In a recent article we introduced some ideas that run contrary to industry wisdom on Hotel Comp Sets. The competitors that hotels include in or exclude from their comp sets influence the results of competitive analysis. Stakeholders in hotel performance should therefore treat comp set selection as a bigger priority than they do today. In this article we use data to show how we know this to be true.

An idea currently doing the rounds in the lodging industry is that hotels can judge the relative quality of their comp sets by understanding which competitor hotels also name them in their comp sets. If a hotel’s comp set contains a high proportion of properties that also name it in their comp set – so the logic goes – then it must be a good comp set. We argued in our article why that is a dubious basis for comp set analysis. A good comp set is one that reflects which competitors a hotel competes with the most directly for business – i.e. bookers. So why judge the quality of a comp set on name-back frequency, which is invisible and therefore irrelevant to hotel bookers?

We promised to begin running our own analysis of comp sets. First we wanted to test the underlying assumption that hotels generally do a good job of choosing representative competitor sets. A good test for that assumption is to compare a sample of STR comp sets to the comp sets generated automatically by a statistical program.

Hotel Compete performs hotel comp set analysis using current hotel and market data, drawn from the sources that bookers use to shop hotels. The process – which is designed to identify which hotels compete the most directly for business – is automated and runs weekly. This provides the opportunity to monitor fluctuations in comp set composition, meaning that when market dynamics change, so should a hotel’s comp set.

For example, if Hotel A does not typically consider Hotel B as a direct competitor, but Hotel B reflags, or radically changes its long-term pricing strategy bringing it into direct competition with Hotel A, then Hotel A should consider adding it to its comp set. The problem today is that no good mechanisms exist in the industry to scan for and recommend changes to hotels’ comp sets.

Hotel Compete uses its process to monitor when a hotel has moved in or out of a hotel’s comp set for more than three weeks in succession. If the changes in market dynamics appear to be permanent – ie not attributable to a temporary blip – then a change to the comp set can be recommended to users.

This process provides a good test for the quality of hotels’ STR comp sets, because the more reflective of current market conditions a comp set is, the fewer comp set changes should be recommended. We ran an initial analysis of 2,833 hotels to compare the difference between current STR comp sets and the comp sets generated automatically by Hotel Compete. The results are in the table below.

Interestingly, the average number of competitors chosen by the statistical process is generally higher than the average STR comp set. This may well be to do with the confidentiality restrictions that limit the number of competitors that can be considered in a STR comp set. Predictably, though, the changes – ie hotels added or dropped week-over-week, are significantly higher when compared to the STR comp sets. This suggests what we intuitively know – that the Hotel Compete process started from a more dynamic baseline than is normal in STR comp sets.

Next, we analyzed the reasons for the changes. Recommended comp set changes are always triggered by some event in the market – in the table below we summarize what the events were. Unsurprisingly, most changes are to do with changes in hotels’ rate strategy. Interestingly, though, the number of competitor change recommendations owing to rate strategy is proportionately higher based on STR comp sets. This suggests that selling rates may be under-emphasized in hotels’ STR comp set selection. Again – this observation that makes intuitive sense, as STR reports focus on ADR, which is a measure of revenue, and not a reliable reflection of a hotel’s selling rates.

Of course, to do a thorough job of comp set analysis, it is important to understand the differences between the STR and HC comp sets. Initial analysis suggests that the differences between the comp sets are substantial – with only 51% average overlap rate for branded hotels and 42% for non-branded. To understand comp sets we must understand not which hotels are in other hotels’ comp sets, but which ones are left out. We will return with more detailed analysis of that critical subject in the next few weeks.