Continuously rethinking revenue management, I have come to the point where I wanted to build a revenue management and pricing solution myself. One of the crucial elements was the recommended pricing. Together with two revenue management teams and technology specialists, we have established a significant list of elements that we wanted to incorporate when it comes down to how we decided on any price for any given day.
The list of elements keeps on expanding and it seems never ending, as we add new elements on a regular basis.
So, if you are thinking of adding new elements in the way you decide on your price strategy, here’s a bunch of ideas to play with.
Automated pricing, price recommendations, price hurdles. Dynamic pricing, market penetration pricing, price skimming, psychological pricing, optimal pricing, bundle or package pricing, promotional pricing, time sensitive pricing, geo-pricing. Sounds familiar? I have tried and tested them all, over and over again.
Since our aim is to build the most dynamic and bespoke yield management solution the hotel industry has seen, I came up with the following items which I believe are worthwhile to be incorporated in a pricing system;
Main data types to incorporate
I am a strong believer that incorporating the wrong data in a price decision is a business killer. Far too often I see too much dependency on how the competition behaves, whilst competitor behavior is rarely the cause of severe direct changes in demand. Hotels buildup demand from a variety of market segments, and the full business behavior should be considered when making pricing decisions. A good system will consider a large a mix of main ingredients in order to make a pricing decision such as:
- On the books data
- Pick-up or pace data
- Average rate of pick-up
- Public pricing behavior
- Pricing of your direct competitors
- Market segmentation
The last item is important to the extent that the influence of non-yieldable and semi-yieldable performance of market segments or channels influences pricing possibilities.
Analysis and decision logistics
Once you have gathered the right data to make decisions, a second logical step is to ensure that you apply the right analysis. Randomly looking for patterns in data might work, but demand builds up in similar ways based on seasonality, day of the week, market segments etc.
Hence a pricing analysis should:
- Look for patterns that matter
- Have algorithms that are built to understand trends
- Allow for manual user overrides
- Measure pricing behavior of similar days (of the week)
- Allow linking of dates that are believed to show similar behavior (i.e. event dates)
A price grid usually consist of different levels of pricing which are separated by hurdles. Pricing grids can work fantastic, but the challenge is that they often only respond to a number of rooms being sold for any specific day. Usually when a threshold is hit, prices move up a level.
When prices are only moved up, it is a one-directional action, assuming demand also goes up. Demand however can stagnate and decline. And a price grid should respond to that.
The price grid must also respond to other demand indicators than rooms sold, being: number of days prior to arrival, and overall demand for a property. In other words:
- If demand goes up, prices can go up.
- If demand goes down, the grid should allow for prices to move into a lower level again.
- If the hurdle is too big, demand stagnates and we get closer to arrival date, the pricing should be readjusted.
Again, as I mentioned prior, pricing grids can work fantastic. Especially when properties recently started with revenue management or are hesitant to become more aggressive with pricing decisions (in a further future). A good pricing grid responds to sales in the very far future, and thus does not require a visual or hands on control of the next 12 months all the time.
To predict yield opportunities, a software uses scientific algorithms based on an extensive amount of data. Dates that require your attention are highlighted and indicate recommended actions to maximize revenue.
The algorithms focus on numerous elements that influence demand and pricing, like:
- Pace of sales/ booking pace – or in other words, how fast am I selling compared to other similar dates
- Business on the books patterns – or how many reservations did I have on the books for similar days, and at which rates
- Pricing patterns – did I raise or lower rates in similar fashion for other similar dates, and did reservations come in with similar patterns?
- Behavior of the property comp set – on average, is my comp set raising rates and lowering rates like your property, or are they indicating to have different demand levels?
- Exception rules for standard deviation – meaning that information for i.e. certain similar dates, that fall outside of a range of similar behavior, do not get included as valid date for analysis. I.e. your pricing for the same day last week was 20 percent higher than this week, hence the data are not incorporated for recommendation.
Now, despite the complexity of most algorithms, the visual output and related pricing should be relatively easy to understand. Hence a (visual) aid should be available to comprehend algorithm output.
Many providers unfortunately completely hide the logic behind their algorithms, only creating more questions than answers.
AI (Artificial Intelligence)
Artificial Intelligence, or machine learning, is likely one of the hottest developments in data analysis. I yet need to find the machine learning solution that is applicable in a large extent to hotel pricing and which has a standalone basis. We have however incorporated machine learning as part of a larger scope of pricing recommendation logic. A pricing solution nowadays should incorporate artificial intelligence that distinguishes user patterns, user overrides and patterns that go beyond the logic of algorithms and complex calculations.
I have an adversity for systems that analyse things without allowing the user any insight in how a calculation and outcome are composed. These so-called magic boxes are full of overly complex formulas and unclear answers. I have faced too many solutions that will not uncover any of the methods or formulas on how they get to a price calculation. I am more for providing a user insight in how a pricing decision is composed by showing core data, and by showing the underlying logic of algorithms.
Multiple steps in pricing logic
- Gather reservations/ on the books data
- Have different price grid options
- Apply algorithms/ calculations
- Decide on a final price recommendation
- Allow the user to override recommendation
- Use AI to learn from the algorithms and user overrides, so this new knowledge can be added to step 4 to allow more fine tuned recommendations
Similar to user insight, users need to have an understanding on how the system can be influenced. Most users have personal thoughts on how pricing calculations should be made. I.e user number 1 wants to incorporate a bit more of benchmarking and a bit less on the books information. User 2 goes for more pick-up weight and artificial intelligence. Number 3 heavily relies on price grids and hurdle models. Hence personalization is crucial. A good solution will:
- Allow users to influence algorithms based on personal knowledge and local knowledge
- Allow users to give weights on elements that lead to pricing decisions
I.e. OTB and pick-up have significant weight in the decision process when coming to a new price. For example 75% of the decision is based on the patterns generated by these two elements, while benchmark information only adds a 10% weight to the pricing decision, or 25%, if the user desires that instead.
Level of detail of pricing recommendations and pricing decisions
Pricing systems may have different levels of detail when it comes down to recommending prices. Various levels of recommendations come to mind for the ideal solution. Such as that the system;
- Recommends pricing decisions per segment/ channel
- Allows for automation of pricing for specific periods of time and specific stay periods
- Allows for pricing recommendations per room type supplement, or a fixed supplement setting
- Understands length of stay pricing (and dynamically via artificial intelligence)
- Assists with management of room supplement pricing and allows for monitoring of sales
UI (User interface) & opportunity identifier
Because pricing often is a result of the analysis of large quantities of data, it’s easy to drown in the data or to get lost in the black and white numbers.
Hence a pricing solution should have visual aids to understand trends in demand that influence pricing on a specific day or period.
In addition, pricing decisions are opportunities. Opportunities require action.
Hence an opportunity identifier, that points users into the direction of action, is a welcome part of a pricing solution.
The Final Word
Summarizing the above, I have been confronted with thousands of questions and challenges related to tools for pricing, yield and revenue management. Having done revenue management for this many years and having a career in the industry, I also came up with my own set of questions, and I have been digging and looking around for answers.
During my search for answers for the best pricing and revenue management tools out there, I have met with people that understand yield in detail, and with less experienced people that shared visions about pricing, often just as insightful.
And obviously, having a decent chunk of revenue management experience, and completely and utterly unable to keep my big mouth shut, I must admit, I have been critical of many theories and solutions.
I ended up building a system – Hotel Scienz – together with a number of colleagues from Xotels and Busy Rooms. And now it’s my turn to receive criticism. I am already receiving thoughts from plenty of clients that have come aboard with Hotel Scienz over the past years. So why not gather some ideas and challenges from the most influential hoteliers and industry players out there?
I look forward to hearing additional thoughts. What would your ideal pricing tool consist of?