Constructor Blog | Ecommerce Search Industry and Product Information

Why You Should Care About Autocomplete

Written by Eli Finkelshteyn | Feb 4, 2015 8:00:00 AM

Many of us in the tech world live in a bubble. We assume things about our users that are not always true, and occasionally this comes out in embarrassing yet eye-opening ways. When my cofounder, Dan and I were working on Search at Shutterstock a few years ago, we realized we were very much doing this with regards to our customers’ spelling habits. Looking through our search logs, we were genuinely surprised to find that somewhere around 10-20% of all searches were obvious misspellings. These were customers coming to our website with a specific goal in mind, to find images of business cards for example, who instead searched for “busness,” or “bisness,” and got terrible results. The thing that really blew my mind was talking with some of these customers and finding out that they *never realized they had misspelled anything*. They thought we just didn’t have what they were looking for. That’s a lot of missed sales and unhappy customers.

We had the luxury of a large engineering team and time to research how best to solve the problem, and one of the best bangs we got for our buck was adding automated spell correction to our autocomplete, so when customers started to type “bisne,” the autocomplete would immediately recommend “business,” they’d click through, and all of a sudden have much better results and a much better user experience. It was a surprisingly difficult system to build right, and that’s coming from a guy with an MA in Computational Linguistics. How difficult it is to build an autocomplete well really shows when you play around with the autocompletes at other companies, even massive ones. Try misspelling “Kindle” as “cindle” on Amazon, for example, or run some slightly misspelled searches at Walmart, or Target and you’ll see how widespread a problem this is. Not only are all of these companies wasting months of developer time and money trying to all solve the exact same problem, most of them are not solving it well. And those are just the big companies. The little ones can’t spare the time or developers to build half-decent domain-specific autocompletes at all.

Realizing this, the idea for constructor.io was born. We use the extensive experience we have in this area to build out autocomplete as a service. Instead of every company needing to either lose huge amounts of their potential business due to misspellings, or huge amounts of time trying to build their own domain-specific autocomplete, they can pay us to make one much cheaper, and much better. We handle the entire service end to end. All our customers need to do is drop in two lines of JavaScript, and they’re good to go.

We invite you to try our service. A/B test it against what you have now and see how much business you’ve been missing. We’re so sure you’ll love it, we offer completely free 1 month trials with no strings attached, and we throw in an extra two weeks if you let us share the results of your a/b test. Get in touch with us, and we’ll have a real human walk you through the entire installation process and answer any questions you have. Give us a shot. You won’t regret it.