Using smart cues to simplify hotel booking decision for our guests

Booking a hotel is a fairly involved decision. You want to ensure that you are making an informed decision. However, unlike most e-commerce purchases, the product on offer here (a hotel) isn’t as standardised (yet), and thus there are a lot of factors to consider.

Some of the commonly asked questions before booking a hotel include -

  • Is the quality of rooms good?
  • Does it offer value for money?
  • Is it at a good location?

These are questions where the answer isn’t exactly a straightforward yes or no, but instead lies on a spectrum and needs a careful consideration of all the factors involved. That’s why every online hotel booking platform shares copious amounts of information about a hotel — reviews and ratings, pictures, amenities, qualitative descriptions etc.

However, sometimes, due to various reasons, it becomes very difficult to parse all of the information to arrive at a decision. In such scenarios, instead of going with the best option, we sometimes settle for the one that appears to be the simplest. One such example is booking a hotel at the last minute.

“When faced with a difficult question, we often answer an easier one instead, usually without noticing the substitution.” — Daniel Kahneman

The case of last minute bookings

Treebo fun fact: We get a sizeable number of same day bookings till as late as 8 pm.

This kind of behaviour implies people are on the move and book a hotel right about when they actually need it. To translate this into numbers — the journey from discovery to purchase is completed within a few minutes, making it difficult to process all the information provided.

Difficult, that is until now.

Introducing smart cues!

To aid guests when booking a hotel, we decided to use smart, data-driven cues to simplify the process involved. Our goal was clear — help guests make well-informed decision minus the preceding research. So, we designed our solution with the following two principles in mind:

  • Display only one message per property, in order to keep things simple.
  • Messages to be based on the most relevant data streams — availability, price and overall interest in the hotel (or popularity).

Based on the shortlisted data streams, we created the following helpful messaging cues:

Availability

Message Text — Hurry! 1 room left

While this conveys a strong opinion about the hotel in terms of popularity, it also acts as a good indicator of how much in-demand a property is. In other words, a hotel with only a few rooms left, most likely indicates that people looking for hotels in an area think this one is the best!

Price

Message Text — This hotel is Rs *** cheaper than usual.

All of us like to shop around to ensure we are getting maximum bang for the buck.This message aids that thought process by showing guests that a hotel is available at a cheaper than usual rate, thereby saving extra few bucks and offering them value for money!

Popularity

Message Text — Top viewed hotel today!

This is a leading indicator of demand. While availability represents if a hotel has already been booked to capacity, this message shows that there is more than usual interest in a particular hotel (for a given set of dates).

The technology behind smart cues

The aim here is to merge three data sources (availability, price and popularity) and have them consumed by one service. A hosts of technologies are utilised for this purpose.

User actions over various touch points across all our platforms are captured via Segment, pushed through an API gateway, streamed to Kinesis Stream and persisted to the date-partitioned S3 via a Lambda (basic enriching) and Firehose. This serves as reference point for computing average viewership of a particular hotel. For the real-time traffic, Spark Streaming is used and the aggregate numbers are pushed to a MySQL RDS.

What the future holds

This is just the first version of smart messaging cues that we have built. With three separate data streams being consumed by one unified service, we will keep churning more fine tuned, granular and intelligent messages. All in a bid to simplify the decision making process of booking a hotel for our guests. Therein lies the power of this technique, and therein lies its importance!

P.s. Watch this space for success metrics and more on our experiment with smart cues!

Thank You!

The rockstars behind this product are Raam Baranidharan, Shanker Sneh, Pragya Jha, Aditya Shetty & myself.
A big thank you to Aditya, Sneh & Avnish for contributing to the post.
Follow us for more such posts from Treebo and staying on top of how product and technology is shaping the hospitality industry!