Whenever you start typing on Google, you may have seen that after a couple of letters, the dropdown starts showing exactly the thing that you are looking for. Amazing right. It’s like Google knows you so well. Well actually it does. A search engine captures a lot of information when a search happens. And based on that it comes up with accurate predictions in the future.
On July 14, 2019 was the Cricket World Cup final. And if you had typed “En” on that day, most likely your first suggestion would have been “England vs New Zealand scorecard”.
Today, June 02, 2021, if you go on Amazon and type “Au”, most likely your first suggestion would be “Automatic Sanitizer Dispenser”. And I am certain, this wouldn’t have been the first suggestion on June 02, 2019.
Google, Amazon and any search engine for that matter has a dynamic Autosuggest algorithm. One that works on retrieval and ranking approaches. The algorithms are obviously different based on context, complexity and business logic. But the underlying concepts remain the same.
Retrieval
The retrieval mechanism is simple. Given an input, it finds a possible set of outputs. Here the input is a letter, a couple of letters, or a partial string. And the output is a set of search queries matching that input.
Ranking
Once the retrieval part is done, there are thousands of matches. But which ones to show and in which order? Here the ranking mechanism is applied. This may take into consideration my personal information. I drive a car. I may be looking for a car wash. Or maybe the context information is considered. My previous search was “car servicing”. So now I may be looking for a car wash. Or maybe geographic and demographic details are considered. When most people in my age group, my gender search “ca” in my region, my city, my country they are looking for a car wash. Or may be based on current affairs. Because of the recent cyclone, most people who start with "ca" end up searching "car wash".
I have just highlighted 4 parameters. Complex ranking algorithms take into account many more parameters, assign them weights. And it's so dynamic that "car wash" may be my top suggestion at 9a.m. in the morning but by afternoon it may have changed to "car bomb near Parliament" based on a political incident.
Now that you have a basic understanding of how Autosuggest works, let me tell you how this helps your users.
According to Nacho Analytics, having a good Autosuggest in Search boosts conversions and sales upto 24%. If you are not paying enough attention already, you should start today.
That’s all folks. Until Next Time.