A few years ago, the U.S. housing bubble burst and the construction industry tanked, which was challenging for every business in that sector. Valley Forge, Pennsylvania-based building-products manufacturer CertainTeed Corporation found itself at a crossroads. A subsidiary of Saint-Gobain, a global manufacturer with around 200,000 employees worldwide, the company faced a crucial decision: How could it continue selling in a distressed market while ensuring that its receivables would continue to be paid?
Treasury & Risk spoke with Susan Delloiacono, the company’s director of credit services, about how CertainTeed began using predictive analytics within its credit-and-collections automation system to identify exactly which customers pose the most credit risk. By basing decisions on more precise information, the company was able to streamline credit decisions and dramatically improve prioritization of receivables within the collections process. Now its accounts receivable (A/R) metrics are outstanding, and employees, customers, and the company are all thrilled with the change.
T&R: What do the treasury and receivables functions look like at CertainTeed?
Susan Delloiacono: Well, the treasury function is managed by Saint-Gobain, but we handle our own credit and collections in the business. In fact, when I was hired in 2010, our five major product lines—roofing, insulation, vinyl siding, gypsum, and ceiling products—each had its own ERP [enterprise resource planning] system and its own credit-and-collections manager.
T&R: That seems inefficient.
SD: It was. We might have a customer that purchased products from each business unit, and they would have five credit managers talking to them about their credit, asking for their financial statements and banking references. Each group was doing the credit-line reviews and collecting for their own invoices. And the credit managers had different criteria for making credit decisions.
Right after I started, one customer filed for bankruptcy. That particular customer was buying from three business groups. I found that that one credit manager was managing the account on a two-for-one basis—so you pay me two dollars, and I’ll ship you one dollar worth of product. Another credit manager was doing one-for-one, so an exchange of dollars for orders. And the third credit manager didn’t even know there was a problem. They sat next to one another, but they never worked together to establish a combined risk strategy that would represent the companywide credit limit that CertainTeed would feel comfortable offering to that customer.
We were just looking at receivables as if we were five separate companies, when the risk rolled up to the corporate level. Job one for me when I started was to combine all our systems so that CertainTeed’s A/R had one face to the customer.
T&R: Did you hear from customers about the hassles this environment created for them?
SD: No, but just because there’s not a complaint doesn’t mean there’s not discomfort. If I worked with a vendor that required me to deal with five different people and provide the same information five times, I think that would be aggravating. And we were very inconsistent in how we approached our customers.
T&R: What about internally? Your visibility into receivables data must have been limited.
SD: When I first got here, the CFO asked me, ‘How much does our largest customer owe us?’ I had to tell him I’d get back to him in half an hour, because I had to find all the account numbers and add everything up and figure out the past-due percentage. That was really problematic in the building-products industry in 2010. Construction took a huge hit in the economic downturn, and it was absolutely critical to CertainTeed to have a consolidated view of each customer’s credit risk.
T&R: Where did you start?
SD: We started by implementing a new software system, AvantGard GETPAID from SunGard. We created a risk score that indicates each customer’s overall credit and collection risk to CertainTeed. The score includes whether they have a bank line of credit, which is critical for us. It includes a few Dun & Bradstreet metrics that reflect how they’re paying other vendors. And then the biggest element of the scorecard is how they pay CertainTeed—not how they are paying their electric bill, but how they are paying us. I’ve found that the Dun & Bradstreet information, on its own, can be woefully deficient. I had one customer that was in a technical bankruptcy, but their Dun & Bradstreet PAYDEX was still 80.
So we developed a risk scoring mechanism within GETPAID that uses SunGard’s predictive analytics models to determine how much credit risk a customer really poses. The model incorporates details about a customer’s payment behavior that a human credit analyst might miss.
One example is the customer I mentioned that was in technical bankruptcy. None of their invoices were past-due until January, so typical collections software wouldn’t have alerted us to the potential problem until then. But the scoring model picked up on the fact that even though invoices weren’t past-due, the customer’s payment behavior was changing. Eighteen months prior, this customer would always discount in five days. That time frame started growing, first to 10 days, then 15 days. Their invoices were net 30 days, and they were still paying on time, so even though there was a clear change, it wasn’t one that I would necessarily notice as a collector. But the predictive metrics picked up on it. This customer’s risk score changed in our system in September, four months before their first invoice became past-due.
Seeing that kind of sensitivity is what sold our CFO on this model.
T&R: How do you use this technology now? Did you also change A/R processes?
SD: When we put in the system, we had credit managers in each of our five siloed business groups doing both credit and collections. Soon after we installed the software, we split out the credit risk function from collections. The skills are very different between those positions. Collections requires people who are accurate and intelligent, but it also requires skills around building a relationship with a customer. So we hired collections specialists and consolidated collections activities for all five divisions with those people. Along the same lines, because credit management involves really analyzing financial statements, we hired accounting-level people to do that job. Now we’re able to do deeper analysis, which has reduced our bad debt, and we’re developing better relationships with customers so we’re getting paid faster.
T&R: So, the credit managers use the predictive analytics to determine a particular customer’s true credit risk before they decide how much credit to extend? And the collections staff use the predictive analytics to prioritize their workload?
SD: Yes. In our previous environment, we segmented customers by type. We knew smaller contractors would be more risky and larger distributors would be less risky, so we had different follow-ups for different classes of customers. Now it doesn’t matter what category a customer fits into. If the risk score shows that a large distributor is a higher risk, we will call that customer more than the small contractors that GETPAID is telling us are lower-risk.
When our collectors look at their portfolio, it’s clear which customers they should attack first. We’re re-scoring our customers on a monthly basis, and when the score shows they are a high risk, they show up at the top of the collector’s list. It doesn’t matter how they did on last month’s score. If you’re at the top of the list right now, you’re going to be hearing from us every couple of days.
T&R: Has this worked?
SD: It’s worked like a charm. Our receivables aging has improved dramatically. Right now, 95 percent of our customer accounts are current. If we include the 1-to-15-day bucket, 99.5 percent of our customer accounts are current. In our industry—or, for that matter, any industry—that’s exceptional. Along the same lines, we look at invoices more than 60 days past-due. That used to be a bear of a number for us. We’ve reduced it so much over the past two years, we’ve reached our stretch goal, which senior management thought we’d never get to. We have reduced our over-60-day invoices by 60 percent this year over last year. We’re at our stretch goal for write-offs, as well.
Another related benefit is that we can be more targeted with our bad-debt reserve. We use our risk score and aging to define a reserve for bad debt. So if we have an account that is current but is showing up as higher-risk, when we allow a new order to go out, we have a better sense of how much we should reserve.
T&R: Is the improvement in your A/R metrics partly a result of the improving economy?
SD: Yes, I want to be clear. The health of our customers has improved. 2009 and 2010 were bad years for anyone in construction, and we’re coming out of it. But we’re not out of the woods by any means. There’s still trouble in the industry. So although some of the improvement in our receivables is organic, due to the improving health of the industry, a significant portion is clearly because we’ve put in place the appropriate processes and key performance measures.
T&R: To what degree has this new approach improved customer service?
SD: It gives us one face to the customer and simplifies the customer experience. For example, we use the system to classify deductions, to look at problems and see if we have an unfavorable trend with a customer. Then we can be a lot more proactive when there are issues. I was just on the phone with a large customer who buys across all our product lines. I asked him to send me his A/P [accounts payable] register so I could compare his Excel file to my parent-level customer statement. I was able to proactively get back to him, outlining the differences, before I received his check. I was able to identify problems and ask him to focus on the missed items. We used to wait to raise issues until after the payment was applied. Now our system allows us to take the shoe leather out of the process for our customers.
On the flip side, we can also classify overpayments. We had an example recently of a contractor that sent us a $20,000 check. Since our cash application had no open invoice to apply it to, the payment became a problem “overpayment” in the system. We looked into it and found that it was a remittance for an invoice that was two years old, which had already been paid. A collection analyst called the customer and alerted her accounts payable contact of the situation. She was so grateful! It was an error on their part, so we immediately returned the money. Through that process, the customer saw that we’ve got her back, and our analyst was able to develop a deeper, trusting relationship with that customer.
One other benefit I should point out is that we’ve also seen a dramatic improvement in terms of employee satisfaction. Now our staff knows that every transaction that appears in their collection queue is meaningful. We no longer have situations where our analysts complain, ‘We know our customer’s pay cycle is on Friday, but the system alerted us to call them on Thursday. The customers are annoyed with us, and we’re spending time following up needlessly.’ Our employees want to add value in every interaction they have with a customer. Now they really do.
T&R: Was there any reluctance within the company to consolidate receivables information across the business units?
SD: There was some. It’s normal. If you have a credit manager you’re used to working with autonomously, you’re going to be nervous about moving to a shared service center. Some sales managers and controllers for our business units were concerned going in. But seeing is believing. All of those naysayers are thrilled with the results, the focus, and the passion we have as a team to deliver results.
At the corporate level, Saint-Gobain is thrilled too. Thankfully, our company is able to support several capital-intensive initiatives, including building new plants and investing in expanding others. It helps the process along when you are turning the receivables as well as we are able to. It is all about providing the working capital needed to support investments in the future of the business.
T&R: So, would you recommend that companies with more traditional credit and collections functions look for ways to incorporate predictive analytics into their processes?
SD: Absolutely. I’m in love with our process; I’m just so proud of it. I don’t even go into my ERP system anymore. I have everything I want in one place on my desk, with no paper, working over a $3 billion queue. I’m loving it.