How Machine Learning Can Help Hotels Via Segmentation Recommendations
Into 2025, this is going to be the ‘year of data’ wherein hotels now have the tools to develop real-time data flows and structured data warehouses to conduct proper AI-driven analyses. Lucrative and alluring in theory; the reality here still far grittier and built on the blood, sweat and tears of IT soldiers who connect the APIs one by one, navigate territorial data ownership policies and rigorously test the modeling for errors, false positives, false negatives and annoying edge cases.
There are many machine learning use cases and business models that will be developed in the coming years and months, but one that looks particularly shiny is in its ability to help with customer relationship management and loyalty building.
Besides solving the ‘unknown traveler’ problem of deducing what a person wants from their OTA email alias, ML can make serious inroads towards moving guests away from the 20th-century segmentations of leisure, corporate and group, and into more personalized, less-generic microsegments. With numerous data sets being merged due to technological advances, brands are starting to get quite granular with their KYC in terms of understanding where guests are coming from channels, device and geography, why they are selecting your property: leisure, corporate, conference, wellness and so on, and what their buying: packages, upselling offers, onsite ancillary spend or others.
Still, the situation on the ground requires grit. You have to get all the data in the right place – cleaned, structured, deduplicated and so on. Common friction points for working with APIs have been, first and foremost, that IT professionals need dedicated time to set up each interface, and then maintain all those established connections with each subsequent software update.
With each new system added to the tech stack, this quickly becomes resource intensive. Here, another type of AI called robotic processing automation has already proven itself by acting as a robot that can directly replace double entry work that has to be done manually because two systems haven’t been integrated to talk directly to each other.
Once you have all this data imported, cleaned to remove duplicates and structured into proper data fields, you now have an enormous treasure trove of numbers – ‘data is the new oil’ as they so often say. While this database is far too vast for a pair of hotelier eyes to pick out patterns, ML is designed precisely for that task. You give it the data; it finds the patterns, however, hidden they may be to the human overseer. The more data you give it, the more patterns it can potentially find and the more accurate its predictions will be.
Besides looking at vast amounts of data and then giving insights into that data, the key to ML is that it can produce a predictive model to optimize for desired future outcomes. Then, once that model is tested out in the field, the best AIs can then use the new data as feedback to improve their own modeling algorithm, further enhancing their predictive power to better optimize for a stated objective. That’s how ML works, then it’s just a matter of defining its goals around looking for patterns that can carve out microsegments and testing marketing purchases, rates, packages and upselling offers to optimize for conversions within that piece of the guest pie.
Where hotels have already seen the most lucrative applications for ML is in the revenue management system, with massive data sets comprising external and internal inputs are computed into an algorithm that can then recommend to the revenue director what pricing will optimize for rooms revenue, occupancy or now total revenue per guest stay. The RMS leaders are doing insanely great things behind the scenes insofar as analyzing rates; now it’s a matter of connecting in various sources of other business data into the CRM or the customer data platform.
It’s this whole notion of recommendations that brings us to the concept of having ML interpret not only how to adjust nightly rates or what response to provide for a website chatbot, but also to look at the multitude of guest profile data and then come back with its own set of microsegments for your revenue, sales and marketing teams to interpret and pivot their planning accordingly.
As of now, all of us are operating under a given set of established business assumptions based on how we were trained and our experience working in hotels. We see the world in terms of leisure, corporate and groups, and many of us have become locked into these guest segments. Recommendation engines based on ML don’t have those same limitations and thus can provide a fresh set of eyes on what your real segments are.
Perhaps this latest technology will help your hotel find an edge over the competition or allow you to deploy the advertising budget more effectively. Maybe it will give you suggestions on what packages will work better for attracting leisure guests or what types of groups are most winnable for your meetings and events business. Just as AI helps us to rethink business assumptions, neither of us would dare assume to know what such a tool would find buried within your hotel’s data.
Our advice is to first chart a path for connecting all your systems, and only then investigate these more advanced ML tools. ML needs a lot of training data to start and lots more feedback data as it starts testing various business models. At the same time, you will also have to confront the cultural, more existential scenario of what happens when the AI finds microsegments that contradict those that your teams start paying off. We live in exciting times!
