Term Frequency Analysis – Part Two

Yesterday, I shared an interactive way for you to see how often popular words appear in online reviews. Today I wanted to share some observations from searches that I performed.

This first graph, below, showing the graph for the word, “awful”, shouldn’t be too surprising. When people have a bad time, they use the word “awful” frequently in their reviews. It becomes less and less popular (trend line moves downward) as the review rating increases. This is obviously because guests use more positive words to describe their stay when they have a great experience.

The second graph compares the word “awful” and the word “incredible”. The trends are almost exact opposites. Again, no surprise since people will rarely use the word “incredible” to describe a bad stay, and rarely use the word “awful”to describe a good stay.

The third graph shows four words that people associate with a good stay: workout, romantic, spotless, juices. While I wouldn’t be a proper Data Scientist without pointing out that correlation does not equal causation, these graphs appear to suggest that these amenities lead guests to have a pleasant stay.

The fourth graph shows four words that people associate with a bad stay: sheet, wallpaper, curtains, payment. Again, correlation does not equal causation but it seems unlikely that you will receive a positive review if you have issues with any of these four aspects of a guest’s stay.

We have already used Term Frequency Analysis to assist in answering some of the following questions:
– How do guests feel about service fees in my properties compared to the rest of the market?
– Is there a type of guest that frequently stays in my market that I am not catering to?
– Are there trends happening in my market that my guests are not aware of?
– What do people expect when they come to my hotel versus a rival brand?

This is just the tip of the iceberg of Revinate research. Let us know your suggestions for other areas of research – or better yet tell me in person at the Hotel Data Conference from September 3rd – 5th.

2 responses to “Term Frequency Analysis – Part Two

  1. Could you explain how the graph works please? To me it seems the numbers on the Y axis around the wrong way.

    The popularity of awful is larger for reviews that received a rating of 5 than it is for reviews that received a rating of 1? Is 1 a postive review?

    1. Hi Rhys,

      Apologies if the graph is confusing. The popularity of a word is ranked similar to TripAdvisor Index or iTunes Rankings. So a ranking of 1 means that it is the most popular word for that rating, and a ranking of 2,500 means that it is unpopular.

      The review values are modeled on the scoring system that TripAdvisor uses, with 1 = worst score and 5 = best score.

      The word ‘awful’ is more popular in 1 star reviews than 5 star reviews so we felt it made sense for it to appear ‘higher’ on the scale at 1 star than at 5 stars.

      Thanks for the comment and happy to answer any other questions you have on the data!

Leave a Reply

Your email address will not be published. Required fields are marked *