Talend Integrates Global Data of Online Food Ordering Service Delivery Hero

Big data for a better delivery service
Talend is easy to deploy and adapt, offering excellent flexibility and stability. And with the connectors that come with Talend, linking up our source systems is child’s play.
Brian Leonard, , Head of Data Warehouse at Delivery Hero

The customer

Order Takeaway Online! That is Delivery Hero’s recipe for success. Founded in Berlin in 2010, Delivery Hero is an online food ordering platform that lets customers order takeaways from participating restaurants on their computer or smartphone. Today, Delivery Hero is one of the world’s most successful online food ordering platforms. It employs around 650 people in 21 countries on four continents and has established partnerships with over 60,000 restaurants. The website of each country features the local brand that customers are familiar with. In Germany, this is Lieferheld GmbH, a wholly owned subsidiary of the global Delivery Hero Holding.

Data management is a key success factor for Delivery Hero. After all, if you don’t know what your customers like, you can’t meet their needs or stay a step ahead by quickly responding to changing preferences. To secure its future success, the company needed to integrate and analyze its global customer and order data in real time. Delivery Hero turned to Talend and its integration technology to help achieve this.

The challenge

Why do more people opt for pizzas in one country and Asian food in another? Why does a high number of orders come in between 10 pm and 11 pm in most countries – yet in some places practically no sales are registered in this time? And how do you manage the perfect mix of participating restaurants? These are just some of the questions that business analysts at Delivery Hero ask themselves every day. On one hand, they want to replicate one country’s success elsewhere, and on the other they want to pass on revenue-generating tips to their restaurants.

“Our success is built on a wide choice of restaurants and fast customer service, but also – and even more importantly – on our ability to evaluate our many data streams,” explains Brian Leonard, Head of Data Warehouse at Delivery Hero. “The raw data is there, but it is scattered across 21 countries and stored in a wide variety of formats. So our challenge was to find a way to harness this big data.”

In actual fact, the business analysts were able to evaluate local data previously, but they had no insight into data from other countries. This was because Delivery Hero did not have a central data warehouse and many heterogeneous databases are in use. These include everything from MySQL through PostgreSQL to Excel.

What was needed was a secure and flexible solution to collect local data from around the world, transform it into a suitable format and upload it to a central data warehouse – preferably within minutes to ensure that recent data is always accessible. In addition, Delivery Hero needed to integrate its CRM solution Salesforce with its backend database. Other aims included an API for big data analysis and support for tools like Excel, BIME and GoodData. This would extend the ability to create and access reports beyond business analysts to departmental users.

Delivery Hero initially turned to Pentaho and shell scripts to meet its requirements. However, these solutions fell short of their performance and flexibility expectations. This prompted the data warehouse team to search for a data integration solution.

The Talend solution

Talend’s solution was the first to be tested based on one team member’s positive experience with the platform in the past. To see if Talend was up to the job, a Talend consultant was asked to come on site and demonstrate how Talend could be integrated and deployed in the existing IT environment. The Talend expert linked up databases from three countries, downloaded their data and transformed it “live” to a local database in Berlin. Test analyses were performed on this data set to confirm the proof of concept (PoC). “Since the trial was so successful, we did not need to carry out further market studies. We were particularly impressed by the fact that Talend’s open source solution is based on Java,” recalls Brian Leonard.

The actual process of implementing Talend took a few days, during which the necessary adjustments were carried out. As a traditional ETL (extract, transform, load) solution, Talend could collect modified data from 14 countries. The initial stage of the process is called the “pre-staging layer with ETL”. In some countries, data is downloaded to the data warehouse every five minutes, which ensures that up-to-date data is always available. After that comes the “staging layer with Talend”. Here, Talend transforms the data from the first layer into a uniform format and loads it to the data warehouse. Less than 15 minutes later, Delivery Hero has all the data it needs to carry out meaningful analyses. Every month, something in the region of 20 gigabytes of data is modified or created – and this volume is rising all the time. “Talend does more than create native link-ups to our databases, it also gives us excellent connectivity to the cloud app Salesforce.com,” points out Brian Leonard.

Benefits

Since the start of 2014, the Talend solution has been extracting data from the countries in which the Delivery Hero is active and loads them into the data warehouse in Berlin. With this central repository of data at their disposal, the business analysts can now run any type of query and use the results to steer the company’s future course. But increasingly it is not just the analysts, but employees from other departments who are taking the opportunity to turn big data into actionable insights.

“The great thing for us is that data integration is as easy as drag & drop, and there is no need for traditional programming,” enthuses Brian Leonard. “Even in cases where programming is required, everything that can be done with Java is also possible with Talend. This is a major benefit for us, because we can call on in-house Java expertise.” Today, Brian Leonard’s team is able to create new jobs in less than 30 minutes and the solution is running smoothly. “On the rare occasion when an error message does show up, we tend to now examine the data sources straight away instead of turning to Talend.”

Ultimately, the data integration solution has brought the company huge time savings. This is crucial, because big data is getting bigger all the time as well as the variety and diversity of the data. “Thanks to Talend, integrating a new country has become child’s play,” says Brian Leonard. “This is because we can re-use existing jobs and processes. An unnecessary bottleneck has thus been removed, paving the way for future expansion.” Now there are plans to deploy Talend for processes other than ETL. Previously, the company managed data quality and MDM with its own solutions, but it is going to evaluate the functions integrated in Talend and adopt them if the trials are successful.