The hotelier’s ultimate guide to occupancy forecasting

In this guide, we’ll revisit the basic fundamentals of revenue management and walk through a step-by-step process to effectively forecast hotel occupancy using real datasets.

Forecasting future occupancy levels allows hoteliers to:

1. Optimize pricing strategies.

2. Improve resource allocation

3. Build realistic financial/budget goals

This guide will serve as a comprehensive resource, providing valuable insights and practical tools to guide your forecasting efforts. And, we’ll use a real-life revenue management situation to learn how to forecast occupancy effectively.

The challenge posed to us:

Management wants an occupancy forecast for the month of October. The good news is: that it’s currently mid-summer, so we still have a few months of lead-time! The bad news: The forecasting method must be a manually-created forecast, meaning: no RMS forecasts, algorithms, or Artificial Intelligence when creating our forecasted occupancy rate for the month.

Sounds scary?

Don’t worry, it shouldn’t be if we follow some logical steps to arrive at a data-driven occupancy forecast.

So, let’s get started.

1. Determine your forecasting time frame

It’s important to define exactly which time frame you’re forecasting occupancy for to cut down on vagueness and confusion. The process of forecasting for a single night, a month, or an entire year will vary quite a bit, so it’s important to define exactly what the date range in question is.

An effective occupancy forecast is clearly defined; for this example, our goal is straightforward enough: to create an occupancy forecast for October. This is much more quantifiable and effective than forecasting “for the next 30 days or so”, or “the next couple of months”.

By keeping our occupancy forecast quantifiable and well-defined, we can use this to compare our forecast to other, previous October occupancy forecasts which can help us when doing other processes such as forecasting.

2. Collect historical data to inform your forecast

Historical performance data

One of the most basic ways in which we can forecast occupancy at our hotel is to first gather data about your hotel’s historical performance for the same (or similar) period you’re forecasting for.

Using a combination of a Business Intelligence tool, a benchmarking tool, and a rate shopping tool, we can quickly assess how our hotel has performed for the month of October for at least the last few years. Data from your CRS or PMS will also suffice.

Analyzing historical performance is a great place to start because it’s really quite simple! By analyzing previous years, we can start to piece together some very basic guidance that will indicate roughly what our occupancy might look like for the coming year.

Digging into our historical data, we can see that the past three Octobers have experienced occupancy performance of (85%, 93%, and 93%) for the 2021, 2022, and 2023 time series.

Knowing this, it likely wouldn’t make sense to forecast 96% occupancy unless there was a major factor that justified this deviation from the norm.

Here’s a look at how easy it is to pull historical performance data from BI, here we are pulling last year’s occupancy and rooms sold production

Here’s an easy visual of room nights sold for the past 3 years

For our example hotel we can see that October is historically a strong month for us. Averaging our 2021-2023 performance we get 1934 rooms sold (or 90.42% occupancy) for the month. That said, this number isn’t a valid forecast though, as historical results aren’t indicative of future performance! It offers some very vague guidance, though.

Remember: most of the time, hoteliers don’t expect to sell every available room every single time (although we certainly would if we could).

Even in periods of excessively high hotel demand, we still have to contend with cancellations, out-of-service rooms, and other factors that can jeopardize the sellout.

For this reason, we will want to calculate the total number of rooms available for sale over the forecast period (2139 rooms) and consider this as an upper constraint, as we know this realistically won’t happen. Our forecast should not exceed this number.

Seasonal trends

In addition, it’s beneficial to collect data on seasonal trends - so, not only how past Octobers have performed, but specifically how October as a month stacks up vs. the surrounding months of the year, and why it performs this way.

This ‘seasonality’ factor is very important as it is much more multifaceted than it may initially appear, and includes many other factors that we will discuss in more detail below (such as events).

Let’s think of a few examples where seasonality greatly impacts a hotel’s forecast:

• A hotel in a warm climate (such as Orlando, Miami, or San Diego) might see an uptick in occupancy due to a robust corporate convention calendar in winter months, a time of year when other more northern cities aren’t a desirable destination.

• Hotels in the Northeastern United States often see a flurry of high demand in the fall when the picturesque foliage begins to change color. As soon as the display is over though, hotels often experience their slowest, most low-demand period of the year.

• A ski resort in the Alps will often see its high and low seasons directly dictated by the level of snowfall that it receives through the winter.

So let’s turn our attention back to our example hotel to learn more about its seasonality.

For our hotel take a look at the average October performance vs. the other months of the year going back to 2021.

October is historically one of the 2 strongest months of the year and is sandwiched between two of the other strongest months of the year. This offers us a broader context around where we should expect October to perform not only outright, but also relative to the other months.

Therefore, it would be illogical to build an October forecast off of last year’s February forecast but may be useful to use surrounding months such as September and November for additional guidance.

Conducting this type of analysis can be easy if you utilize a business intelligence tool with flexible reporting - simply export historical data for each month for your desired date range (consider pulling at least a few years back). If you don’t currently utilize a BI solution, this data can likely still be sourced via your PMS and organized in a spreadsheet for quick calculation.

3. Account for unique factors that impact occupancy

Now that we have a well-defined timeframe we’d like to forecast for, and then some additional context about historical performance and seasonality, it can often be helpful to assess any unique factors that may influence any future findings we will encounter as we begin to analyze our data.

Many other factors can influence an occupancy forecast - while we won’t be able to list them all, here are a few common factors:

• Renovations - perhaps our hotel is renovating and has many hotel rooms out of service. Perhaps a competitor is renovating which is boosting demand at our hotel.

• Fluctuations in market supply - perhaps many new hotels are opening in your area which is diluting occupancy at your hotel.

• Natural disasters - For example, if a competitor hotel experiences flooding while your hotel remains unaffected.

• Macro-Economic-level market trends - political unrest, recession

If you find that any unique factors are affecting your hotel, be sure to note them and refer back to them if you find any unexpected

For the purposes of our forecast this year, thankfully we didn’t have to grapple with many of these factors.

The only one that made an impact on my forecast decision making was reviewing the lower-than-average October 2021 occupancy performance.

After reviewing this, it was unsurprisingly caused by a slow first two weeks due to post-covid event cancellations which continued to plague this market well into 2021 and even into the first half of 2022.

4. Conduct data analysis

Now we’ve arrived at a crucial junction, it’s time to pull back the curtain and do some detailed data analysis. On some additional factors.

Pickup and Pace

Pickup and pace are two crucial KPIs that are always top of mind for revenue managers. These are two metrics that we can use to continuously hone our occupancy forecast over time. Valuable questions we can ask include: Are we already pacing ahead of where we thought we would be at this point in the booking curve? Are we seeing pickup in an unexpected segment that will require us to revise our forecast up?

Thankfully, tools like Lighthouse’s Business Intelligence allow us to look at pickup and pace side by side in a convenient view. Examining our pickup and pace charts in Revenue Insight, we can see that our pace is up by a strong margin (+138 room nights).

The dashed line in the chart below shows where we might expect to land for occupancy if we experience last year’s level of pickup, which In this case would be 2099 rooms sold, or 98.13% occupancy.

Much like the rolling 3-year average example above, this also isn’t a valid, standalone forecast by itself. We can’t expect this year to play out in exactly the same way as last year, but this number is another interesting possible outcome, and good food for thought, but does not include well thought out, rigorous analysis.

And what about pickup?

In the last 30 days, it has been exceptionally strong when compared with last year, but much of that pickup is specifically due to a corporate group (more on segmentation later) that was recently added for the week of 10/28.

You can see a screenshot of the group as it appears in BI in the screenshot below. As we’ll discuss later, this group has pros and cons, but is a big reason why our pickup and pace metric looks as promising as it does.

Demand

Additionally, understanding the anticipated level of future demand makes occupancy forecasting much easier. Strong, varied demand drivers create numerous opportunities for high occupancy from different types of travelers.

A good starting point for demand forecasting is to review an event calendar that shows various events along with their expected attendance.

Below is a comparison of our 2023 vs. 2024 event calendars for the month of October. As it stands right now, the event calendar looks quite a bit softer this year, especially for weekends (and especially for Saturdays in the first half of the month).

In addition to the simple count of events, we can also get anticipated attendee data for many of the events on the calendar.

Looking out across the events for this coming October, our anticipated attendees attributable to known events appear to be about 20% lower than last year’s anticipated attendee counts.

While the situation may improve as we get closer to fall, there does appear to be a slight decline in the number of conventions and conferences for 2024 and their anticipated attendee counts.

That said, we shouldn’t immediately assume that occupancy will be lower - as we are just beginning to scratch the surface of our analysis. For now though, let’s anticipate that we may have some minor headwinds this year, especially on weekends.

Stay Pattern

Guests stay on different days of the week, and for varying lengths of time. Knowing their average length of stay and which days they tend to stay on can lead to some crucial insights that will lead to a more accurate occupancy forecast.

At our hotel we can see that we have a relatively balanced stay pattern with occupancy ramping up through the week, ultimately leading to our strongest nights for occupancy: Friday and Saturday.

Remember when we discussed the demand component, and our lighter event calendar this year? Looking at pace by day of week we can see that our weekdays are driving all of our positive pace, and as might be expected, weekends are slipping behind.

This aligns with our original assessment that there were potential weekend headwinds for occupancy.

Market Mix

Market mix is a crucial component to create the most accurate forecast possible. It’s one thing to say:

“We anticipate 1000 room nights this month”, but it’s an entirely different level of detail to say: “we can anticipate 250 room nights of government segment business, 500 room nights of retail business, and 250 room nights of tour-group segmented business.”

Which would you rather hear in your next revenue meeting?

There are infinite ways to incorporate market mix analysis into revenue management analysis, but a good starting point for using market segmentation to create an occupancy forecast would be to analyze current booked production.

Recalling our pace chart from earlier, We know that we have a solid variance but can we explain this by market segment? As it turns out we can! A deeper analysis reveals that we are experiencing an uptick in Transient - Qualified Discount and Transient Negotiated Business. Also Group SMERF and Group Corporate are adding nice positive variances. Our largest negative variances are in the Transient Discount, and Group Social segments.

Another aspect of market segmentation we can analyze is cancellations for a certain segment. Knowing that much of our positive pace is owing to the group segment, we will want to pay careful attention to that segment’s tendency to cancel to ensure an accurate occupancy forecast. Looking at the cancellation to booking mix for 2023 and 2022, it looks like roughly a 10% cancellation rate is very reasonable to expect.

5. Forecast future occupancy rates

Now let’s put pen to paper and actually generate our forecast. To do this, many revenue managers will use spreadsheets, but for our purposes we will use the forecast + budget module within Lighthouse Business Intelligence.

To start, I created a user forecast in Business Intelligence using Last Year’s performance as a template. This will be the template that we modified based on our data analysis and findings above.

Based on the softer event calendar and lower attendee counts for the current year, I reduced our weekend to-be by several room nights on both Fridays and Saturdays.

Some reasons for the change:

• We are already pacing behind STLY for weekends, with a softer event calendar.

• We weren’t consistently selling out last year on Fridays and Saturdays, and all indicators are showing that it will likely be even more to sell out this year.

Using our findings from the Market Mix and Stay-pattern analysis above, we know that the week of 10/27 will likely present some challenges due to the large group allocation on the books for 10/28 and 10/29.

• From a stay-pattern perspective, this will reduce stay-through availability and negatively impact the Sunday. For this reason, I reduced our to-be’s by a significant amount to account for this disruption in the stay pattern likely to occur because of the group.

• Next, I accounted for group wash and reduced to-be’s by 10% accordingly. This is in line with the cancellation rate that the hotel has seen over the past few years for this segment. While SMERF groups are up in pace, we still must account for group wash

Weekdays (except for the Sunday mentioned above) look strong; to-be’s were slightly increased for several midweek “peaks” to better reflect how the hotel will likely perform based on current occupancy trends. We can confidently forecast this way due to our findings in the pickup and pace section above, where we found positive midweek trends in pickup and pace in certain corporate segments.

So what was our forecast after all?

1,998 room nights sold (93.4%) occupancy +- 1% to account for variability.

All signs are pointing toward a reasonably strong October with performance very similar to the last two years but with a slight shift in segmentation favoring negotiated and group corporate.

Weekends will likely be slightly softer due to a less active event calendar and underwhelming recent transient pickup, but will likely still be relatively strong. Weekdays will likely see a slight lift to rooms sold due to stronger group and qualified discount production.

Here’s how our occupancy forecast stacks up with both LY’s totals, and the algorithmically generated BI forecast. Notice that we are quite close to BI’s forecast, which is built on a sophisticated forecasting model:

Improve occupancy forecasts with a predictive market intelligence solution

Forecasting your hotel’s occupancy can be a complicated and involved process for even the most sophisticated hoteliers.

Compiling historical data, and performing data analysis on a variety of variables can be a daunting task for hoteliers who aren’t armed with quality data.

That said, with quality data and tools, hoteliers can generate high-quality forecasts with relative ease. There are major benefits to incorporating a user-forecasting process for your property if you so choose.

A user-created forecast is always a welcome compliment to an algorithmically generated RMS forecast and or a Business Intelligence tool’s forecast, as the more forecasts a hotelier can analyze and weigh, the more accurately they can predict their hotel’s likely performance.

If you’d like to learn more about the tools used in this guide, visit us here.