“With Databricks and Fivetran, we will be able to significantly improve marketing insights in the future. From a technical standpoint, the two tools interact harmoniously together and the integration feels very native.”
Jan-Niklas Mühlenbrock, Team Lead, Business Intelligence & ERP
- One single source of truth provides a holistic view of all marketing channels to optimize digital advertising spend and to enable all decisions to be data driven
- Data democratisation gives everybody in the company access to reports to make better strategic business decisions
- Fivetran saves cost equal to one full-time engineer’s salary (60,000 to 80,000 € per year)
- As a first-party Microsoft provider, Databricks was the perfect fit for Paul Hewitt’s existing Microsoft infrastructure
- The Modern Data Stack empowers Paul Hewitt to shift business from 50% traditional retail to 100% eCommerce during the pandemic
- Future-proof data architecture sets Paul Hewitt up for machine learning
Cloud Platform: Microsoft Azure
Business Intelligence Tool: Power BI
Outgrowing old practices
Early in its online journey, Paul Hewitt used Supermetrics for analysing advertising spend – but the business soon out grew its capabilities. The analytics team, three employees at the time, were manually entering data from Supermetrics into spreadsheets to determine which channels were delivering the best return, in a time consuming and error-prone process.
“It took us about a week and a half to update and consolidate all the different data sources into one dashboard,” said Jan-Niklas Mühlenbrock, Team Lead, Business Intelligence & ERP. “We had exceeded the capabilities of Supermetrics for the amount of data we wanted to obtain.”
To meet the needs of an increasingly complex supply chain, the company had invested in an ERP system, Microsoft Dynamics NAV, and began to make data available for analysis with Microsoft Power BI. The business decided to take its data strategy to the next level with a cloud data platform where data from across the business could be integrated into one place. The objective was to transform into a data-driven business.
Combining the best tools
Databricks, a first-party Microsoft provider, was a natural choice for the data analytics platform as Paul Hewitt already had Microsoft Azure cloud platform in place. Another advantage was the way Databricks eliminated data silos, enabling holistic analytics and data science as well as future machine learning use cases.
“Databricks is a one-stop shop that enables data engineers and analysts to work collaboratively, which was a really big deal for us,” explained Mühlenbrock. “We have a small team and wanted something that would be easy for everyone to use. Databricks, as a destination, helped accelerate our goal to democratise data.”
With the combination of Azure and Delta Lake by Databricks providing the data lake, Power BI the visualisation tool, the final piece to complete the Modern Data Stack was automating the pipeline from the data sources. The team had manually built connectors as a proof-of-concept – a time-consuming process plagued by broken connections, API throttling and irregular updates. It convinced them of the importance of searching for a preconfigured solution.
Several ELT providers were considered, including Funnel, but none of them compared to Fivetran in terms of functionality and ease of use. During the evaluation discussions, the case for Fivetran only grew stronger. “The Fivetran team were very patient with us, answering every question in a transparent way that was a big help in our decision-making process,” said Mühlenbrock. “And they make onboarding as easy as possible for the customer.”
With all the architecture components in place, it was relatively straightforward to integrate heterogeneous data from a wide variety of channels into the data lake. Fivetran was set up immediately, automatically loading the data from multiple advertising channel APIs to better inform the marketing spend.
Jan-Niklas Mühlenbrock is convinced by the new architecture. Paul Hewitt’s single source of truth now accelerates the path to better reporting and enables dashboards that surface “a treasure trove of data'' to more employees. Fivetran connectors took over the heavy lifting for the marketing analytics team, enabling a holistic overview of their channels for the first time, including Google Analytics, Facebook Ads, Pinterest, Microsoft Ads and TikTok. Paul Hewitt had attempted a holistic view prior to implementing Fivetran, but abandoned it because it was too complex.
“With Databricks and Fivetran, we will be able to significantly improve marketing insights in the future,” said Mühlenbrock. “From a technical standpoint, the two tools interact harmoniously together and the integration feels very native. We had to do very little maintenance or customisation; it simply works and it’s very reliable.”
The company has calculated that replicating Fivetran’s level of service, maintenance and speed of bug fixes would require a full-time data engineer, costing somewhere between €60-80,000 per year. The time it now takes to turn insights into actions is a lot shorter and the benefits are very tangible.
“As Team Lead, it's been nice for me personally to no longer have to explain to the marketing department over and over again why data is not up to date or accurate,” said Mühlenbrock. “It’s been a huge win because it leaves more time for me and the team to focus on work that adds more value to the business.”
The data set-up Paul Hewitt has in place now also enabled them to quickly shift their business from around 50% traditional retail to 100% eCommerce when the pandemic hit. Long-term business benefits will be about fulfilling the company’s goal to become data driven, democratising data by enabling non-technical people to access insights that inform better decision making. Jan-Niklas Mühlenbrock is confident he has the architecture to make it happen.
“Databricks and Fivetran are the best tools on the market in my opinion. Even if we don't use the full capability right now, we are setting out for a future where we will be able to perform machine learning and explore different use cases,” he said. “We are just at the beginning, but I have no doubt that we will continue to yield insights that will have a very positive impact on our business.”