Artificial intelligence (AI) is at the forefront of hospitality technology heading into 2024, and in looking at specific tasks where it can be deployed, a very lucrative application is in attribute-based shopping (ABS). Alongside other pursuits like dynamic pricing and adept channel management, the ability for hotel guests to select individual rooms, configurations, services, add-ons and ancillaries in an a la carte manner has long been sought after as a way to bolster net revenues (some postulate by as much as 10% for the average property) without any significant upleveling of the physical product, and now with some AI bolted on this newly unlocked value may finally be achieved.

Before diving into why AI works for ABS and why your brand may need it in 2024, two prominent obstacles need to be addressed, both from the transient guest’s perspective as well as from the hotel manager’s side of things. While many have thus far blamed legacy systems as the hindrance, this is hardly the case as leading industry vendors have all deeply considered ABS modules as a way to provide more value for their clients. Instead, the challenges are mainly psychological.

Stating the Objections
From the guest’s point of view, it’s a matter of ‘shopper’s paralysis’. You give a guest four room types and they have no problem selecting which one fits their needs and budget. You give a guest those same four room types along with varied check-in times, F&B packages, high floor or low floor, close or far from the elevator, ocean view or no ocean view and others like these, and it becomes too much. Instead of making a choice, the user abandons the cart and goes elsewhere for their accommodation needs.

How about solving this with a purchasing sequence within the internet booking engine (IBE) so that these variables act akin to a prix fixe menu? This can work, but then you must also consider the stepchild of shopper’s paralysis – decision fatigue. Especially when it comes to the psychological pain of putting money on the line, making a decision is the most energetically taxing function that our brain performs. Ergo, the more monetary decisions you laden a guest with, the more displeasing the booking process becomes and, again, the more likely they are to abandon the cart. In this case, the customer may already be exhausted having compared upwards of 20 to 30 different lodging options, and throwing ABS on top of that would be overkill.

Onto the back office and the situation is one that every hotelier around the world can relate to – room blocks and VIPs. Hoteliers need to maintain control over where they place guests right up to the last few weeks before arrival so that they can maintain flexibility for accommodating a big group contract that came in or someone that deserves a higher-tier room based on their loyalty status. For these cases, ABS would ‘lock’ specific rooms and possibly create more work for managers who then have to untangle groups or VIP requests around these upsold ABS room assignments.

How AI Helps ABS
The mechanics of A/B testing different options and offers as they relate to machine learning (ML) presents a profitable avenue for both incremental revenues and insights to guide future capital improvements or a full-property renovation.

At its core, ML is just pattern recognition based off of a multitude of data – namely, thousands upon thousands of guest interactions with a specific brand as well as those derived systemwide from all hotels using a vendor’s platform – then optimizing to better fit that pattern towards a given outcome. In a general sense, the more data a machine has, the more patterns it can decipher and the better it can fit its behavior towards that stated goal.

So, if the objective is to maximize the number of bookings, the AI may determine that the best route is to not deploy ABS within the IBE. Why exactly? Answering that isn’t within the machine’s purview, but the human overseer may judge from the insights found by A/B testing different purchasing sequences that it’s because of shopper’s paralysis interfering with conversion maximization. Contrarily, if the objective is to optimize for revenue per room reservation, then the AI may find that compromising slightly on occupancy with well-sequenced ABS of the rooms inventory alongside other saleable products delivers a better overall topline.

There are indeed some prominent vendors who have built incredible online shopping experiences that incorporate elements of ABS and ML directly into the IBE. Getting into specifics for all the possibilities and advantages for specific hotels would take pages, but rest assured, they’ve found a way to thread the needle between customer decision fatigue and net revenue maximization. When considering how a prix fixe menu relates to the entirety of the guest journey, the two other areas where A/B testing can occur are within a prearrival upselling platform as well as within the hotel app for when the guest is onsite. Again, vendors are on the case, especially for the former of the two where ML can also observe the optimal day out from arrival for when guests are most likely to spend on ancillaries or individual rooms.

Separate from all the transient merchandising considerations, another prominent area where AI is reaching the forefront is in robotic processing automation (RPA) wherein a machine can be trained to connect disparate systems together when a strong interface doesn’t exist or there are problems structuring the data. While not expressly a function that current vendors can solve for right now, with ML’s core purpose being pattern recognition, RPA should be considered as a means of observing the behavior behind all group bookings so that a hotel has a better sense of what that segment’s volume will be according to pace reports and when the best time is to serve up ABS offers.

Why ABS Needs to Start in 2024
We’ve talked a bit in abstract about what the machine can do with the user interaction data it is provided, but we haven’t yet emphasized how the machine learns. It’s relatively easy to understand what A/B testing is or even how a Monte Carlo experiment can work to deliver a range of possible revenue outcomes with a high confidence interval. What’s more difficult is, like the obstacles, the human factor of being patient with the learning process.

That is, it takes time for the machine to test within a given range of variables in order to decipher what actually works to get guests to purchase more at any given moment as well as deliver actionable insights that can be used by hotel teams to improve product messaging, contextual delivery, sequencing, packaging, new types of onsite programming or capital expenses that maximize returns.

Because all this hinges on big data sets, every digital interaction that your guests currently have with your hotel where all the booking variables aren’t being recorded then tested is an opportunity that’s lost. This will likely require a lot of deliberate thought about what IBE, CRS and upselling platform you are using to start putting ML behind your online shopping experience, but it’s a task that should be undertaken inside of 2024 because otherwise lots of money is being left on the table.