Here is an example from a real and very popular car searching website, and at the point of specifying my preferred price range I can only pick one option. So I might click £20,000 to £22,000 because that is roughly my budget. But...

...budgets are rarely so straight forward. What if the perfect car was found but costs £22,100? This website would never show me that car because it is outside of my price band. The “computer” doesn’t understand that I might be happy to pay an additional £100 on top of £22,000 to get the car I want. I could of course pick the next band up, but I can only pick one so that won’t work. Or alternatively I could avoid specifying a price band at all, but then I get a list of rubbish cars for “spares and repair” listed from £100.

More often than not, (for me anyway), my budget in these cases falls on a boundary. in the above example I my budget might be £20,000. So do I pick £18,000 to £20,000, or £20,000 to £22,000? As a user this frustrates me, and frustrated users are far less likely to spend money with your company or recommend your website to friends.

Some websites have given this some consideration and have tried a little harder, but have still failed. Take this second real example from another very popular car searching website...

On this website I can specify any min price and any max price. A slight improvement, but still anything outside of that range, even by a single pound will be ignored. And notice the distance selector too; I can ask to be shown cars within 20 miles of my home, but I would then never see a car that is 21 miles away.

Adding the human touch

A technique we implement on various websites is the “fuzzy searching” algorithm which overcomes the typical exact matching as follows:

• Ask the user to specify their important criteria, but use the phrase “roughly” where applicable. For example, “Roughly what is your budget?”
  
  
• The user might type in £20,000 (or pick from a list of £1000 increments).
  
  
• Other criteria is selected for example Diesel fuel, leather seats, Distance up to roughly 30 miles away and so on.
  
  
• The search algorithm starts by looking for exact matches to all of the criteria and gives these 5 stars out of 5. Excellent matches.
  
  
• The search algorithm searches again with a slightly fuzzier selection, so the £20,000 might now become £18,000 to £22,000. Any matches are now given 4 stars. Pretty good results.
  
  
• If enough results are found we might stop now, but if only a few results are found we might go fuzzier still by searching without the diesel selection, or up to 40 miles.
  
  
• And so on. The search getting fuzzier until something is found.

This sounds complicated to implement, but the user experience is fantastic. Picture this real example...

A user is searching for:

     “A diesel car, £20,000, up to 30miles away.”

A typical website says:

     “Sorry we have nothing that matches.”

Our website says:

     “We have a car that matches but it’s 32 miles away. Another that is diesel at £18,999, and a car that fits all other criteria but it’s petrol not diesel.”

Conclusion

Building a search routine with a human touch, and with a level of fuzziness, and a star rating system, is of course more complicated than designing a more typical and straight forward yes/no search routine. But the difference in usability is outstanding. Visitors are happier, visitors are shown more products closely relating to their requirements, and more visitors make a purchase.

Why don’t all websites offer improved search facilities like this? Most developers are very logically minded which works well for most of their productive day, but fails terribly when trying to consider the way that the human mind works and how visitors might get the best out of a website.

Improved search facilities are also out of reach for a lot of developers. Anyone with a database can ask it to find records where “distance=20” as it’s as simple as literally writing “distance=20”. But not many developers would know how to perform complex, fuzzy, rated searches, and even if they did, making them run fast is another tricky task.

Final Thought

Google has proved itself to be the king of searching, and why? Consider again the website where you search for cars where you have 10 different entry fields for fuel type, number of doors, distance and so on. Google would never do this. Google has a single entry field where you can type in anything you are looking for.

Would it work if a website selling cars used this technique? Imagine a single entry field where a user might type “black diesel BMW, around £20,000, near London”. Does this work? I’ve never seen it done, but I think that it might!