is being configured, can start improving the way the content fragments are indexed and weighted.
Tackle quality of the content but do it wisely
There are typically always also quality issues with the content and those need to be fixed as well.
It is recommended, though, to conduct the analysis of the fragments before drilling down to the quality of the content; when you have identified the most valuable content fragments, you can put all your efforts to improve the quality of the content in those fragments rather than trying to fix everything.
When improving the quality of the content, the search engine log will help you finding what people are searching for but cannot find. Remember to use that information in your content strategy as well.
Your vocabulary and user’s vocabulary
The second very large part in solving the relevancy issues is to ensure that the information architecture (see part 2 of this blog series) is sound and solid.
The key part is taxonomy which manages all the baseline of your e.g. product information but also “folksonomy” is very vital. Folksonomy is the words the user uses when looking for your products. And yes, they can be very different from your “official vocabulary”.
The search log will help to mitigate also this conflict by showing the difference between how the company defines the language compared to how the user express themselves while searching.
It is also important to pay attention if the users are searching with natural language “full sentences” or are they using simple words? This will help on how you tune the capabilities of your search engine and the eventual investment in development you must do help the user find what they are looking for.
- If the Site search return bad results, don’t start the blame game. Form a task force from different digital competences and start solving the puzzle together piece by piece
- If your pages are composed from content fragments, find out first the most useful content. Then optimise it both by ensuring that search engine gives additional weight to it when indexing and by improving the copy quality
- If you struggle to find a way to start solving the puzzle, start by doing a comb-through for the search engine logs