Why We Changed the Popularity Ranking Algorithm
Over the past decade, the volume of reviews and opinions on TripAdvisor has increased exponentially – from six million in 2006 to over 350 million in 2016. Today, travellers are adding 200 new contributions to the site every minute.
Due to the speed at which travellers are sharing their experiences on TripAdvisor, we found instances where a newly listed property was able to skyrocket to the top of the rankings on relatively few 5-bubble reviews. We called these properties “fast-risers”. Over time, TripAdvisor travellers would submit more reviews on the fast-risers, causing them to settle into more stable, accurate rankings. Unfortunately, while this transition was happening, TripAdvisor consumers may not have been seeing the most accurate rankings for the destination. Fast-risers would temporarily enjoy higher positions before falling significantly in the rankings, and other businesses would appear lower than they would have otherwise.
To improve our site experience for travellers and businesses alike, we’ve enhanced our Popularity Ranking algorithm. The enhanced ranking results in a more accurate representation of a business’s performance over time. These changes were not undertaken lightly and were carefully designed and tested to improve our rankings algorithm in some very specific ways, while maintaining the accurate standings of existing properties with great reputations among TripAdvisor members.
In designing the enhancements, we analysed the hundreds of millions of reviews we’ve received over the past 15 years and how they impacted the rankings of properties over time. Based on this research, we designed the enhanced Popularity Ranking algorithm to value the quantity and consistency of reviews more significantly than we have in the past. Doing so helps stabilise the ranking for all businesses, reduces fast-riser behaviour and creates a more accurate overall ranking for our travellers. We’ve tested the enhanced algorithm extensively – both internally and by analysing how travellers interacted with the site as we’ve gradually rolled it out.
When did this happen?
In January 2016, we switched to using the enhanced Popularity Ranking algorithm for Restaurants and Attractions on TripAdvisor. Between February and April, we rolled out the algorithm for Hotels. During the roll-out, there were some cases where the rankings would move independent of new reviews being collected. Those ranking changes were due to the enhanced algorithm accounting for more and more of the calculation for each property over time.
The enhanced algorithm is now being used on the vast majority of the site, and the resulting Popularity Rankings have stabilised.
How the Enhanced Popularity Ranking Algorithm Works
Our goal was to design an enhanced algorithm where a property settles into a stable ranking more quickly and avoids the fast-riser behaviour. The Popularity Ranking continues to be based on the quality, recency and quantity of reviews that a business receives from travellers. The final algorithmic input, for a small set of properties that violate our guidelines, are fraud penalties that exist to protect the integrity of the content on TripAdvisor.