Search is not just the place your users go to when they are lost or have something specific to find. It is also a place where your users go to to have a conversation about their needs. Avinash Kaushik remind us that search is the only place that gets used more than 10% of the time. So, how are we designing this conversation? In this presenation for the IXDA Singapore community, I share a framework to design the search experience.
26. Content modelling
Content modelling is a process of
exploring and defining how content
types can be chunked and linked to
add value to customers and
businesses.
http://www.pebbleroad.com/perspectives/4-view-approach-
to-content-modelling
40. PARAMETERS
Location Identity Activity Time
ACTION
Push Pull
Hidden actions
Towards a Better Understanding of Context and Context-Awareness
https://smartech.gatech.edu/bitstream/handle/1853/3389/99-22.pdf
41. What will a context-aware app do
when I search for recipes?
42. What will a context-aware app do
when I search for recipes?
43. Identity: who usually writes a retrospect?
Time: Can I know when a retrospect is due?
Activity: Can I know when a retrospect is
being written?
62. The real work
starts after
you’ve
implemented
search.
Tuning
63. “The impact of search on business
performance depends more on the
level of investment in a skilled team of
people to support search than it does
on the level of investment in search
technology.”
Martin White, author of Enterprise
Search
64. https://www.youtube.com/watch?v=m4RaGKwvMbY
“There’s absolutely no feature on your
website, none, that gets used as much
as internal search…
Your header, left nav, footer, main body,
promos on the right…None of them will
get used 10% of the time…
And site search is used probably much
more…”
Avinash Kaushik, author of Web
Analytics 2.0
68. How many users are using the search box?
What are the top xx places users begin their searches?
What are the top xx search queries?
What are the top xx pages reached through search?
What are the queries with 0 results?
What are the queries with 0 click-throughs on the SERP?
How many times was search used immediately after the first SERP?
What are the queries that result in more than 3 SERPs?
How many users used facets or filters?
How many users use “best bets”?
…
This is how many people know search – the box and the 10 links.
This is how our clients know search.
Search is a like a black box to many.
I would like to open this black box of search and take you on a journey of understanding search so that as designers we can create better search experiences for our users.
Search is a people-machine interface. People pose the questions and the machine answers. The expectation is that the answers satisfy us.
The answers may satisfy us when the questions are pretty basic such as, “How many pounds in a kg?”/].
When when the questions are fuzzy, which is most of the time, the answer is fuzzy too. We are unhappy when this happens.
It turns out that search in fuzzy situations is an iterative process. After each answer, we may evaluate the results and ask a different question till we can get the answers that satisfy us.
What we find changes what we seek. Search is a conversation.
http://pages.gseis.ucla.edu/faculty/bates/berrypicking.html
Marcia Bates showed how the search behaviour or the infoseeking behaviour is like a berry picking session. We don't to to the berry forest with a plan. Rather, we pick a place to start and then based on the quality of the berries we get we make our way to the forest.
It turns out that when we search there are many types of answers we’re looking for – from simple info retrieval to complicated info seeking. These are called search modes.
Google is great for info retrieval.
Google sucks at the first iteration of a fuzzy query.
Amazon is better for learn and explore modes around a single collection.
This is an interesting example for a fuzzy question on which locations in Vegas will give good rental returns. Instead of returning a list of 10 pages this website shows a map, color-coded according to rental yields. Which one would you prefer, this map or a list of 10 pages?
When we conduct search workshops, we start with a job story or a user scenario. The scenario revels the search intent. When many people create their scenarios we can get a good understanding of the type of answers they are expecting from search.
This matrix matters. It turns out that double-savvy searchers search differently than double-novice searchers. So having a good understanding of who is doing the searches is important. We can design differently if we think that a majority of the users are tech-novices.
<div>Icon made by <a href="http://www.freepik.com" title="Freepik">Freepik</a> from <a href="http://www.flaticon.com" title="Flaticon">www.flaticon.com</a> is licensed under <a href="http://creativecommons.org/licenses/by/3.0/" title="Creative Commons BY 3.0">CC BY 3.0</a></div>
This is not the Burj Khalifa. It is a picture of the Burj Khalifa. This is a resource.
We then describe resources. This description is called metadata.
When you have many resources, you get a collection.
When you want to search a collection, you need metadata. No metadata no search.
How to get metadata? The hard way is to do it manually when we are creating the content. This is the preferred way when you control the content such as in an app or a website. There are automated ways of creating metadata but the results have not been that encouraging.
Here is an example of a content model. Modelling is a creative act. What you include could point a way to innovative solutions and new business models, as we shall see shortly.
Google extracts metadata automatically. That is why they can show the metadata under Search tools. Google has the luxury of having access to millions of documents from which to gather metadata. They also have 1000s of people creating complex algorithms to extract metadata.
During our search workshops we use the scenario cards to extract content types. We can have an exercise to model the content types.
<div>Icon made by <a href="http://www.freepik.com" title="Freepik">Freepik</a> from <a href="http://www.flaticon.com" title="Flaticon">www.flaticon.com</a> is licensed under <a href="http://creativecommons.org/licenses/by/3.0/" title="Creative Commons BY 3.0">CC BY 3.0</a></div>
Usually happy hour ends as evening progresses, but not for this bar. This bar is in a stadium. Foot traffic reduces in the evenings. This bar is reacting to its context.
What would you get if you Googled for “Olympics”.
Today, you get these results.
http://searchengineland.com/olympics-medal-count-info-128855
But if you did the same in July 2012, you would get the medals tally. It totally makes sense. Google is reacting to the context its users find themselves in.
What is the most relevant answers for these searches if you think about the Singapore context?
http://picjumbo.com/download/?d=HNCK2298.jpg
Yes, mobile is all about context.
http://searchengineland.com/google-now-reminds-pay-bills-205379
This morning I searched for Art Friend. There are two branches, one closer to home, from where I made the search and one away from home in the city. But when Google shows me the results it shows the cone that is closer to home, even though the city branch is the flagship branch.
What is the most relevant answers for these mobile searches if you think about the Singapore context?
It turns out that there are many ways to view context. Here are a few.
Based on context, you can either push info to users or contextualise info when it pulled by users. You can also do nasty hidden things like what Uber does, but automatically using the info without telling users.
Remember the receipe content model we built?
When I search for recipes I get only the ones I can make from home. How does my search app know that? Well, it checked that ingredients I have in my refrigerator. It also knows that I like pasta and that few ingredients were missing so it tells me that there is a supermarket on the way home where the ingredients are available. How does it know that? Well, it checked with the supermarket!
Is this all possible now? Technically yes.
Continuing with our workshop, we extract context from the scenario cards we get plus we conduct interviews to better understand how a content type is used.
This is the stuff that designers know. They know the interface elements that go into building the search interface. But like what we’ve seen, don't forget the journey to the interface. Users, content and context will change your interface design.
The big book of search interface patterns.
If search is a conversation then the search snippet is the place where the conversation takes place. So, yes, we design more than 30 snippets for 30 content types if we include device context.
If we want to delight users when they search, then having a thesaurus is a must. It helps the user to better understand the resources that we have. A thesaurus is difficult to create but once it is done it is relatively easy to maintain. However, your search engine should be tuned to use the thesaurus when making ranking and display decisions.
UF = used for
BT = broader term
RT = related term
NT = narrower term
Yes, to offer a good search engine, you must do these well. But how many of us can create a search engine from scratch?
We have two powerful, open source search engines.
If you need a quick start, try these search engines in the cloud.
A good read if you want to know the enterprise context and what it takes to set up a search team.
Search results follow a long tail type of curve. So we can start at the short head. Pick the top queries and try to understand the performance of the search from these queries.
You can get beautiful answers to the performance of your site or app by asking these simple questions and looking at the search logs for answers. Remember, search logs capture the user’s intent.