The Anatomy of Search Suggest

If you are developing a nice and usable search engine on your website, you have most probably thought of implementing a ‘search suggest’ feature that gives useful suggestions to the user in a drop down menu. But have you considered that the search suggestions can be all different depending on what you want to achieve with the feature? Let’s have a look at the anatomy of Search Suggest – read before you start the development or improvement of your own suggest-implementation. The usability results can differ a lot depending on what model you choose.

(A Danish version of this blog post is available)

The idea behind Search Suggest

Google popularized the suggest functionality in 2008, when they introduced it as an experiment. The goal was to help the user formulate queries, reduce spelling errors and reduce keystrokes.

Google udbredte og populariserede suggest-funktionaliteten i 2008.

Google released their Search Suggest feature in 2008. Today it is a common thing on many websites.

Google’s suggest feature was later named Autocomplete. Other names for similar features on other web sites are “type-ahead” and “search assist”.

But Google’s version really only represents one of at least four types of Search Suggest:

  • Popular search queries
  • Mini search results
  • Index lookup (auto complete)
  • Search history

And as always when you start categorizing, we need a hybrid version that combines some of the above in different variations :-)

Type 1: Popular search queries

This type of Suggest shows what other queries users have entered and it puts the most popular ones at the top. It is probably the most common version due to the fact that it was the version Google originally released. When the user picks a suggestion, the query is executed and the user ends up with a set of (hopefully usable) search results. 

Search suggest of this type is especially good for:

  • Facilitating novel query reformulations
  • Encouraging exploratory search - the user is easily inspired to relevant searches
Amazon anvender Suggest der viser, hvad andre brugere har søgt på. Bemærk, at nogle af søgningerne er afgrænsede til et bestemt område af butikken - en snild måde at få flere brugere til at filtrere på.

Amazon uses a Suggest that shows what other users have searched for. Note that some of the search queries are limited to a certain area of the shop (“in Movies & TV”) – a clever method of getting more users to use a relevant filter.

Københavns Kommune anvender også en visuel simpel udgave af en Suggest der foreslår "gode søgninger". Nytteværdien er åbenbar - det er ikke nemt som bruger at huske på tekniske termer som "barselsrefusion" - men søgefunktionen hjælper til her.

The Muncipality of Copenhagen uses visually simple version of a Suggest feature that suggests “good queries”. The usefulness is obvious (at least if you understand Danish!) – it is not always easy to remember technical terms such as “maternity refusion” – but the suggest feature is a good helper here.

Zalando bruger suggest-funktionaliteten aktivt til at få brugerne til at søge mere specifikt og dermed bedre. Der er helt sikekrt flere brugere der blot ville skrive "cowboybukser" end "cowboybukser denim". Men sidstnævnte foreslås som det primære - og det er nok et godt gæt på hvad brugeren egentlig leder efter.

Zalando adjusts the suggest feature actively in order to make the users search more specific and precisely. In Danish, blue jeans are called “cowboy trousers” by most people, but more correctly, they are called “denim trousers”. Zalando puts “cowboy trousers denim” as the top suggestion in the dropdown menu even though most users have probably in fact just entered “cowboy trousers” in the search field.

Type 2: Mini search results

A completely different, but also popular version of Search Suggest could be called ‘mini search results’. Actual search results are shown in the drop down menu – updated search results are shown for every keystroke, hence the nick name “instant results”. When the user picks a result, he is directed straight to the content and not to a traditional set of search results.

This type of suggest is especially good for:

  • Promoting specific items or products
  • Support quick and efficient search

You should strongly consider adding a link called “See all search results” at the bottom of this kind of search suggest drop down. Also, add an indication of how many results the search query will result in. User studies carried out by Vertica shows that users can otherwise misunderstand the mini results as being the complete set of results. If the product is not among the limited amount of results in the drop down menu, many users conclude, that they are not at all in the web shop.

Dpreview har dybdegående anmeldelser af digitalkameraer og lader brugerne komme direkte til indholdet via suggest-funktionen. Der henvises dog også i visse søgninger til kategori-sider ("See all Nikon products")

Dpreview has in-depth reviews of digital cameras and lets the user go directly to the content via the suggest feature. The user is also given direct access to category pages (“See all Nikon products”)

Irma hjælper kunden med at komme hurtigere igennem supermarkedet. Man kan lægge i kurven direkte fra suggest-dropdown'en - så det er ikke engang nødvendigt at gå til produktsiden inden næste søgning foretages.

Danish supermarket Irma helps the customer to a quick shopping experience by letting him add items to the shopping bag directly from the suggest drop down. It is not even necessary to go to the product page for the product before the next item can be added.

Gyldendal præsenterer et stærkt kategoriopdelt mini-søgersultat i deres Suggest. Måske forsvinder overskueligheden en smule - men på den anden side åbenbares det for brugeren at Gyldendal har andet end bøger - der er fx også forfatterportrætter og kurser.

Danish publisher Gyldendal presents mini results that are strongly divided into categories (Titles, Authors, Articles, Courses). It may be less at-a-glance, but on the other hand it is revealed to the user that the publisher has other offerings than books. Also note how the presentation is tailored according to the content – user ratings and number of comments for the books are included as particular relevant pieces of information.

Type 3: Index look-up (auto complete)

Index look-up or ‘auto complete’ can be confused with popular search queries, but they are by nature all different. They are look-ups in the full product catalog or database of the content presented by the website. The content in this type of Suggest is so to speak more static because it represents a dump of the product database while not being influenced by the users’ interaction with the search field. Unlike the mini results, the list is not always sorted by relevance, but e.g. by alphabet.

Index look-up is especially good for:

  • Helping the user when accurate and efficient data entry is critical.
  • Letting the user select from a finite list of names or terms.
  • Exposing the complete catalog when other users’ search terms are irrelevant for the user. E.g. when searching for road names or zip codes.
  • Exposing good search terms before you have reached critical mass in the amount of search queries your users have made. It can be first step towards a suggest that will contain popular search queries.
På Boligsiden kommer registeropslaget rigtigt til sin ret - her er det stednavne og ikke f.eks. de malende salgstekster der søges i. Det giver en præcis fremfinding af det, som brugeren leder efter.

On the Danish site Boligsiden (real estate site) the index look-up comes in handy. Road and city names are suggested instead of the free text words from the (some times) graphic sales descriptions. It gives a precise search function for what the users are actually looking for.

På dsb.dk er registeropslaget supernyttigt. Det er væsentligt at brugeren får præsenteret de faktiske stationer - hvad andre brugere måtte have søgt på er f.eks. mindre vigtigt. Opslaget er i øvrigt relativt tolerant (man kan f.eks. stave Aarhus som Århus) - et registeropslag er ikke nødvendigvis  lig med rigiditet.

On dsb.dk (web site for the Danish Railways) index look-up on the station names a highly useful. It is essential that the user is presented to the actual names of the stations – what other users may have searched for is less important. The look-up is, by the way, relatively tolerant (you can enter Aarhus as well as Århus) – an index look-up is not necessarily rigid.

Smartguys suggest

Smartguy’s suggest feature lists terms directly from the product catalog. It leads the users to search terms that actually exists, but it is at the expense of more commonly used searched terms that users actually prefer. I have a suspicion that a primitive search engine is the real reason for the functionality here – a search suggest with popular search queries would probably be more useful here.

asdfadsf

Danish book seller Saxo’s suggest feature includes book titles and authors. The index look-up in the book titles is quite efficient at first glance, but it is also a bit buggy in real use – the lookup only works on the start of the titles. So if the word “motiverende” is found in the middle of the title, it will be completely absent here. You cannot expect your users to know what product names or other content start with in a case like this.

Type 4: Search history

It is fairly rare, but this type of suggestions can be useful if you want to support the returning visitors on your site. If you have users who log in, you can connect the history to the user profile and present the data across devices. When the user e.g. searches on his smartphone, he can see what he searched for on the laptop – and vice versa.

dba

On DBA.dk (Danish equivalent to Craigslist), the suggest drop down only shows the user’s search history. The number of results from the last search is presented next to each search query.

Type 5: The Hybrid

The four basic types of Search Suggest can, of course, be combined. And to a to a great extent, they are. With hybrid models, you can support multiple user needs at once, but beware: If the feature is not designed meticulously, it will easily be at the cost of usability. It is essential to identify what you really want to support the user in doing, and then focus on that task.

Apple

On Apple.com popular search queries and mini results are combined quite succesfully.

vvs

You do not need to be Apple to make a good Search Suggest. The version on Lavprisvvs.dk (Danish discount plumbing parts shop) is built after the exact same model as Apple’s. At the top, you have popular search queries, and below you find the mini results. Furthermore, a nice and clear link to more results (“Vis alle”) is visible at the very bottom of the drop down.

wh

Whiteaway combines a series of different suggest features in one drop down menu, and it still gives quite a good overview. However, the product photos are absent at the cost of a slightly less inspiring suggest drop down – at least compared to other examples.

homedepot

At homedepot.com we see an example of Search suggest on steroids! It has popular search queries, top results for each of these queries and several kinds of extra content is promoted. The intention is good but usability and efficiency of use has almost totally disappeared. Whoops! The site would undoubtedly serve the users better with a simpler suggest feature. The extra content can be better presented on the full search results page where you have more screen real estate.

Think of these usability guidelines when you have chosen the appropriate type of Search Suggest

I hope my categorization can help you in making some qualified decisions when you develop your own Search Suggest. However, before you release it, walk through this usability check list to see if you forgot some important details:

  • Performance is absolutely crucial in a well functioning search suggest – if it is too slow it will only be a source of irritation. A simple user test with a real user who uses the feature will quickly reveal whether it is useful. Expect to go for response times well below 1 second in order to get a good experience. The search engine should have fast response times but optimization of pictures is also a factor when creating a fast search suggest.
  • Return key should work as submit - some users will still prefer to get a full set of results and it is important that this part of the search function works as expected. When the users presses return, the search should be executed as if the suggest drop down was not there at all. Some users does not even look at the web page when the enter their queries – so they never see the suggest drop down. Consequently, it can cause great confusion if the search is not just executed when pressing the return key.
  • Support use of arrow keys - users should of course be able to use the arrow keys to navigate the drop down menu and select an item and perform the corresponding action with the enter key. If you have more than one action per line (e.g. “put in basket” and “go to product details”) you should define what is the most important action and assign that to the enter key stroke.
  • Test the suggest feature on a smartphone and a tablet - very often it is more irritating than useful on small screens. Many sites choose to completely remove it because of space and performance problems on mobile devices.

Have fun developing your Search suggest! If you have experienced a useful suggest feature, you are very welcome to share it in the comments below. I would love to hear from you. Or maybe you have experienced an example of usability horror in a Search Suggest that you want to warn your fellow blog readers about? Feel free to share!

2 thoughts on “The Anatomy of Search Suggest

  1. Pingback: Anatomien bag Search Suggest | VERTICA BLOG

  2. Of those, five had more thsn 6,000 parts peer million ”.

    Keyy Ingredient: Zea Mays: Cobtains aan anti-caking agent and is a healthier alternative
    to talcum powder. All manufacturing and packaging facilities undergo a rigorous inspection.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s