Background
A travel agency specializes in high-end corporate retreats in exotic locations. They offer a turn-key service, organizing everything: hotel rooms, charters, luxury transportation, visas.
Most of the corporate clients schedule special activities during those retreats, such as skydiving or horseback riding. The agency typically also offers optional activities to the travelers, for a fee: wine tasting, hot air balloon trips, arts and crafts workshops, etc. Those optional activities are a big source of revenue for the agency and allow them to reinforce business relationships with local partners, facilitating future opportunities.
Over the last couple years, however, a new trend that threatens the agency's profits has appeared: travelers book many optional activities, then cancel at the last minute. This causes logistics nightmares for the agency and often takes the profit margin out of some activities.
The gilded cage
The agency evaluates many scenarios to solve the problem:
- Require a non-refundable deposit
- Require pre-payment in full
- Make cancellations tedious by forcing travelers to call during business hours rather than do it online
However, all those scenarios are coercive by nature, which isn't compatible with the agency's brand. The business thrives because they offer a VIP service to their corporate clients; any policy or practice that threatens that image of 5-star service is bound to damage the bottom line.
Therefore, the management team is adamant that any new policy must add value for customers, not just for the agency. This is not an easy problem to solve, and hours of brainstorming with senior staff only lead to weak suggestions, such as improving the user experience by adding more photos/videos in the activity selection website.
Big data to the rescue
The agency's management team are strong believers in data, and have baked this obsession into every system the agency operates. Emails, CRM events, hotel bookings, travel schedules, even seating arrangements in charters: everything is carefully logged in a state-of-the-art data lake hosted in the cloud. This magnificent asset has served the agency well over the years, and they figure that the answer to their latest problem lies somewhere in some hidden pattern in the data. Since the agency is too small to have data mining experts on staff, they reach out to us whenever they need data-driven answers.
There are usually two kinds of problems involving data mining: either one is looking for an answer to a specific question, or one is just trying to identify some kind of patterns. In our case, we're looking for patterns, so we employ cluster analysis techniques.
We immediately notice that there's a greater similarity between activities that are cancelled than between travelers who cancel activities. This tells us that the agency has a better chance to solve their problem by looking at the way optional activities are packaged and offered than by trying to coerce or convince travelers to remain committed to their selections.
The next thing we do is look at the customer journey during the selection of optional activities. In most cases, the selection is done online using a website that follows a typical e-commerce workflow (nice catalog with pictures and videos, shopping cart, etc). We also assume that given the profile of the typical traveler in those corporate retreats (executives and senior management), administrative assistants handle the actual task of selecting activities, possibly based on minimal verbal instructions from the traveler. The timestamps on the interactions with the website tend to confirm this theory: almost every registration process is done during business hours.
This analysis tells us two things:
- An activity is either selected or not; there are no ways to detect interest unless the activity is selected
- The emotional impact of the pictures and descriptions is lost on most travelers, since someone else handles the registration process
Data-driven decisions
We share those findings with the management team and propose the following:
- There should be some kind of "maybe" option in the selection process; travelers could use it to indicate interest for an activity without making a commitment
- Any activity that gets a lot of "maybe" should be repackaged/rebundled asap
- More energy should be spent on adapting the offering based on live data than on improving the shopping experience, since the travelers themselves don't interact much with the website
The agency immediately starts working on implementing those suggestions, and this leads to a tremendous success. Not only does the profit margin on optional activities increase, but the quality of the offers improve consistently as the new "maybe" option provides the agency with valuable feedback, in real time.
