Trivago Bought An AI Startup That Eased Peoples’ Fears About Giving Up Data

The hotel search engine acquired the German machine-learning startup Tripl, which harnesses users’ social media activity to make recommendations.

Trivago Bought An AI Startup That Eased Peoples’ Fears About Giving Up Data
[Photo: Flickr user Anne Worner]

All types of online services are scrambling to meet the AI-driven personalization standard set by Amazon and Netflix. Fast-growing international hotel search engine Trivago is taking that path by acquiring machine-learning startup Tripl, which analyzes users’ social media profiles to recommend itineraries and link to offers from third parties like hotels or tour operators.


“It’s an angle that we so far have not had that much experience in, and the acquisition allows us to catch up in that dimension,” says Trivago’s CFO, Axel Hefer, who did not disclose what the company paid for Tripl.

But personalization requires personal information, something consumers in the two companies’ home territory of Germany are especially guarded about.

Tripl learned about privacy concerns through trial and error, says founder Hendrik Kleinwächter. With the acquisition, he’s now a software engineer in what Trivago calls its “user profiling team.” That may sound ominous to privacy hawks, and Kleinwächter says any gathering of user data requires being up front and providing transparency.

“Basically, if you tell the user really what you are doing with their data then people are more open-minded, but you should definitely tell the user what’s going on,” he says.

User signups jumped when tripl offered users options.

When Tripl launched in July 2015–mostly to a German audience–users could sign up only through Facebook, potentially sharing oodles of profile information such as friends, relationship status, hometown, and birthday.

“The key feedback we received from the users was, ‘Hey why can I only register with Facebook? This is kind of annoying,'” says Kleinwächter.


But in October of that year, Tripl added the option to instead sign up with an email and answer a short questionnaire about preferences–a simple thumbs-up/thumbs-down for vacation goals such as: Culture, Party, Relaxing, Romance, and Luxury.

Adding this option lowered the rate of people who bounced off the site without signing up. But the surge in sign-ups wasn’t simply due to new users opting for an email login over Facebook: Registrations through Facebook Login also went up by 23%.

“Based on our experience, we did an experiment, and people then started to use Facebook more often because they had an alternative,” says Kleinwächter. “In the one case you are forced to use Facebook, and in the other you have choice,” adds Trivago’s Hefer. “And choice always feels better than force.”

Kleinwächter believes that people warmed up to Facebook registration partly because it required less typing–just two clicks in most cases. He also thinks that users saw the benefit in providing more information for better recommendations–at least after Tripl explained it more clearly.

That’s borne out in other industries with far-more sensitive data, such as WebMD’s collecting of user health information. The trade-off between privacy and service is clear for people stumped by a medical problem, and users have opted to give up a lot of details in the hope of finding help. WebMD’s straightforward privacy policies were crucial to the success, concluded a study by Deloitte Insights.

“We actually did a couple of A/B tests” with language around how data would be used, he says. “We tested based on what we put on our landing page, like giving the user better explanations. Then the sign-up rates increased [by 30 to 40 percent], because they knew exactly what we were doing with the data.”


Tripl didn’t save copies of all the previous wording and iterations, but the current version of its privacy policy is straightforward at a merciful 660 words–breezy for German legalese. Trivago’s privacy policy, by comparison, runs about 1,750 words in German and about 1,900 in English.

The gist of its policy: The user data Tripl collects goes into its tailored recommendations. The brief text explains that data might go to another firm if the company is purchased (as just happened with Trivago), and that it could be shared with law enforcement bearing proper warrants. The privacy page also details how people can delete Facebook cookies so the social network can’t follow their activities on Tripl, and explains users’ rights under Germany’s strict federal data protection law.

Users who don’t want to connect to Facebook can fill out a survey instead.

Will AI Keep Trivago Growing?

Despite some hiccups, Trivago has been a powerful success story. But as AI and other technologies upend all online services, it could fall behind in a heated race for your travel dollars.

Founded in 2005 for travelers in Germany, Trivago quickly expanded throughout Europe by 2009 and now searches 1.8 million hotels in over 190 countries, says the company. Trivago is a meta search engine, indexing the hotel rates offered by over 200 booking services, on a platform available in 33 languages. Although Expedia bought a majority stake in the company in 2012 for about $632 million, Trivago continues to index all providers of hotel bookings and may send users to Expedia competitors.

The site opens with Google-like simplicity–a white page with a small, multicolor logo and a search bar where users enter a city name. Beyond the home page, though, is a shopping geek’s paradise, with well over 100 filters. They range from the elementary like star ratings and price ranges, to nitty-gritty items including availability of beach umbrellas and a cosmetic mirror, to lifestyle aspects such as suitability for “party people.”

Trivago’s advertising has been equally down-to-business. Most commercials feature a plain-dressed “Trivago guy” or “Trivago girl” pointing at superimposed parts of the website like hotel listings or preference sliders: essentially, thriftiness porn. Ads have diversified in the past couple of years, with one in the madcap style of Wes Anderson’s The Grand Budapest Hotel, for instance, and another featuring dreamy sequences of a same-sex couple in Barcelona. But Trivago’s popularity comes from hard numbers like exhaustive search options or exclusive pricing deals, not a trendy brand image.


Until recently, that’s been more than enough to satisfy investors. A 2016 IPO opened at $11.20 per share, a value that more than doubled by July 2017. But the price is now back around $11, after Trivago revised growth estimates down on September 6–from 50% to 40% per year.

Trivago offers extensive search options but not much guidance helping people figure out what they want to search for.

For all its success, Trivago’s edge depends on fickle factors that can change quickly, like a drop in advertising on the site. And regardless how extensive its features and effective its algorithms are today, Trivago is locked in constant battle against well-funded competitors like Priceline, TripAdvisor, and Google itself. A company’s dominance is fragile if it’s based on a technical performance edge in an era of ever-evolving technologies.

In acquiring Tripl, Trivago may be deepening its offerings beyond pure thrift and utility, beyond checkboxes and sliders, beyond the people who already know just what they want: AI could help the rest of us figure out what we want. For recommending destinations and itineraries, “any data that you have in addition is obviously helpful to refine your search algorithm,” says Hefer, the CFO.  (New services are in the works, he says, but he won’t give any details.)

RelatedThe Subtle Power Of Default Choices

At the very least, Trivago will have to convince potential customers that its searches beat the competition because the company knows travelers best. And Trivago can only do that by making people feel comfortable sharing more data about themselves.

About the author

Sean Captain is a Bay Area technology, science, and policy journalist. Follow him on Twitter @seancaptain.