Tamr first brought its data unification technology to the market as a general purpose solution to help companies in their quest to become truly data- and analytics-driven enterprises, providing a next-generation means for them to clean and connect disparate data in an automated and scalable way. When DATAVERSITY® spoke to Tamr co-founder Andy Palmer in late 2014 for an article on data curation, he discussed how using Machine Learning and semantic triple stores to address the enterprise data unification issue offered a great opportunity for businesses to gain 360-degree views of suppliers, customers, products, or whatever their needs might be to inform analytics and address hard business questions.
At the time, he pointed to one unnamed enterprise that was putting the technology to work to optimize spending, making sure to get the best price for all products it buys across the entire company. Now, spend analytics for the cause of strategic sourcing is a primary use case that Tamr has settled on as giving businesses, large and small, the biggest opportunities for success from the holistic Data Management enabled by its technology. “Data is in the forefront for our customers like GE, Toyota Motors Europe, GSK and Thomson Reuters,” says Nidhi Aggarwal, Global Lead of Strategy and Marketing at Tamr. “They understand the importance of data preparation and how that lets them become data-driven.”
More companies of every size need to come to the understanding that they’re leaving a lot of money on the table by not undertaking sourcing activities in a systematic way and better managing spend for both indirect materials (pens, papers, marketing) and manufacturing items in the broadest sense of the term. For a heavy industrial company like GE, metal is such an item; for a bio-pharma like GSK, it’s the chemicals that go into new drugs; for an information purveyor like Thomson Reuters, it’s data that drives information-infused products and services.
While everyone’s excited about the potential of Big Data to help them increase revenue, Aggarwal says, less recognized by most companies is the fact that “a dollar saved in terms of the cost of [manufactured] goods sold contributes directly to the bottom line.” In comparison, when companies make an extra dollar in revenue, only about 20 cents goes to the bottom line, while 80 cents goes to cover producing those goods.
“ROI tends to be very high if you do spend management effectively, if you are sourcing the right way,” she explains. At the same time, risks of inappropriate sourcing can be minimized – such as acquiring the same vital part for manufactured goods from a handful of suppliers all located in the same geographic location, which means that an entire supply chain can come to a grinding halt should that region experience political or natural calamities.
The Data Behind Strategic Sourcing
Today, many larger companies do engage in strategic sourcing to some degree, going beyond who’s going to give them the product they need at the cheapest price, but even then it’s a limited practice. Many bigger companies have only about half their spend under management, and typically solely at business unit levels. That’s true even when the same item is used in multiple parts of the enterprise, which means the business may be missing opportunities to increase volume savings.
Additionally, while big businesses likely will review $10 and $100 million supplier contracts with a fine-tooth comb, “the majority of their spend is below the $500,000 or $200,000 level and that’s where the real opportunity to save money and reduce risk lies,” Aggarwal contends. 75 to 80 percent of savings opportunities lies in data that’s not being analyzed, she says.
Strategic sourcing should be based on accomplishing spend analytics using all available sourcing data to inform a holistic view of spend – and not just for individual business units or departments, but company-wide. What is the company buying? Where is that product used across the company’s operations? What are the alternatives to it? What are the different prices? What is the financial profile of the supplier or suppliers being used? What is their on-time ratio? What locations are they in? What political or natural events are those areas subject to? And in a world of fast-paced mergers and acquisitions, have they just been bought by another supplier in the field, which could have an impact on competitive pricing? Such questions are only a select few that inform the necessity of strategic sourcing and spend analytics.
In sum, enterprises need to look at data using a full range of attributes at a wide breadth and deep level of granularity, and for sub-category items as much as for big-ticket buys, Aggarwal says. “The $10,000 contract should be looked at as closely as the $100 million one,” she says.
But to complicate matters, not all the attributes that need to be taken into consideration are going to be found within internal company data. Third-party data is going to need to come into play, as well. It’s unlikely, for instance, that any enterprise is tracking their suppliers’ M&A movements, or looking at how often certain geography is affected by extreme weather. But such information needs to be brought into the mix from outside sources to drive smart sourcing and not be placed at risk or disadvantage.
Change the Dynamics
It’s not that there isn’t a desire improve spend analytics and increase strategic sourcing prowess. The problem is that the fundamental practice on which it’s made possible – cleaning and connecting data from a variety of venues in a variety of formats – has been such a manual and tedious process. It takes so long and requires so much labor that it simply becomes unaffordable for most companies, especially on an all-encompassing scale.
Tamr is taking on the problem, automating the process with Machine Learning and helping it along with expert sourcing knowledge. Its Machine Learning algorithms clean and classify about 60% of data from multiple sources. Then, sourcing and procurement managers at user organizations weigh in; they are asked a few targeted questions to help train the Machine Learning model further so that 95% successful classification can be achieved, according to the company.
It’s an approach that makes it possible to continue to bring in and rapidly classify more data sources from around the company and outside it, all in the service of driving more effective spend analytics.
“The experts give valuable information to the Machine Learning model, and when it sees other data like that in the future it can make classification recommendations with a high degree of confidence,” she says. “This lets us provide high accuracy and scale to more data sets.”
Taking this approach, Aggarwal says, fulfills the true promise of Big Data Analytics – that is, that every single spend decision should be optimized, thanks to lowering the cost and speeding the process of cleaning, integrating, and classifying the data that informs those decisions. “The future as I see it is that every spend decision is driven by more data – holistic data – with all spend under management and people making effective decisions and saving 1 or 2% on each transaction,” she says, which can add up to their saving 20 or 30% more than they are today on sourcing.
“Saving money on every transaction in the longer term makes strategic sourcing really valuable,” Aggarwal says. “Every time you make a decision it will be a better decision because it’s driven by more data, and you’re lowering your risks by bringing in third-party data. That eventually saves millions of dollars for an organization on an ongoing basis.”
