A faster pipeline of data

How automation improved contract analysis and accuracy

Summary: Flowserve Corp., a global flow control system manufacturer and service provider, wanted to save time, cost, variations and errors in its contract reviews. Grant Thornton helped Flowserve plan, develop and implement a machine learning and optical character recognition solution that can process more than 90% of contracts 6x faster, with more accurate results.

Problem

With more than 300 locations around the world and almost $4B in revenue, Flowserve Corp. is a global provider and servicer for pumps, seals, valves and other fluid motion products. The company has 180 customer service centers that provide parts and services, and it contracts with a wide range of global customers.

To make sure that all of its sales contracts were compliant with the new ASC 606 Revenue Recognition requirements, Flowserve employees manually reviewed the relevant details from contracts across many document formats – including some that were up to 100 pages long. Then, employees determined the performance obligations and revenue method to apply to those contracts. These manual reviews involved different employees and standards worldwide, were time-consuming, inefficient and introduced a risk of human error. Even contracts from the same customer might follow a variety of formats. Lastly, the manual reviews left contracts trapped in paper files where leaders couldn’t access or analyze data across the range of agreements.

Flowserve wanted an automated solution to identify and classify contracts, quickly extracting the customer, offering, promises, price, terms, conditions and other details from their paper contracts. The company also needed to identify relevant contract data element exceptions and to store the details in a system that could fuel compliance reporting and business analysis. In all of this, the company needed to ensure that the contractual data remained compliant with security requirements.

Solution

The Grant Thornton team assessed options and developed rules to outline the company’s requirements. Based on these, the team determined that Ephesoft patented machine learning technology and optical character recognition (OCR) offered market-leading capabilities that would best meet the client’s needs.

“When we started this discussion, we were looking at multiple solutions,” said Grant Thornton Transformation Advisory Managing Director David Dominguez. Flowserve wanted a cloud-based modern technology process with advanced capabilities to automate data extraction, indexing, classification and validation while identifying exceptions for relevant contract data elements. The company was impressed with the machine learning, optical character recognition and intelligent character recognition functions that Ephesoft offered. “We knew the capabilities of the Ephesoft tool – it’s a proven solution,” Dominguez said. “So, we demonstrated the product and the company liked what they saw.”

The Grant Thornton team began a proof of concept (POC) project that used Ephesoft technology to scan paper documents and automate contract processing for the accounting and finance team. Although the scope was limited, it still required the solution to extract 54 fields of data from contracts spanning multiple nations and currencies. To achieve success, the project needed to prove that it was a feasible solution to achieve both higher productivity and fewer errors.

During the POC, the team found that contracts used a wide range of different terms and conventions to identify similar information. The team refined and expanded the list of terms that triggered automated tasks, creating more complete lists that automatically handled a wider range of term variations. “It allowed us to gain additional clarity on requirements,” said Grant Thornton Transformation Advisory Manager Will Morse. “It let us pivot and proved out some things that made the implementation so successful in the end.”

The team also designed a stopping point into the solution. Flowserve wanted to be able to stop the automatic processing so that employees could selectively review extracted data before it was entered into the cloud-based revenue recognition standard. “They wanted to have everything stop so that they could still do checks, but without having to hand-key the data,” Morse explained.

Once the team refined the POC solution, it was ready to expand. “We used the pilot as a springboard into the full implementation,” Morse said. The solution’s capabilities quickly scaled up to serve more business areas and process more data. “One of the technical specialists said he was impressed by the volume of data the solution was processing,” Morse recalled. Flowserve was also impressed by the solution’s adaptability to updates and modifications over time.

Benefits

Now, Flowserve can complete contract processing as much as 6x faster. “For a complex contract with multiple change orders, it took an average of 4–6 hours for a reviewer to make a revenue recognition determination,” Morse said. “Now, they’re down to about an hour.”

Flowserve processes about 10,000 contracts per year, containing more than 500,000 pages of information. The new solution can now process more than 90% of those contracts automatically. By eliminating manual data entry, the solution decreased errors by almost 90%. It also automatically extracts key terms and conditions so that processors can identify which ones might be in other contracts and ensure standards.

The Grant Thornton team planned the project so that Flowserve could maintain it going forward, but Flowserve has also found it helpful to call on Grant Thornton support. “They’ve had some people change roles in this practice,” Morse said. “They want to re-engage us to help get new people up to speed.” Meanwhile, Flowserve has also continued to expand the solution’s scope.

As the Flowserve implementation has expanded, more employees have been impressed with the results. “They’ve noted the flexibility of the system – how it can capture almost anything,” Morse said. “The full implementation was designed to initially address the top 70 high-volume customers. Now, they have a new top-70 list of customers, and they are working to adapt to that additional set.”