Since our last major product update in April 2014, we’ve added over 450 cities across 11 countries, and have always actively asked for feedback from our foodies across the globe to work towards creating the best restaurant search experience possible. Of the thousands who have obliged, one recurring point that’s emerged has been about too many restaurants being clustered within a small rating range. Until now, restaurant ratings were being determined by a weighted average of absolute scores, which resulted in this clustering within a range. But what foodies really want is to be able to compare two restaurants to decide which one is a ‘better’ destination. The only way to solve this is by moving from an absolute rating to a more classroom-style grading model, where the distribution of scores in every city on Zomato is normalized, resulting in ratings in each city being distributed over a normal curve.

As illustrated in the example below, the earlier system (the orange line) put over 50% of the restaurants in the range of 3.2 to 3.5, which doesn’t really help one decide which restaurant is ‘better’.

Once the absolute rating curve has been flattened to create a comparative distribution curve (the red line), fewer than 30% of the restaurants will fall within the 3.2 – 3.5 range. We’ve also cleared out a lot of spam based on improvements in our spam-control filters, so there’s less noise getting in the way when you’re trying to decide where to eat.

There are plenty more features and improvements on the way and, as always, we’d love to know if you have any suggestions for things you feel we could/should be doing differently.