No self-respecting enterprise applications vendor these days can be without a strong artificial intelligence (AI) and machine learning story to spin to the market. But applying AI meaningfully is not as easy in a business-to-business (B2B) context as it is for a consumer-facing business (B2C).
AI thrives on huge volumes of data, which comes easy when the likes of Microsoft, Google, Facebook or Amazon want to teach computers to parse everyday human speech or recognize photos of cats. It's somewhat harder to amass enough raw data if your machines have to learn to parse specialized product descriptions or recognize behavioral patterns among buyers in specific industries. It takes patience to build up a big enough data sample, and smart algorithms to guide the machines in their learning.
Spend360 adds significant new heft to Coupa's existing investments in both these elements. From a data point of view, it has built up a huge volume of spend data over several years from more than a hundred customers spread across most industry verticals. Named customers include HSBC, Nissan, BAE Systems, Britvic and the BBC. In a press statement, Coupa says that bringing this data together with its own will create a data warehouse that has "processed or analysed more than a total of $1.3 trillion in spend — and counting."
Even more valuable is the analytics clout that Spend360's algorithms can apply to that data. Analysts at independent sourcing specialist website Spend Matters assess the company as "best-in-class" in its field:
In the spend classification area, Spend360 has leapfrogged most competitors with its recent releases in its approach to cleansing and classifying data ...
Spend360’s use of deep machine learning is now competitive with human accuracy levels (i.e. 98%).
Classification is one of the fundamental prerequisites for successful analysis of the kind of data that B2B applications deal in. Here's how Adeyemi Ajao, co-founder of HR analytics startup Identified, which cloud HCM vendor Workday had acquired as the foundation of its predictive analytics tools, explained its importance back in 2014:
From an engineering perspective, to get analytics right you have to solve three distinct problems: you have to get the data, you have to classify it, and you have to apply the right algorithms.
It’s data classification that really constitutes the key of the problem.
But whereas Identified's classification analytics were largely rules-based, where Spend360 excels is in its use of deep learning technology to automate the task of classifying business spend data. So while Identified had to manually examine recruitment profiles to create a definition tree that can classify the 600 different ways of saying 'nurse', Spend360 has developed algorithms that can train machines to automatically normalize and classify supplier definitions or other aspects of spend data.
That was the real attraction of Spend360, says Donna Wilczek, VP Strategy and Product Marketing at Coupa, who I spoke to late last week about the acquisition:
Coupa historically did not classify the data in anything more than a rules-based manner.
We spent a lot of time in 2016 evaluating a lot of the vendors in this space and finding there was no comparison to Spend360.
Their secret sauce is all of the algorithms they've written around machine learning.
The startup, which was headquartered in Guildford just outside London and also had offices in Seattle and Sydney, may have brought Coupa some unique intellectual property, says Wilczek:
We are in the process of having our legal team review their assets and we believe there could be some assets that come out of this acquisition in terms of patents.
Initially, Coupa will sell Spend360's classification and normalization capabilities to select customers as standalone services. But work is progressing in parallel to bring those services into the Coupa visualization layer so that the Spend360 capabilities become part of its unified platform. "That effort is happening now and the timing is TBD," says Wilczek, who adds:
Right now we're concentrating very heavily on the integration of the companies, and that we first focus on getting those core assets into our platform, into our Perfect Fit insights.
More efficient buying
Later on, Coupa hopes to apply the team's machine learning expertise to help identify ways in which customers can make their buying processes more efficient or more thrifty. That will draw on the vast library of data Coupa has built up as a multi-tenant SaaS vendor of all the different ways its customers and users have transacted with and configured the platform, Wilczek explains:
It's not simply the spend and transactional data that Coupa has. What we also have is that all of our customers have always been multi-tenant cloud. So we have the structured data around how they have configured their spend systems to process those transactions.
We also have the results — what results have they achieved through this configuration and their usage of the system?
The insights are not simply how are you doing in terms of spend, but now you can also start getting into performance. And if you were to adjust your configuration, what would the likely results be?
Spend360 will help us to mine that data ... If we can figure out ways the machine can do the work for the person, to improve that person's results, their efficiency, and reduce effort, those are all key considerations for us.
This direction was already highlighted by CEO Rob Bernshteyn when he spoke to diginomica's Derek Du Preez in October, shortly after the company's successful Nasdaq IPO:
We have the opportunity to do the same things that IBM is just starting to do with Watson and what Salesforce is attempting to do with Einstein.
Coupa was certainly eager to close the deal with Spend360 once it had made up its mind. Lawyers who acted for the startup have told of how they were contacted on December 15th with instructions to complete the transaction before the end of the month. By working over the Christmas period, they were able to close the deal on December 30th.
For all the advances and current hype around machine learning and AI, it's important to remember just how difficult it is to produce useful results from its application to B2B processes and data. Coupa seems to understand these constraints, and to recognize the path it must take to deliver value from the technology. This is a significant acquisition that brings unique capabilities as well as highly sought-after deep learning talent, and gives Coupa a strong competitive edge in its field.