Good bye to Miami

At least based on the experience of more than 360.000 user reviews Will Smith would have to rewrite his song text. Customer satisfaction for hotels in Miami Beach is significantly lower than for the other top 10 US destinations. Las Vegas is best in class, but Boston and Washington D.C. are following closely.

trivago combines in its hotel search besides its own user comments the reviews of multiple sources as for example US based tripadvisor, European holidaycheck or booking portals like expedia or Booking.com. We use this unique validity of our hotel rankings not only for our price comparison portal but publish continuously destination rankings for top European cities that get widely spread in the media.

For you I took a closer look at the top 10 US destinations, for which you can find 3421 hotels within the trivago database. We accumulated the ratings of 366310 user reviews for these hotels, which means an average of 107 reviews per hotel. I do not want to bother you with statistics, so I just tell you that 366.000 data points for ten cities lead to extremely valid results where also little differences can be interpreted. If one of you likes to calculate the 95% significance level for this sample, feel free to post it.



The grades you can see here are based on a scale between 10 and 100 points. Different portals use different scales for their reviews. trivago normalizes them between 10 and 100 to make them comparable. The hotels in Las Vegas lead the table with 77.9 out of 100 points, followed closely by Boston and Washington D.C. with 77.0 and 76.4 out of 100.

The difference to Miami Beach with 69.1 points might not sound not so remarkable, but you have to keep in mind that this is not the mark of a single property but of hundreds of hotels where a lot of effects can be leveled out. If you compare hotels on trivago you will realize that 9 points in the overall rating of a hotel make a huge difference in the perceived quality.

OK some very clever among you might say that this number might more give feedback about the hotel structure within the city than the overall expected quality of hotels in Miami Beach. A reason could be that there a lot more five star hotels in Las Vegas and Miami is more serving the lower market segment.

A deeper look into the data unveils that at least for a comparison between Miami Beach and Las Vegas this is not true. In the next graph you see the overall liking of four star hotels in the different cities. So all effects that occur due to a different structure between high and low class hotels are excluded.



Even so the average ratings are generally 4 points higher for four star hotels than for all hotels, the ranking shows pretty much the same results. Las Vegas is still at the top, Miami four star hotels with a 9 points gap at the bottom of the list. Interesting: San Francisico made a jump from rank 7 to 2 in the four star list. Here it proofs true that the bad results in the overall list is a reflection of the hotel star structure.



San Francisico has significantly more 1-2 star hotels and less 5 star hotels than for example New York City.

What may be an excuse for San Francisco does not hold for Miami Beach. The split between the different hotel star categories is pretty much the same as in Las Vegas. Nevertheless there is one additional difference. A standard hotel room in Las Vegas costs in February (due to the trivago price database) on average 150$, in Miami Beach 284$.

Brits want to see it all

trivago Europe maps unveil diverse travel behaviour of Europeans. While Spanish travellers more or less stick to their roots, British people like to travel all across Europe.

One thing you might never have been realising on trivago: The navigation structure is totally localised. So surfing on the British version gives you a different experience than surfing on the Italian version. The dots shown in the map are a reflection of the local travel behaviour. That gets even more interesting when you compare the same region. You can do that with each area and will discover interesting differences. The map of Europe is only one example.

This are the British top destination on trivago.co.uk. The dots are spread widely all across Europe. The top tourist destinations like Canary Islands Andalusia, Majorca or Tuscany are covered. The top cities are London, Paris, Berlin, and Madrid.


A totally different view on Europe is offered on trivago.es. The Spanish travellers are much more home focused. You can see many orange Region marks that cover the different Spanish holiday areas. Only a few country marks (green) outside of Spain.


The German Europe map on trivago.de is somewhere in between. It has a strong weight on German Austrian and Swiss destinations (probably because its winter season). Nevertheless the main European countries are covered (including Turkey). The Germans share the British passion for Majorca and the Canary Islands.


Interesting on the Italian map of Europe on trivago.it: We see only rare dots in Spain. London, Paris, Berlin and surprisingly Prague are the most important metropolises outside of Italy. As the Spanish people, Italians really like to travel in their home country,


The French focus in Europe is currently on the French Alps and the South of France.


A contrasting view on Europe: The map of the continent on trivago.jp. On one hand you see much more blue city dots that cover nearly all metropolises. on the other hand you discover a clear focus in the middle of Europe. Japanese travellers seem not want to go far from established tourist routes and much less visit the south, totally ignoring countries like Greece, Turkey and Portugal.

TV impact on visitor curve

As our January campaign came with a bigger budget, we had the chance to do research on some major prime time spot events, as for example during the German TV premiere of "Rush Hour 3" and "Next".

We combined the visitor curve 10 minutes before to 20 minutes after airing of 18 prime time spots. Each of them had a minimum reach of 1 Million viewers. You can see the results in the attached diagram.


The red curve shows the unique visitors. As you can see the effect already starts during the spot, with a peak in minute one. Interesting is that the short term spot effect ends very soon. After only 5 minutes more than 95% of the effect is covered. The search event (green) shows nearly the same behaviour only 2 minutes delayed. The curve of the click out (sales) event is much more flattened with a little peak in minute four. The effect is still very visible after 20 minutes and holds on for an even longer time.

Even so you could generally expect an outcome like this, the extend in which the effect of TV triggered visitors can be cut down to only five minutes after the spot surprised us. And so it led to some major changes in the way we measure TV success.

All effects are factor related to each other. These factors are quite constant over time as long as you do not change something in the structure of your page or do a big change in the creatives. So you can always calculate sales as a factor of searches as a factor of visitors.

Some might say that sales is the only important figure to measure. Sure. The question is what is the best way to measure it.

It depends on what you do, but sales events might be pretty rare. With let's say a visitor conversion of 1% you have to attract a lot of visitors with a spot to get into an area where volatility is not an important factor anymore. For example a prime time spot reaching 2 million visitors might results in 1000 short term visitors on your web page and with a conversion rate of 1% lead to 10 sales. Sometimes you will have 5 and another time 15 sales, but is that something you would like to base your decision on if a spot does pay out?

With a conversion of 1% the accuracy in which you can measure visitors is 100 times higher than the accuracy of measuring sales events. If you take the finding above into account, that the distribution of sales event is much more flattened as the peak in visitors, this ratio gets even higher.

Doing a TV campaign with 50 spots a day you will definitely not be able to separate the sales effects of each spot, but you will nearly always be able to separate the visitor effects. So the failure in your data by extrapolating visitor to sales events will be much lower than the failure of attributing a to high/low effect due to the randomness of small numbers.

New trivago TV spot

Since end of January we released the second Version of our TV commercial. We changed only little details, but it seems to work even better than the spot we started with end of December. Thanks to the guys from fat lady , who already did great work in the production of the first spot, we found a new voice for trivago. It creates exactly the authentic character that we had in mind when we started, and it fits even better to the relaxed atmosphere of the music. Max and Becko heard at more than a hundred different voices and we casted 11 of them personally (one of them starsearch winner Martin Kesici) at production day. Really tough decision. Different characters, nearly all of them would have made a great fit.

Also many thanks to the creative people at omstudios. This times we had to do only little changes but they were again right to the point. Great that they could fit us in their busy schedule although they had to work long hours for the Audi A8 presentation in Miami.