TripAdvisor: Rapid product development and content generation
TripAdvisor grows to be the largest travel site in the world. Why? This post explains the secret of its strategy and operation model.
TripAdvisor is an example of effectiveness because it becomes the largest travel site in the world by rapid product development and content generation process.
TripAdvisor’s business model is to provide travel information, gather travelers , and monetize traffic. TripAdvisor provides over 200 million reviews about hotels, attractions, and restaurants and attracts travelers by enhancing search engine optimization, search engine marketing, partnering with other travel related sites, communicating its members via e-mails, and providing its application. TripAdvisor also offers a price comparison feature with which travelers can compare prices of hotels from commercial partners including online travel agencies (OTAs) and hoteliers at a glance. Clicking one of prices shown leads travelers to a partner and TripAdvisor receives commissions from the partner. This cost per click (CPC) model accounts for over 50% of TripAdvisor’s revenue.
In the internet hospitality industry, each player competes in getting travelers by various features such as price comparisons and incentive programs and providing travel information. Moreover, travelers go to other services by just one click when they are not satisfied with features and content on a site. Therefore, in order to satisfy travelers and win the competition, TripAdvisor has to keep improving its products at a faster pace and generating high-quality travel information than its competitors do. TripAdvisor’s operating model syncs with these strategic objectives.
In terms of product development, TripAdvisor streamlines its product development process. Each product manager owns certain product areas with key metrics such as traffic and conversion rates and works together with designers and engineers. This development process is a highly data-driven process. Product managers analyze data, define issues, ideate solutions, develop minimal viable products with engineers and designers, test the products on the development server, run A/B tests on the live site, analyze the results, decide whether to implement the products or not, and iterate this process. Especially, testing with minimal viable products rather than fully developed products shortens this cycle. With this streamlined process, TripAdvisor keeps its high development speed.
Moreover, the number of tests at the same time is also notable. Each product manager handles multiple products concurrently and hundreds of A/B tests are running on the site. The big difference between A/B tests at TripAdvisor and in Team New Zealand is the easiness of tests: virtual products are cheaper and faster to test than physical ones. To take advantage of this virtual product characteristic, TripAdvisor develops solid backend infrastructure and enables the fast and large-scale product testing.
In terms of travel information, TripAdvisor relies on its community. Travelers submit their reviews on properties by clicking the “write a review” button and completing forms on TripAdvisor. TripAdvisor, like Threadless, generates a large volume of content everyday with little content creation by itself.
Moreover, TripAdvisor keeps review quality high by moderating reviews. Submitted reviews go to TripAdvisor’s moderation process in which moderators confirm whether reviews are qualified and authentic. Firstly, algorithm assesses submitted reviews and judges if their quality is high enough. If the quality score is high, the review is published on the site. Otherwise, human moderators check the review manually and decide whether it should be listed or rejected. By using and constantly improving algorithm process, the company deals with nearly 300,000 reviews per day with a fairly small number of people. In addition, algorithm based moderation absorbs volume changed caused by seasonality as travelers write more reviews in travel peak seasons. With this combination of algorithm and human moderation process, TripAdvisor keeps review quality and deals with variability caused by seasonality. Moreover, this algorithm becomes more accurate with more data through machine learning process.
In conclusion, TripAdvisor’s operating model of product development, review generation, and moderation supports its fast and constant product improvements and content generation. The accumulated product improvements and reviews sustain the company’s competitive advantage. From TripAdvisor’s case, we can learn about how to continuously improve the service. Especially, the following tactical options are worth considering when we launch a web service: 1) rapid A/B testing, 2) community based content generation, and 3) automation by using algorithm.
Sources:
- TripAdvisor
- Why TripAdvisor Excels at Product
- My 2.5-year working experience as product manager at TripAdvisor.
Disclaimer: This article does not represent the company’s view.
Thank you for your post Aki! It is very interesting to see how TripAdvisor relies on customers to provide the content, but in return offer great value to customers that benefit greatly from the reviews others post. It can even be said that they become more valuable as they have more content and more information to share with customers. It will be even more interesting to see how they maintain the community they have created and their competitive advantage.
Thanks Carolina! Yes, I agree that TripAdvisor becomes more competitive as they have more content. Community management is one of key operations at TripAdvisor to maintain collaborative environment. The company has dedicated resources for that purpose.
Hi Aki, great post. I really love organisations that act fast and you described how TripAdvisor does that. My only concern is about the quality of the content – what happens if you move to fast and provides product that do not meet the quality criteria? Will that make customers run away and never look back? I guess that choosing the product managers has significant importance in such company where they have amazing independence.
Hi Shimon, thanks for the comment! Yes, that is one of the challenges I faced at TripAdvisor. For example, Hebrew is a unique right to left language, which tends to cause issues if we only care about left to right languages. In order to prevent such issues, we developed preemptive measures for potential errors. Actually, I was a guy who managed non-English site quality projects. Love to talk about them if you are interested!
Aki – Thanks for the post! I find it really interesting to get your perspective on the inner-workings of TripAdvisor. It sounds like an exciting place to work where the speed of the job matches the speed of the product you are providing. I am also fascinated by the community TripAdvisor has built. You mention Threadless in your post and I’m wondering – how does TripAdvisor incorporate customer input into their operating model. I see that there is a lot of A/B testing done, but does TripAdvisor solicit input from the community in other ways – either to improve the site or to increase the number of reviews?
I love TripAdvisor for planning my vacations – thanks for your work!
Hi Dan! Yes, I actually ran one project to solicit customers’ feedback by using UserReport. http://www.userreport.com/
We collaborated with the company and solicited feedback on our desktop and mobile site at first with a 10% of traffic, then rolled out the feature for weeks until we got enough volume. One interesting finding was that some customer needs are universal across sites. i.e. people love to see how their reviews are evaluated “helpful” by other travelers. We tweaked our review functionality based on that feedback.
I really liked this post, Aki!
It sounds like TripAdvisor’s edge over its competitors lies the quality and robustness of its user-generated content. 300,000 reviews per day is an incredibly large number. It’s impressive that such a large fraction of reviews can be vetted without any human touch. Do you know if TripAdvisor also has concerns about users being overwhelmed by the sheer volume of reviews? And beyond the daily human/algorithmic filtration of reviews, what bigger measures does it take to encourage the production of high quality content?
I would be really interested to hear more about the sorts of A/B tests that have led to larger changes and about how the team decides whether or not to implement a heavily debated new product. I like your point about A/B testing being much easier on a web application than on a raceboat, but it does seem that the might be reason to worry that too much testing might make for a worse user experience for the unknowing test subjects.
Thanks Will for your comment! To your first point, TripAdvisor has a review guideline and rejects reviews which do not meet that criteria (lack of details, bad language, inappropriate content, etc.) so that it maintains high quality content. How to show helpful reviews is an important point of user experience design. The company uses various criteria to surface helpful reviews up to the front (i.e. Amazon does a good job on this point).
I understand your concern that many tests may confuse users. This is actually true. Especially when we implemented a large interface changes, major metrics decline in many cases because existing users are confused. So, what we focused on was to accumulate small changes and improve metrics on a weekly basis. Major interface changes are not often: let’s say every quarter or every half a year.