Outliers

Malcolm Gladwell wrote a book called Outliers a few years ago that examined what makes people successful (his other book, The Tipping Point, has become a must-read for many businesspeople worldwide — particularly those involved with marketing).  Outlier is really a statistical term that describes an observation that is numerically distant from the rest of the data.  Sometimes this indicates a measurement error, other times, it indicates the existence of a heavy-tailed distribution (which Chris Anderson also wrote a book about called The Long Tail which described how the Internet changed business).

Recently, in my TripAdvisor Master Class, I talked about the idea of Outliers and how they applied to travel.  The responses were so positive that I thought I’d share one of the examples here.  We analyzed English reviews that were written about hotels in each city on the TripAdvisor tour to see how positive or negative the sentiments in the reviews were.  As we are independent from TripAdvisor, we have our own technology to do the analysis that’s vastly different and in more detail than what the summary numerical ratings provide.  To make a long story short, our technology basically rates each applicable review on a scale of positivity by auditing the words used by reviewers to describe their experience.

Here then, is a distribution graph of reviews over a 16-month period on our scale.  This graph here is about Bangkok hotels.  The X-axis represents the range of rating, and the Y-axis counts the number of reviews we found in each range.

Interestingly enough, the distribution of reviews by the positivity of sentiments expressed resembles a normal distribution curve.  I won’t get into the actual ranges but I will say the ranges covered by the yellow rectangle are neutral to slightly positive and comprise of the majority of reviews.  This backs up TripAdvisor’s assertion that more reviews are positive than negative, and that the average rating (when you only consider TripAdvisor’s scale) is above 3 (average) on a scale of 1 to 5.

The area in the graph where I denote ‘Outliers Here’ is where it gets interesting — in that the reviewers were extremely positive or negative (i.e. different from the ‘norm’ reviewers covered in the yellow area).  Who were they?  What can we learn about them?

So we did an analysis to see of the countries that contributed outlier reviewers, which countries had the most positive to negative ratio?  And which ones had the most negative to positive ratio?

So it turns out that for a set of primarily English-speaking countries, the United Kingdom had the highest positive to negative ratio (roughly 6 positives to 1 negative).  Singapore, on the other hand, had the lowest (2.7 positives to 1 negative).  For a different set of primarily non-English-speaking countries, Finland was most positive (7 positives to 1 negative), whereas France was most negative (3 positives to 1 negative).

And what about Thai people?

You’d be surprised to find that Thais were actually more negative than the French or the Singaporeans (2.4 positives to 1 negative).  The interesting thing was… the hoteliers in the room were not surprised by this discovery.  Perhaps we’re more critical than visitors of hotels in our own country?  (by the way, in a separate analysis on a few more countries this hypothesis held up…)

So what does this mean for hoteliers?

Well for one, our independent analysis corroborates that TripAdvisor reviews (at least in the cities we looked at) is generally more positive than negative.  This isn’t surprising when you look at their guidelines for posting, which actually prohibit insulting language and profanity… hence reducing the number of outlier negative comments.  Given that TripAdvisor rejects a certain number of reviews they consider to be fraudulent or inappropriate, the general positivity makes sense.

Also, in knowing the outlier behaviors by country (or by other characteristics like age or gender), is interesting for reputation management.  To date we’ve focused a lot on improving operational performances.  But some criteria are culturally subjective (e.g. when is staff ‘overly’ friendly? how large and varied do the selections need to be for the breakfast buffet?).  So doesn’t it make sense to have a strategy for managing reputation in social media with a cultural point of view as well?

Social media is interesting in that if you’re able to focus, you can find rich data about people that were previously inaccessible.  The proper use of this to delightfully surprise guests — especially the ones that have statistically shown to be positive outliers — can yield rewarding results.