Arriving at the perfect price at the right time for each listing is no easy feat when it comes to vacation rentals. And as the pioneers of dynamic pricing for short-term rentals, we understand this well.
To get accurate pricing for vacation rentals, advanced analytics and models are needed. These models should rely on data that is timely, clean, and validated at scale, because at the end of the day, the quality of data directly affects the model's performance – and low quality data will ultimately result in suboptimal prices, bad bookings, and frustrated owners.
At Beyond, one way we achieve the most accurate pricing with the most reliable data is through our practice of Clustering: Our team of global revenue managers analyze data from tens of millions of listings and groups them based on similarities - making tens of thousands of Clusters across the world. This allows us to create comparable sets of inventory for each listing that accurately determines the right price.
So, what does this mean for hosts & property managers looking to maximize revenue and bookings? Here I’ll delve into our stance and methodology on Clustering at Beyond, and how it plays a role in maximizing vacation rental revenue.
How Beyond Builds Clusters
Step 1: Owning the Data Pipeline
Data quality is essential to achieve superior pricing outcomes - which means that having direct oversight, access to, and ultimately ownership of data is a core differentiator for revenue management systems. If all pricing providers rely on the exact same data points, there will be minimal difference in the pricing strategy. As the key ingredient in a pricing strategy, first party data collection is imperative for the most accurate pricing, and your pricing models need to have complete control over the whole flow, from collection through to the final output.
Given how important data is for accurate dynamic pricing, here at Beyond, we invest heavily in both property management system (PMS) connections and online travel agency (OTA) data-gathering techniques – and have been doing so for over ten years. As a result, we have the largest data set of STR data available, and it is proprietary. This means that when anomalies occur or trends change, we don't rely on other companies to tell us what's happening. We instead rely on our global data and data science team to understand and handle any adjustments efficiently. For example, when Airbnb started to add longer listing_id’s that needed special handling or if a PMS changes how it separates out fees from rental rates, owning the data allows us to understand the impact, and how to adjust pricing models accordingly.
Owning our data pipeline is crucial to setting the right prices - and is ultimately the key ingredient in delivering higher Revenue per Available Night (RevPAN), ADRs and Occupancy rates to our global customer base.
Step 2: How We Establish Markets
Okay, so having your own data source is great, but what do you do with it? How does it go from trillions of data points to something we can use?
First, we go back to the basics. As I've already mentioned, data is used for advanced pricing schema - but it's also used in everyday conversations with our customers and amongst our team. Which is why we separate our data into conventional Markets, This is usually based on the general understanding of an area and how people talk about them, i.e., Chicago or Pais as an urban area, Dallas-Fort Worth with two names, but one market, or Istanbul with two continents, but one market.
At Beyond, we think in terms of metrics, events, day-of-week (DoW), and seasonality, so we require Markets to be similar for at least one of these. However, don’t worry, this is just the starting point – we don’t price all of Chicago the same.
One of the most common uses when talking about Markets is about supply, which has grown substantially over the past few years. While this increased supply means more competition, the silver lining of this means that there are more listings in which to create comp sets, and therefore, we can make these comp sets more accurate. As an example, below is the location of 1,000 of the +16k STR locations in the Smoky Mountains and even 1K is a lot to try to mentally map to a comp set.
Step 3: Gathering and Cleaning the Data
Now we get to the data that powers our core algo, primarily Airbnb, Vrbo, and hotel data from Expedia, we first focus on speed, or data freshness. Like most scraping operations, speed is important as 1) you typically can’t go back and scrape history (think days in the past) and 2) Markets change quickly, especially in the short term. In other words, old data can sometimes be worse than no data. Therefore, we constantly monitor our scraping machines, ensuring we scrape each calendar about every other day or markets that have shorter booking lead times, we scrape them even more often so we don’t miss out on optimizing pricing for that last-minute occupancy.
Once we obtain the data we store it in massive databases that allow for quick retrieval. This is important when dealing with large volumes of data, such as every price and day of Airbnb on every day ever. This allows us to run algos, but also have our analysis team run reports on the data at scale.
Another valuable point of having data at scale stored in a more dynamic database is that we can run our cleaning logic faster, which again helps with data freshness. In cleaning, we focus on removing the inactive - or ghost - listings that are not actually active, blocked days as opposed to booked days, as well as any listings that are not competitive in nature or getting bad reviews, or no reviews.
Step 4: Auto-Clustering
Now that the data is stored and cleaned, we break down the markets into the comparable inventory sets or Clusters. These Clusters are eventually where we aggregate the data in order to get the valuable metrics out (occupancy, price, lead time, etc.).
We do this by matching together similar, valid listings by bedroom size and grouping with a focus on geography. In the example below, you can see the polygons create a grouping of similar listings taking into account different neighborhoods and centers, like around a convention center or airport. We also take into account neighbors that may function differently, such as a central business district that is busier on the weekdays, and a bar area that is more popular on the weekends. We have never relied on phone codes or zip codes as they don’t align with the geography of how STRs function. Even better is that this is logic that can be run at any time as more or less homes get added to OTAs, we are always jostling boundaries, adding and removing Clusters to ensure accuracy.
Step 5: Clustering from the Human Perspective
Advanced logic is wonderful, but to get the perfect comparable inventory set, sometimes it's not enough. Our last step in creating Clusters is bringing in a Beyond trained Revenue Management Analyst to review the output and adjust according to their institutional knowledge and local market feedback. The same way a property manager wants to review their pricing, we want our experts to review our groupings. This is especially true for the most high earning STRs, they are usually the most unique and this needs to be accounted for. We spend time adjusting boundaries to make up with expectations, but also to make sure the geo-attributes that affect price, like ocean front or ski in/out, are taken into account.
Once these Clusters are designed and reviewed by our RM team they are all set for the real magic to start, the pricing algo!
What Does This Mean For Vacation Rental Hosts & Property Managers?
Beyond's accurate data set and meticulous approach to clustering play a pivotal role in maximizing revenue and bookings for vacation rental businesses. By owning the entire data pipeline and investing in our own data collection, Beyond ensures the highest quality and most up-to-date information, giving us a competitive edge in the dynamic pricing market.
The process of establishing markets and gathering clean data further enhances the accuracy of our comparable inventory sets, allowing for precise pricing tailored to each listing's unique characteristics. Auto-clustering based on multiple factors and the human perspective provided by trained revenue management analysts ensure that our algo accounts for even the most unique and high-earning vacation rentals. Ultimately, Beyond's proprietary data-driven approach and advanced algorithms provide property managers and hosts with the confidence and tools they need to make informed pricing decisions. All in favor of earning the most money!
As the pioneers of dynamic pricing for short-term rentals, Beyond continues to lead the way in revolutionizing the vacation rental industry and driving success for property managers worldwide. With Beyond's robust and real-time data set and innovative techniques, property managers and hosts can unlock the full potential of their vacation rental business, achieving the perfect price at the right time for each listing.
Ready to unlock more money from your vacation rentals? Sign up for Beyond!