Consumer goods maker Unilever is halfway through building a global enterprise data warehouse, a project which has been locked into the critical paths of high-profile business programmes from its inception.
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
Since planning the EDW project in 2010, Unilever has eschewed the more painless route of divorcing its data warehouse from major business projects until ready.
Greg Swimer, vice-president for IT, business intelligence, says: “One of the easy things to do would be to take the EDW away from the critical path of these business programs and say we’re building it in the background and, when we’re ready, we’ll launch some high profile things on it.
“We went for a strategy that took us onto to the critical path of delivery for some big and important programmes in our business. Six months after mobilising the EDW build with our first big release to several hundred users in North America, supply chain analytics went along side an ERP go-live. It forced us up the learning curve of release, test and progression very early on. It was a very healthy dynamic for the EDW programme.”
Read more on enterprise data warehousing
Learn about the analytical advantages of enterprise data warehousing
Objectives of Unilever's data warehouse project
Unilever started implementation in May 2011. It is set to replace around 40 data warehouses and data marts the company has built over the past 10 years.
The data warehouse was designed to support two objectives.
Firstly, Swimer and his team wanted to “unleash the power of information” by exposing analytics consumers to as large a data set as possible.
“This is the big data story,” Swimer says.
“It is about giving the best real-time analytics to the business, handling the data explosion and enabling tens of thousands of business users to mine data quickly and find insight.”
The second objective was to give more mainstream business users access to standard reporting from anywhere in the world within three clicks of a mouse.
To achieve both these objectives simultaneously, Unilever decided it needed a completely unified data warehouse infrastructure and strategy.
“In global companies like Unilever there can be as many forces pulling away from unification as there are pushing towards it,” Swimer says.
“We concluded that if we did not have an infrastructure for getting the same data to wherever it needed to be in real time, we would end up building a new system every time, to a new functions specification, a new server build, a new interface and new data marts. We would never be able to keep up with demand.”
Layered data warehouse approach
The system was built on three layers.
Firstly, the BI team built a document repository to store standard company reports. The user interface was based on Microsoft SharePoint.
The second layer was a global business intelligence suit supporting multiple applications including sales reporting, sales analytics, supply chain analytics and procurement analytics. This was built on the Microsoft Business Intelligence toolset and Tableau’s data visualisation software.
Finally, beneath these two layers sits the EDW itself, with Teradata at its heart and SAP technology for master data management and extract transform and load functions, Swimer says.
Enterprise data warehouse governance built-in
The technology project is supported by a central IT team, as well as an information services team, managing governance and master data. However, regional teams and business functions, such as supply chain, each have individuals responsible for information services. This way, the governance for the EDW is embedded into the business.
According to Swimer, moving to an EDW in one project requires a careful balance of business priorities and technical requirements. Business teams who demand application capabilities did not want to be first or last to receive applications and were concerned about their existing dependencies. IT teams must avoid the whole EDW being defined by the first project and must build in capacity to create security and availability in the system, which do not belong to any single project.
To get around these dependencies, Unilever created a delivery model where any given project – sales analytics for example – is run by a distinct team who own it from beginning to end. These are business-funded teams, with very strong business governance, Swimer says.
But each project team is bound into a common technical framework with seven technical teams governing front end, data warehouse, enterprise ETL, testing and release. Those “towers” own the common standards, they govern the release timetable and they govern the infrastructure.
Unilever's progress so far
So far, the project to simplify and speed up access to reporting has been “wildly successful”, Swimer says. Launched in the third quarter of 2011, it has gone live in 45 operating units, with 6,000 commonly used reports. Eighty-five per cent of business users now say they have improved access to reporting and information.
Meanwhile, the analytics side of the EDW is also claiming successes.
Ten of the planned 17 capabilities are now live, and it is on track to finish on time during 2014. One capability, which harvests retail point of sale data, has lowered costs and also improved business performance, Swimer says.
“Our own customer teams are coming back to us and saying that with the insight they have got through these tools they have been able to drive revenue with retailers,” he says.
But the greatest benefit comes from increased agility in BI and analytics, as the EDW is able to support new use cases as if it were a utility, without needing new hardware or software.
To take advantage of its growing capability, Unilever has launched a user training programme which it says will help it become an analytics power house. It expects 20,000 users to be benefiting from the EDW once it is complete in 2014.