As hotels invest in AI for a plethora of tasks (e.g., pricing, forecasting, conversion optimization and guest experience personalization) many are discovering that the main barrier is not the algorithm but the mental model behind the data.
NB: This is an article from EHL
Subscribe to our weekly newsletter and stay up to date
AI has plenty of processing power but can fail when it is trained on flattened representations of guest behavior. In essence, hotels are applying advanced technology on top of outdated assumptions. Personas remove the variation, context, causality and real-time signals AI needs in order to be precise and adaptive.
Personas are Outdated
The first issue is that personas are built on averages, while AI works on individual-level variation. A persona is an aggregate profile, in other words a “typical guest” created by compressing many observations into a single narrative. That may simplify internal communication, but it comes at the cost of accuracy. Yet, AI creates value from differences that include slight variations between similar guests, changes with the same guest over time or reactions to small contextual shifts. Two guests assigned to the same persona may respond very differently to a rate increase, a cancellation rule, a loyalty benefit or an upgrade offer. More importantly, the same guest may react differently from one trip to the next even on business trips, for example. When hotels train AI on averaged assumptions, they blunt the precision the technology is meant to deliver. Averages make management conversations easier, but they often hide the signals AI needs to support better decisions.
Decisions Depend on Contextual Factors
This averaging problem leads to a second, distinct limitation. Personas also assume guest behavior is set in stone. However, even a well-built persona struggles in hospitality because guest decisions are often dynamic. Hotels serve guests before booking, during booking, during the stay and even after departure. A room night is shaped by travel purpose (business, leisure or bleisure labels hardly suffice), travel distance, trip duration, reimbursement rules, urgency, budget pressure, trip companion, cancellation risk, visible alternatives, and many other situational factors. These variables do not operate in isolation; they interact.
An Exhausted Guest is a Captive Customer
A simple example illustrates how guest behavior shifts depending on situational factors. I generally prefer dining in restaurants and rarely use room service. However, last year I attempted to order room service on three occasions. At first, this seemed inconsistent with my usual behavior, but the underlying pattern became clear upon reflection.
