Inside Infor’s Data Science Lab
August 17, 2016 Alex Woodie
If you’ve noticed subtle improvements to your enterprise software lately, you’re not alone. For the past two years, IBM i ERP giant Infor has been using data science techniques to optimize various business processes that are automated by its software. And the best part may be that Infor is offering its Dynamic Science Labs expertise to customers for free and that improvements get wrapped back into the product.
The big data revolution is impacting many aspects of our lives, including how we work and play. The technologies and techniques that the big Silicon Valley Web firms like Yahoo, Google, LinkedIn, and Facebook developed to capture and track user behavior for things like Web search results and social media optimization are increasingly finding use in other places, including software product development. So it’s no wonder that a forward-looking software company like Infor–which has its eyes set on disrupting the ERP duopoly that Oracle and SAP currently enjoy–would seek to leverage this potential difference-maker.
About two years ago, Infor co-president Duncan Angove spearheaded the creation of the Infor Dynamic Science Labs (DSL). The Cambridge, Massachusetts-based lab currently has a staff of about 20, including numerous data scientists who are experts at turning large volumes of data into actionable insight.
Ziad Nejmeldeen, the chief scientist of Infor Dynamic Science Labs, says there are two main areas where Infor seeks to leverage the power of big data and data science, including forecasting and recommendations.
“The recommendations we’re making [include answering questions like] what inventory should you maintain? What price should you offer? Or on the CRM side, what product should you be producing and which customers should you call?” Nejmeldeen tells IT Jungle. “All these questions that can be asked, they’re answered by doing something with the data that already resides in the ERP.”
Up to this point, the Infor lab has worked with just a handful of customers, including users of Lawson S3 and M3, Infor CRM, Infor LN, Enterprise Asset Management (EAM), Infor SX.e, Infor WFM, and Infor A+. Among those, M3 and A+ run on the IBM i operating system (although the Java-based M3 also runs on other systems). The plan calls for eventually touching all major Infor products in every supported industry, which would include a much larger number of IBM i shops.
The first client was a hospital looking to optimize inventory that was managed with a Lawson ERP system. By analyzing the flow of goods in and out of its warehouses and building a model that replicated the movement, Infor was able to create an inventory optimization solution that reduced inventory by 15 percent while maintaining 99.9 percent product availability, Nejmeldeen says.
The idea is to eliminate the need to have humans periodically compile data and manually analyze inventory needs, which is largely how companies do it today. “We want to build software that’s going to do it for you,” Nejmeldeen says. “It’s going to be on autopilot, you won’t need human intervention, and once it’s running, you’ll be able to get updated answers as often as you like.”
Another DSL engagement involved building lead scoring, next likely purchase, and pricing models for an Infor CRM customer. Many marketing organizations today hand-code their own customer matrices using a tool like Excel, but the matrices can only slice the customer data in so many ways, which makes them of limited use. Infor sees an opportunity to take the marketing automatization to another level with the power of data science.
“What winds up happening is that [the matrix] is provided to sales reps who are the front line for talking to customers,” Nejmeldeen says. “But those sales reps have power to override the prices. So the matrix becomes simply a guideline with a lot of overrides at the very tail end of it.”
Infor’s data scientists can build a big data solution that uses the entire purchase history of the company as the baseline for automated recommendations. The recommendations will almost certainly raise profit margins, but what’s equally important is explaining how Infor got the answer. Without the proof, the sales reps are likely to go back to their old ways of overrides, eroding profitability.
Infor doesn’t charge for DSL services. “We do this completely for free,” Nejmeldeen says. “We get paid with our feedback. We go in and do a POC and deliver a set of results, and at the end of the POC, they get immediate benefit, and we get the feedback. We get to walk away with a solid design. We take our development team and sell it.”
In some cases, the DSL engagements will yield a bolt-on solution that Infor can sell to existing customers, while in other cases, Infor will use the insights to improve existing products.
In the future, the DSL group plans to tackle big data challenges in other industries, including vehicle routing in the distribution business, boosting margin for manufacturing customers by optimizing the order of things manufactured, and improving inventory and markdown optimization for retailers.
The availability of ERP-resident data from more than 70,000 customers gives the Infor DSL an edge that highly specialized boutique firms can’t fathom. “We’re not siloed by industry,” Nejmeldeen says. “If you look at DSL, every member of the group can work on any one of these projects or any industry and can switch over when the time is right.”
Neither is the DSL siloed by operating system, which is good news for Infor’s 14,000 IBM i customers. Nejmeldeen says his group has been approached by a number of System 21 shops in the fashion industry that are interested in manufacturing and warehouse optimization, as well as by EAM customers who are looking for ways to optimize spare parts and predict asset failures. Those projects are a bit further down on DSL’s “to-do” list.
Most of the DSL’s engagements have used internal data that’s stored in the ERP and the associated database. However, there’s a raft of other data sources that the DSL could tap into in the future, including social media data and senor data from the Internet of Things (IoT). “RFID is making a huge comeback,” Nejmeldeen says.
Getting the DSL ramped up has been a challenge, particularly in regard to attracting data scientists, who are so tough to find that they’re often dubbed “unicorns.” “Finding the people we need is not easy,” Nejmeldeen admits.