‘Sometimes all you have to do is ask, and it can lead to all your dreams coming true.’ – Randy Pausch
The two most important factors which lead to a buy/no-buy decision on an e-commerce store are price and reviews. While price gets a lot of attention, it is strange that review collection has historically not received the same amount of love and care. The result? These are reviews from a certain familiar e-commerce store.
If this makes you cry, you are not alone. You have to trawl through so much plain text to find some meaningful insight (if you do find some insight, that is!)
Can we do better? If we look at how some newer companies are tackling this in other verticals, we find that the evolution has already taken place.
The biggest leap forward in the rating space was by Uber. Uber made ratings very easy (just a tap!), and they also added relevant topics to know more details.
Our hypothesis was that we can improve e-commerce product review experience and quality by using similar methods. We decided to ask users to rate the products they bought and allowed them to write the reviews as they generally would, and asked them a few additional questions which could be answered in just a click each.
However, we faced one hurdle. How do we pick the questions to ask? To solve this, for a set of products, we performed natural language analysis on some content available publicly and determined key topics that people care about. Additionally, to reduce the long list of topics we clustered semantically similar terms to ensure the topics don’t seem repetitive when shown to users.
Half the customers were shown a usual review (star rating and comment box) and this formed the control group. Let’s call it group A. The other half was shown the usual review and 3 additional questions relevant to the particular item that they had bought. Let’s call this group B.
The results were nothing short of stunning. The customers in group B showed significantly higher engagement and the response rate was an astounding 300% higher compared to group A. Asking about the relevant topics led to more involvement and more reviews. The information rich structured responses were a far more valuable source of data. What’s more, we rotated among the topics to get opinions on a much wider set of topics without overwhelming the customers with long surveys. The upshot of this is that the usefulness of the reviews dataset goes up as we collect more and more reviews. Isn’t that how it should’ve been to begin with?
Understanding existing reviews can help you collect better reviews with more engagement.
This is an unstoppable trend. Some of the world’s leading food delivery companies are now changing the way they collect reviews too.
Are you still collecting reviews the old fashioned way? Get in touch!