Consumer goods giant Unilever has revealed how it is using SAP’s in-memory computing platform Hana to transform its operations.
Unilever - the €50bn global company that controls brands such as Dove, Hellmans and Lipton - is initially using Hana to increase its profit margins by accelerating its profitability analysis.
Hana combines an in-memory database with an analytics engine on a single platform. Its in-memory storage, column store and parallel processing architecture is suited to rapid analysis of large datasets from multiple sources and making predictions in real or near-real time.
This year Unilever began using Hana to carry out profitability analysis (CO-PA) for its operations in the Americas, using SAP’s Hana-based CO-PA accelerator, which is designed to speed up OLAP queries. It has just implemented a Hana-based CO-PA accelerator in Asia, for operational planning in India and plans to begin using the in-memory platform to carry out a range of reporting, analytics and forecasting tasks across all regions worldwide next year.
Thomas Benthien, director of ERP CoE for finance and ES at Unilever, revealed that in proof of concept tests Hana was able to return planning and forecasting information far more rapidly than a hard-disc based systems. Tests compared product cost and transfer pricing analysis using SAP BusinessObjects Planning and Consolidation 7.5 (BPC) software running on a disc-based system to BPC 10 running on Hana. BPC 10 on Hana was able to return results from 200 million records within 30 seconds, compared to 446 seconds for the standard BPC 7.5 system.
The difference in performance between the two platforms was less pronounced when carrying out MSO cash-up calculations using fewer records. BPC 7.5 returned results from 13 million records in 59 seconds and BPC 10 on Hana within 47 seconds.
In practice Hana is allowing the company to more accurately price its products, Benthien said. The company is using Hana to refine how it calculates the price of products, for example calculating how fluctuations in the costs of raw materials, and other factors that affect production costs, should be reflected in a product’s price. Benthien said using Hana to carry out product costing will allow the process to be carried out on a weekly rather than quarterly basis.
“Being able to do that much faster allows us to be much more responsive to the changes we see in the raw material market,” he said.
“It means we get much better insight into the actual costs and profitability of our products.
“With a € 50bn business getting your margin right and really understanding the profitability of the product better, even by just one per cent, already pays back all the investment. There’s a huge benefit,” he said.
The speed of analysis would also aid Unilever in choosing which of its many suppliers to use, he said, providing it with more timely information on which supplier would provide the best margins.
Deploying Hana-based CO-PA accelerators allowed Unilever to explore the benefits of Hana with minimal disruption, Benthien said. Unilever still stores the CO-PA data in traditional relational database systems and that data is then mirrored on the Hana accelerator platforms where it can be used to carry out rapid OLAP queries.
“The accelerators were one of the most mature ones [uses of Hana] to choose from. We’ve now shown the value. This was a way to get a foot in the door. Now we can talk about how to redevelop code and put core functionality and additional capabilites onto Hana,” he said.
The core functionality that Unilever hopes to move onto Hana includes rescheduling of sales orders, optimisation of product cost calculation and analysis of purchase orders and contracts, he said.
The company has purchased seven 64GB Hana appliances, which include the database and analytics engines running on hardware sold by SAP’s OEM partners. Unilever has loaded regional CO-PA data onto separate Hana appliances. With the heavy compression enabled by Hana’s column-store this CO-PA data is able to be reduced to about 30GB. Every transaction involving CO-PA data is written to both the traditional database system and a Hana appliance using the SAP Landscape Transform replication technology, Benthien said.
It took Unilever nine months to initially implement the Hana appliances. Benthien said that it only took five to six weeks to connect the appliances to its infrastructure with the help of SAP’s Rapid Deployment Service, and the rest was planning and getting buy-in from the organisation.
“Once you have set it up it’s very fast and the risk profile is very low because you are still writing to the classic database,” he said.
Benthien didn’t want to reveal how much Unilever paid to buy Hana appliances and licence software to run on it, saying that Unilever’s pricing wouldn’t be a fair reflection of cost as it varies based on what organisations wanted to do with Hana.
Starting next year the firm also hopes to work with SAP to begin moving its entire SAP Business Suite ERP system onto Hana.
Unilever has been working to consolidate its ERP systems since the early 1990s. Over the past 20 years it has moved from more than 200 country-based SAP ERP systems to regional systems, with the goal of having a single global ERP system by 2015.
Having a single scalable ERP system is central to Unilever’s plans to increase operational efficiency and support its ambitious plans for growth. Unilever wants to become a €80bn company by 2020. Benthien said Unilever believes it is on track to hit this target, with estimated revenues of €50bn this year.