Users visit TripAdvisor.com to research vacation options, book hotel reservations, and post reviews and ratings of their experiences at hotels. One way TripAdvisor makes money is through an annual subscription service in which hotels pay for the right to display links to their own websites and other data within their profile pages on TripAdvisor.com (see image.)
Whether a hotel renews its subscription is largely determined by the number of click-throughs (CTs) and related bookings it receives through TripAdvisor.com, according to company Director of Analytics Michael Barry. This is a metric TripAdvisor can materially impact by increasing or decreasing the level of marketing related to specific clients.
While the company can’t control what users write in their reviews or how positively or negatively they rate hotels, TripAdvisor can control, to an extent, its clients’ CT rates. Naturally, TripAdvisor wants to retain its most valuable clients, so the company’s analytics team built models and uses advanced Big Data analytics to:
- Predict how many more CTs are required for a given client to renew;
- Calculate how much increased marketing activity will be required to reach the required number of CTs for that client to renew;
- Determine if it makes economic sense to do so.
For some clients, Barry and his team has found, even a significant increase in CT rates only marginally increases its likelihood of renewing. For others, even a modest increase in CTs dramatically impacts the odds of renewing. Knowing this through advanced analytics, TripAdvisor can focus its finite marketing efforts on the latter and not the former.
Underpinning TripAdvisor’s analytic efforts is a custom Big Data platform. It uses Hadoop to store and process web log data; Hive to query and aggregate the data into tables; SQL Server for reporting against aggregate data; and JMP, visual discover software from SAS Institute, to further explore the aggregate data.
Action Item: The TripAdvisor use case above highlights several important considerations every Big Data analytics practice should consider. First, identify which performance metrics you can control and those you can’t. Next, marketing and other resources are finite, so use analytics to prioritize customers for targeting based on relevant value metrics. Finally, don’t get locked in to one specific technology or vendor, but rather find the right mix of tools to get the job done.