How to mess up Search in five easy steps?

You know implementing and maintaining search is challenging. It’s easy to mess up.

Chances are you’ve already heard several horror stories.

Or probably, if you work with search, you’ve already made some of those mistakes, too. (Shhhh, I’m not going to tell your manager it was you! 😉 )

In the last 15+ years, I have seen so many search mistakes I cannot even count them. The main reason, in my opinion, is that search has a very special position in every organization’s life:

• Search is considered to be a “black box”, users don’t understand how it works. At the same time, their expectations are super high.
• Your managers think search should be “easy”.
• You know it is one of the biggest challenges of your life but you want to prove you can deal with it.

So you start doing things. You listen to the users’ complaints as well as the stakeholders’ requests, and you go for it. You read the internet and feel like a superhero who is ready to change the world…

Until you run the first PowerShell script to update the search configuration.

Or you change something in the service application settings and the result is a very ugly “Something went wrong” message instead of the intended result set.

You name it.

You’ve been there…

 

Five Steps to mess up Search

Let me collect the five most common mistakes one can make while configuring SharePoint search.

#1 – Add everything, literally EVERYTHING to search, without understanding your users’ real needs.

Yes, in enterprise search “Everything” search is still a thing, but this doesn’t mean it really has to contain everything. For example, the first thing I always remove is AllItems.aspx and other list forms.

The next consideration is not to include archive files in “Everything”. It’s ok to display them as a separate search vertical but in 99% of the cases, users find the outdated and archive content irrelevant, and even disturbing.

You name it.

Talk to your users, listen to them and remove everything that is irrelevant. After that, “everything” will be relevant and useful for them.

 

#2 – Add content to search without understanding what it is.

Do you know what kind of content is in a source you’re just about to crawl? How many items? What kind of items? What types of documents? What metadata? Who is everything relevant for? How are the users intended to find the content? What do they want to do with the content?…

Ask questions, and then ask more questions.

If you don’t know how much content you’re about to crawl – how you can be sure you’re not breaking the whole search environment by a full disk or a content processing timeout?

If you don’t know what the users exactly need and how they want to use the content, how you want to present the result for them? (Do you really think the out-of-the-box “ten blue lines” experience is great? That it cannot be any better?…)

 

#3 – Don’t pay attention to relevance.

Relevance tuning is one of the darkest black boxes of search.

Or at least, this is how it is considered. However, with some guidance, there are some small tricks that can make a big difference. For example, you don’t have to change the ranking model, you can also impact ranking by Query Rules.

Do you think tuning relevance is not important? – Ask your users how many times they could not find a document they were looking for on the first three pages of search. How many times they gave up, and asked a colleague to send over the intended document by email instead…

 

#4 – One Search to rule them all.

Search Center is a good thing if you want to have “everything” search (see #1 above).

But did you know search can be behind so many applications? Some examples:

• customer portal;
• call center application;
• template browser;
• “MY” dashboard;
• project dashboards;
• team dashboards;
• etc.

The common thing? – Each of these aggregate information from various sources (sites, site collections, in many cases even from different source systems!). By the traditional integration model it would be challenging to implement these applications, but by using the search engine as the “integrated database” behind everything the implementation can be much easier.

But these applications don’t even have to be huge, complex ones. Even a simple “my tasks” web part added to your team’s landing page can provide huge benefits to your users.

Search is much more than “enterprise search”.

 

 

#5 – You don’t need a test environment. Let’s do everything in production, immediately.

Search is not the place where you can easily show yourself smarter than you are. “Fake it until you make it” is not an approach that works here. It becomes visible very easy, very soon. So don’t try to fake it.

Learn it instead, and then test, before you go to production.

The challenging thing with search is that in most organizations, production environment cannot be replicated as a test environment, with all the content. Which makes sense, but make sure, all the relevant content is represented on your test environment to be able to do functional tests there. (Performance testing is another story, though).

Do not fake it. Test it until you make it in production, instead.

 

My goal with this post was not to scare you. You’re still here, reading this post, so I suppose you’re somewhat interested in search.

In this case, I have good news. Implementing and managing search the right way is possible. Moreover, it is learnable. Let me help you!