HolidayCheck AG has been operating a leading portal for travel and hotel reviews in the DACH region since 2003.
2 x
faster search results
12
minutes for automated deployment
10 %
increase in average yield per user session
5,000
fully automated quality assurance checks
Initial situation
HolidayCheck AG has operated a leading portal for travel and hotel reviews in the DACH region since 2003. Full-text search is one of the most important navigation tools for customers, whether they use the website or one of the native smartphone apps.
The quality of the results provided by the existing search services was not sufficient to provide the modern user experience expected in the highly competitive travel market. The speed of the search was also no longer acceptable. The code had become unmaintainable due to a long history and lack of testing. Any meaningful maintenance or even expansion of functionality was associated with considerable risk. Furthermore, there were different search implementations for the various search channels, i.e. the website, the mobile website, and native applications. Different results made it difficult for users and stakeholders to understand search results. This meant that there was no possibility of rapid, customer-oriented optimization.
HolidayCheck was faced with the task of fundamentally overhauling this infrastructure in order to be able to respond more quickly and reliably to all market requirements. codecentric AG's Munich office and its Elasticsearch experts were called upon to provide support with the Elasticsearch technology.
Solution
One of the major challenges was to support customers within a short period with a competitive search function that could also cover many travel market specifics, such as a multilingual index for hotel names.
HolidayCheck also wanted better protection against regression so that it could test search improvements faster, more often, and with greater confidence. A very large number of automated tests were created to simulate the real-world behavior of users in order to ensure the quality of the search solution before each release. For this purpose, over 5,000 search terms were extracted from website tracking based on various criteria (frequency of search queries, average revenue per search, etc.) and fed into automated tests.
User behavior is now also used to achieve higher quality of results. For example, clicks on search suggestions are incorporated into a dynamic ranking system in order to maintain a flexible and learning search system.
Continuous load testing was used to demonstrate in advance that the new search solution can respond to a significantly larger number of queries in a shorter time while using the same hardware. This meant that the launch was risk-free for HolidayCheck.
Result
In collaboration with the engineering team at HolidayCheck, the excellent knowledge of Elasticsearch and the software skills that the codecentric experts brought with them improved not only the performance and quality of the search results, but also regression reliability.
Pair programming and intensive reviews resulted in a very lean solution that more than met expectations. After a short time, the first A/B tests began, which, due to the positive results (a 10 percent increase in average order volume and higher click-through rates), quickly led to the legacy system being replaced in an iterative process. Thanks to the knowledge transfer that has been achieved, the internal team can continue to work at the same speed even after the codecentric team has exited the project.
The newly created, fully automated, and tested infrastructure means that new requirements can now be implemented very quickly, with a turnaround time of just days. HolidayCheck is now able to respond immediately to changing customer needs in the area of search.
Any questions about the project?
Are you interested in a customized solution for your company? Let's talk without obligation.
Head of Sales
Further projects of codecentric AG
Find out about other successful projects that we have completed with our customers. Perhaps you will find inspiration for a use case in your company here.
Jürgen Wurster
Head of Sales