French Tech Leader Cegid Generates €15M Additional Volume Annually with AI-Driven Decisions

We use AI to predict the probability of invoice payment. Increasing the number of deals by 20% amounts to approximately €15 million in additional annual volume.

The Need: More Models in Less Time with the Same Team

At fast-growing Cegid, the predictive analytics team must meet ever-expanding demand fueled by frequent acquisitions.

The French company offers cloud services and software solutions for accountants, small businesses, retail, and corporate clients serving 350,000 businesses across 150 countries. 

Cegid’s analytics team takes on a growing list of business challenges, including predicting the likelihood to get paid on invoices and the propensity of customers to add services. As with most companies, the greatest analytics challenge is creating more models in less time and minimizing the technical skills and resources required to do so.

“We wanted to reduce the need to code while also accelerating model development,” said Joao Henriques, Head of Credit Risk and Data, Cegid.

A One-Stop AI Solution

When he joined Cegid, Henriques saw the opportunity to bring in DataRobot. From his experience at a previous company, he knew it offered the one-stop capabilities the team needed to automate the analytics lifecycle end to end.

We wanted an all-in-one platform. DataRobot is a no-code solution that covers the full spectrum of developing, deploying, and managing models. I’m convinced this is the best tool in the market for us.

The company turned to Portugal-based Passio Consulting for support in deploying the platform and applying it to specific business decisions. Passio brings business intelligence, artificial intelligence, data warehousing, and data virtualization solutions to help its clients transform data into a competitive advantage.

At Cegid, the solution integrates with Amazon Web Services and the company’s data lake using application programming interfaces (APIs). Cegid then analyzes insights in Microsoft PowerBI and Excel.

Faster Decision-Making While Managing Risk

Cegid applies the platform primarily in its invoice factoring business. To help clients with cash flow, Cegid extends financing to them based on open invoices—rather than clients waiting 30 or more days for end customers to pay. For their service, Cegid keeps a percentage of invoice totals.

The company has modernized factoring by turning it into a one-click process for clients. Using the AI platform, Cegid gains the necessary end-customer details to decide whether to pay or deny the selected invoices.

With a Payment Prediction Model (PPM), they evaluate the probability of payment for each invoice. Analytics also help them set optimal interest rates. Those insights help them manage risk in their decision-making.

“We use machine-learning models to support decision-making on all individual deals that we are making in terms of our factoring business,” Henriques said.

They’re also exploring models to improve collections on those invoices, as well as others to assess the likelihood of Cegid clients to convert to factoring clients.

20% More Deals in One Year = €15 Million

With the model developed using DataRobot, Cegid increased the approval rate on factoring deals by 20% without increasing risk—driving significantly more volume in funded invoices last year.  

“We’re using AI predictions on payment probability for every decision on invoices,” Henriques said. “Increasing the number of deals by 20% amounts to approximately €15 million in additional annual volume.”

DataRobot was the perfect match for a small team looking to put more models into production. Cegid makes decisions, such as setting interest rates, with much greater accuracy and has already seen a payback on their investment.

Deploying Models in 50% Less Time

Using DataRobot automated machine learning capabilities, Cegid expedites model experimentation and development, allowing them to test scenarios quickly. Data scientists traded coding for time to refine their data and evaluate business problems.

Though some data scientists were initially skeptical about the platform, Henriques finds they have changed their opinions after seeing the productivity of AutoML firsthand.

We accelerated model development—cutting the time to deploy models in half. We can now use our resources on analytical studies and business intelligence.

Cegid looks forward to applying the platform to more business challenges in its factoring division and beyond and knows it can rely on the expertise at DataRobot to help maximize its results.

“When we have technical questions, we get a rapid response from our account manager,” Henriques said. “We have only positive things to say about DataRobot support.”