Moving Expenses to Procurement with Procurable Insights

Originally Published October 22, 2021 – Updated April 19, 2023.

At Coupa, one of our three core values is “Ensuring Customer Success.” This means we need to go above and beyond to find value for our customers, even when they aren’t looking for it. Powered by Community.ai, Procurable Insights is a key capability within Coupa’s expense management solution that delivers this kind of value. The solution uses a state-of-the-art and patent-pending artificial intelligence (AI) algorithm to identify expenses that could have been procured for additional savings.

For those new to the terminology, expenses are submitted after the item or service has already been purchased. This is often called “post-approved” spend. In contrast, a procured item or service goes through a procurement process where suppliers are identified, negotiated with, and possibly even subjected to an RFP. This is often called “pre-approved” spend.

Our analysis shows that 10–15% of our customers’ post-approved spend are items or services that they or other members of the community have procured in the past. Identifying this spend helps customers drive future savings by encouraging users to submit orders instead of expenses or get the item at pre-negotiated and contracted prices.

How we solved the problem

Coupa’s unique solution to this problem is typical of the AI work we do. Identifying procurable expenses is not something currently being done by other firms. The solution requires us to combine advanced AI algorithms along with our more than $4 trillion of community spend data in order to identify patterns across many companies.

Since the problem is new, we had to start with first principles and imagine how a layperson would frame the challenge. For example, someone might think: “If we have a historical invoice for this item that an employee has just expensed, then it is likely procurable. However, there are some expense categories (like meals, taxis, etc.) that could not be realistically procured, even if we have some past invoices for them.”

We then had to translate this logic into a process that a machine can handle:

1. Assign every expense category using a classification approach (airfare, meals, lodging, etc.).
2. Select categories that would be realistically procurable (office supplies, gifts, etc.).
3. Compare the expense description to all historical invoice descriptions. To make it more accurate, use a semantic similarity measure. (For more discussion on semantic similarity, see our post on duplicate invoices where we talk of creating procurement-related natural language processing.)
4. If we find an expense description with high similarity in the same customer’s invoices, then it’s “procurable by you.” If we find an expense description with high similarity in invoices with the rest of the community, then it’s “procurable by community.”

A technical challenge we faced was how to compare the extremely high volumes of expense lines against invoice lines that we’ve processed for our customers. To address this, we used a combination of clustering and sampling to reduce the number of invoice lines for comparison by 90%. This allowed us to find a practical and efficient solution.

The flow chart below outlines how our algorithm works. You can see that the approach stitches together many different steps to solve this unique problem.

We’re proud to announce we have a patent pending for this work. We hope that this innovation will help organizations everywhere spend smarter together.