Machine Learning from PMS Data Extraction Enables Personalization and Segmentation | By Adam Mogelonsky

In HN's latest THEMATICS World Panel, we posed the question of how primary data extractions from the property management system (PMS) can be used for machine learning (ML) to the grow a hotel or heighten productivity (see World Panel Viewpoint). With a range of astute viewpoints from respondents, there are two main themes that the answers revealed.

The first major use case is personalization. This may seem obvious as it’s been at the core of all AI thought leadership and company AI explorations for quite some time, but it nevertheless deserves some unpacking. What AI enables is the democratization of guest personalization, whereby having a solid backbone of smart automation technologies can allow hotels to deliver towards a specific guest or specific guest context.

Here are some examples of where ML can be applied to PMS data when combined with guest data from other key sources like the CRM or POS:

• Analysis of a guest’s booking history, billing details and preferences to predict service requirements or what will ‘surprise and delight’ during the next reservation with the brand, in addition to helping with error recovery situations or with any service that would be aided by being able to respond in real time
• Beyond algorithmic rate optimization recommendations, which many RMS vendors already excel at, ML can help define upselling strategies and purchase propensity algorithms
• Bridging the gap between sentiment analysis and rate strategy to develop models for enhanced dynamic pricing and agile personalized marketing, by, for example, rapidly identifying high-value customers and using one-to-one offers to entice trial, knowing that the lifetime value of said customers will far outweigh any loss leader incentives
• Similar to customized marketing, with the right models loyalty programs can be specialized to offer more attractive and individualized perks, rather than adhering to the classic model of tiered loyalty member classifications

Related to all guest personalization applications, the next evolution of this is the development of enhanced or AI-guided segmentation and microsegmentation. This isn’t to say that the traditional mainstays of leisure, corporate and group no longer apply, but they are too broad to work as optimally as possible in today’s hyper-fast, neuromarketing-driven world. Data connections and structuring allows ML to test assumptions and make recommendations about what actually motivates various guests to buy or spend on ancillaries.

Here are some ways that ML can help with customer segmentation:

• Create audiences or microsegments to verify or modify human biases about a hotel’s ideal customers so that companies can more effectively find high-value guests or develop more productive marketing offers
• Comp set benchmarking to reveal which properties are direct competitors for each given segment as well as highlight areas for improvement
• Using topic or sentiment analysis, as derived from sources like social media listening tools or the CRM, to find relationships with PMS data in order to guide rates, packages and marketing

As you can see from just these three, AI-enhanced segmentation has many similarities to what can be done on the personalization front. No doubt there are plenty more use cases that will be made apparent as hotels deepen their exposure to ML, genAI and other tools. Common throughout, and especially for PMS data extraction, is having strong integrations or interfaces, or using a customer data platform (CDP) or robotic processing automation (RPA) to connect all the siloed databases together.

Visit this World Panel discussion

To close, if there was a third area to touch upon for where primary data stores like the PMS can have applications for hotels in the present day, this would be team efficiencies. For instance, in accounting, AI can now help with error detection in financial audits or invoice processing. For housekeeping, predictive room status changes and more accurate occupancy expectations can help to define the cleaning priority list or more flexible start times, while merging data from major citywide events can help with long-term staff scheduling. Thirdly, a large language model (LLM) may soon be used to interpret SOP manuals or operations guides in order to facilitate faster customer support, microtraining and the ability for team members to ask questions when focused on a specific task. Lots of applications, so best reach out to your tech vendors to learn more!

Contributing World Panel experts included:

• Frédéric Brayer, Vice President Accor Tech Sofitel, MGallery and Emblems
• Luis Weir, Senior Director, Integrations and API Strategy, Oracle Hospitality
• Tomeu Fiolv, Global Hotel Technologies Director en Meliá Hotels International
• Dave Berkus, Managing Partner at Wayfare Ventures LLC
• Fergus Boyd, Hospitality Consultant
• Darko Vukovic, CEO and Founder - PolyAPI
• Adam Mogelonsky, Partner at Hotel Mogel Consulting Ltd.
• Juanjo Rodriguez, Founder, The Hotels Network
• Mark Fancourt, Co-Founder at TRAVHOTECH
• Vincent Somsen, Hospitality and Travel Tech Advisor
• Simone Puorto, Founder | CEO | Futurist
• Uli Pillau Founder and CEO of apaleo