Coupa applies the hive mind of customers to supplier risk
- 8 years of spend management data from across its customer base gives Coupa a head start in applying AI to functions such as supplier risk
Back in 2009, spend management vendor Coupa made a prescient and astute move — it began signing up customers to its benchmarking program. In exchange for making their anonymized data available to be pooled with that of other customers, they would benefit from insights into how their performance measured up against the community. For example, a company might see that it was taking much longer than average to complete a certain type of purchase, which would show a need to streamline some of its processes.
At the time it was still three years before Google researchers would demonstrate deep learning technology that could work out for itself how to recognise a cat, triggering massive investments by the likes of Google, Amazon, Microsoft and many others that are now fueling rapid advances in AI. But that early decision has put Coupa in pole position to be able to start delivering practical AI insights to its customers today. Just one of its 525 customers opted not to take part in the benchmarking program — a single business whose transactions amount to little more than a rounding error in the $360 billion of spend that customers now have under management with Coupa.
All that customer data — along with third-party data added in the course of recent acquisitions — is now available to inform what Coupa is calling "community intelligence." And as any AI expert will tell you, it's the scale of data you have access to that truly determines the success of any machine learning project.
Applying AI to supplier risk
While most enterprise application vendors talk up AI as an enabler for the as-yet-unproven benefits of being able to converse with their applications instead of having to tap in commands, Coupa CEO Rob Bernshteyn was able to stand on stage at the vendor's annual conference yesterday and demonstrate AI applied to a real-world business problem — functionality that's available to its customers now under early release.
Called Risk Aware, it is a supplier risk management capability that evaluates a variety of metrics and then flags up risk factors if there are problems at a supplier. Some of the metrics come from third party sources such as financial data, court filings and news sentiment, collated by newly acquired data aggregation platform Riskopy. Other metrics come from transactional information collected from across the Coupa customer base, flagging up warning signs such as above-average disputes and overages. All of this is analyzed and presented using tools developed within Coupa and at its recent data science acquisition Spend360.
The results are displayed within the context of live business transactions, putting supplier risk data where it's needed, at the time buyers are taking decisions on placing new orders or renewing contracts. That's an important differentiation from conventional, standalone supplier risk tools. The software also brings together additional information, such as the total value of spend with the vendor and how significant it is in specific spend categories, and is also able to automatically suggest potential next steps.
The new capability is the first fruit of Coupa's acquisition late last year of Spend360, whose existing store of some $1 trillion in spend data has been merged into Coupa's own data store. An important part of that process was aligning the Spend360 data with Coupa's own data structure, beginning with supplier data where Spend360 has particular strengths in using machine learning to normalize supplier records. That made sense, says Bernshteyn, because of the immediate value that could be realized from collating historic transaction data for the three million suppliers in Coupa's network.
We just said, okay, let's start with one entity. We could have started with products. We could have started with service. We could have started with price. But we said, 'Who are people spending money with? Let's understand who they are first. Let's normalize that, enrich that, classify it, structure it, put it into central store, and then start thinking about how to share insights that hang off of that entity.'
The Riskopy acquisition then added the ability to collate information from external data sources, rounding out the risk analysis.
More killer use cases
This is just the first of many ways Coupa can apply machine learning technology in the procurement space, says Bernshteyn:
It's a use case of applying AI onto a big data set that I think is going to work. People have been interested in supplier risk for a long time.
We've got a lot of ideas on what we can do with community intelligence that apply to other data sets like items, and services, and price points, and delivery times, and all kinds of things. So I think this whole vision of making companies smarter about the way they spend money, with this community intelligence concept, I think we're in a good spot ...
Certainly, we'll see where it goes. That's why I'm really excited to leverage this community's intellectual horsepower so they can give us ideas on other things we could do with this. It's such an asset that's growing for us, it's a shame if we can't figure out those killer use cases.
Those "killer use cases" could prove hugely valuable, as Bernshteyn told attendees in his keynote yesterday:
We don't have everything figured out but we know we're sitting on gold. We're sitting on trillions and trillions of dollars of spend data, growing in spend volume, and if we work together we can all get smarter and smarter and we can have competitive advantage against any competitors in our given industry because we know how to help our company spend smarter. That's the vision we're pursuing.
This is a classic example of how a SaaS vendor can use its pooled history of customer transactions to add value, but Coupa has played its hand particularly well. I'll have more from my conversation with Bernshteyn as we discuss Coupa's expanding footprint in enterprise procurement.