- Less time configuring, more time analysing
- Tasks that took 60-70% of data team’s time now automated
- Eradication of data loss and bad data
- More time to fine-tune self-service reports and A/B testing
- Pipeline: Fivetran
- Source: MySQL, Zendesk, Salesforce, Stripe, Google Ads, social media...
- Destination: Amazon Redshift
- Business Intelligence Tool: Looker
Spanish start-up Wallapop runs a thriving second-hand marketplace via mobile app and web that customers use to snap pictures of items to sell. Since launching in 2013, the company has built a base of 30 million users, trading more than 100 million items across Spain.
The Challenge: Extraction Distractions
A data-driven business that only makes decisions supported by analytics, Wallapop has no shortage of sources for insights. Around 15 million active users per month perform more than eight million searches per day across a catalogue of several million products. The Business Intelligence unit is responsible for analysing everything related to growing the customer base, and shares reports with teams in Shipping, Sales, Advertising and Marketing.
More than half of the people who work at Wallapop have access to data and use it to inform decision making, particularly at the board level. Enabling a fast and smooth extraction path to the data warehouse from key databases was essential – from MySQL, which collects all the data from the Wallapop app, and Zendesk, which captures customer service interactions and other sources like social media.
The problem was proprietary code in MySQL that made ETL time consuming and inexact. Ongoing product tweaks in Zendesk required the data team to always stay up to speed on the latest iterations. Both databases were workable, but it was taking too much time to extract the data and deliver it in a cleansed format into Amazon Redshift, the company’s cloud-based data warehouse.
“Our analysts were spending 60-70% of their time just making sure that the data was uploaded properly. You are always wondering whether or not the data is correct, which is simply not the right way to work,” said Sergio Rubio, Business Intelligence Manager at Wallapop. “We wanted to spend more time on performing strategic analysis.”
No More Configuration Complexity
Fivetran has made it happen. An alternative Amazon product was originally considered and tested with Zendesk, but the solution lacked depth and capabilities, “and was not at Fivetran’s level,” according to Rubio. The automation is what has impressed him most. When his team grants permission for Fivetran connectors to transfer data from MySQL to the data warehouse, database tables are preselected, cleansed, organised and then synced at a preferred frequency – be it every five minutes or every 24 hours.
The data team no longer wastes time manually creating ETL processes, extracting data and uploading, because Fivetran uses pre-built schemas for the databases to do it all automatically. Any new tables or columns are synced with the warehouse with no duplication. There is no complexity in terms of configuration or impacting the performance of databases as transfer occurs; it’s now a process that runs in parallel.
“You don’t have to spend a week configuring it and stop everything, leaving you without access to the data. In that sense, Fivetran is extremely useful and easy to use,” said Rubio.
More Time for Deeper Analytics
Data analytics lifecycles define how information is gathered, processed and eventually analysed for more insights. Optimising each stage will accelerate time to value, and the Fivetran piece is often the most valuable of all. No data has been lost since Wallapop has been using Fivetran -- a direct consequence of eliminating manual configuration -- and there is real trust in the quality of data that’s loaded into Redshift.
The biggest benefit is that it has allowed the Wallapop Business Intelligence team to focus on business-critical analytics, carry out more A/B testing, and help identify process improvements that will deliver tangible business benefits.
“One of the objectives of our team was to facilitate self-service analytics in Looker. Instead of making sure that the data is correct in Redshift, we can focus on making that happen,” said Rubio. “The Fivetran connector enables fast connections to get all the data we need into our data warehouse. Since we have been using it, we have greatly reduced the impact of event loss or not being able to trust the data. It gives you a lot of peace of mind.”