Latest / Cash Flow Conversations with Fexingo: Working Capital, Receivables, and Small Business Finance / How to Predict Customer Late Payments Before They Happen
Transcript
- Lucas: Let me tell you about a wholesale distributor in Atlanta that had a pretty typical receivables problem. Average invoice paid in 52 days, terms were net 30, so they were living with a twenty-two-day gap they couldn't explain, customer by customer. Luna: That's a huge chunk of working capital just sitting out there, especially if you have thin margins. Lucas: Exactly. So they did something that sounds obvious but almost no one does at the small-business level: they pulled every invoice from the previous two years, tagged each one with whether it was paid early, on time, or late, and then looked for patterns. And they found that about sixty percent of their late payments came from just twenty percent of their customers. Luna: So the Pareto principle applied to payment behavior. Lucas: Right. But here's the twist: those customers weren't the ones who looked risky on paper. They had decent credit scores, they placed large orders, they'd been customers for years. The pattern was in the timing of their payments relative to their own past behavior. So the distributor built what I'd call a payment propensity score — a simple number from zero to a hundred that predicts the likelihood an invoice will be paid late. Luna: How did they actually calculate it? Were they using machine learning or just a spreadsheet? Lucas: Spreadsheet to start, actually. They weighted three factors. First, the customer's average payment speed over the last six months relative to terms. Second, the variance in that speed — a customer who pays anywhere from day 15 to day 60 is less predictable than someone who always pays on day 35. Third, the seasonality: a lot of their customers in landscaping equipment paid slower in Q4, so they factored that in. Luna: That third one is interesting because it's not about the customer being bad, it's about their business cycle. Lucas: Exactly. And that made the scoring feel fairer. Once they had a score for each customer, they segmented them into three buckets: green zone scores above 70, meaning high probability of paying on time; yellow zone 40 to 70, meaning a decent chance of delay; and red zone below 40, meaning almost certainly late. Luna: Then what did they do with that segmentation? Lucas: They changed their outreach timing. For green-zone customers, they sent a standard reminder three days before the due date. For yellow, they sent a friendly email five days before, and then a phone call on the due date itself. For red, they started the conversation ten days before the due date — not demanding payment, but asking if there were any issues with the invoice, offering to set up a payment plan if needed. Luna: And did that actually move the needle? Lucas: Over six months, their average receivables aging dropped from 52 days to 31. The biggest improvement came from the red-zone customers — they weren't being chased, they were being helped. And the distributor didn't lose a single customer from the more aggressive outreach because it came across as proactive support, not harassment. Luna: That's a really concrete example. But I wonder — does this approach require a lot of data or technical know-how? Because for a lot of small businesses, even pulling two years of invoice data might feel daunting. Lucas: It used to be a big hurdle, but now there are affordable tools that integrate with QuickBooks, Xero, FreshBooks — they automatically calculate payment propensity scores. Some are even free for basic use. The key insight is that you don't need a data scientist. You need a system that turns your existing invoice data into a simple signal. Luna: And I imagine the behavioral economics piece is important too. There's a concept called 'payment momentum' — if a customer has paid late three times in a row, they're more likely to do it again because it becomes a habit. Lucas: Exactly. That's why the scoring model is so powerful. It catches those habits before the invoice is even due. And the distributor found that once they broke the cycle for a few customers by offering a small prompt-payment discount, those customers actually shifted into the green zone permanently. Luna: So it's not just about predicting, it's about intervening early enough to change the behavior. Lucas: Right. And that's the practical takeaway: you can't fix a late payment after it's 60 days overdue. But you can probably prevent it if you know it's coming. And the data to know that is already sitting in your accounting system, just waiting to be used. Luna: One risk I want to flag though — if you're too aggressive with red-zone customers, especially your biggest accounts, you could damage the relationship. How do you balance prediction with trust? Lucas: That's a fair concern. The Atlanta distributor actually tested this by not telling customers they were being scored. They just adjusted their own internal processes. The customer only saw a more helpful and timely communication. And they actually got positive feedback — some customers said they appreciated the early check-in because it helped them prioritize their own payables. Luna: So it's almost like a customer service upgrade disguised as a collections strategy. Lucas: Exactly. And the other thing they did was use the scores to decide which customers to offer early-payment discounts to. Instead of offering two percent off for paying in ten days to everyone, they only offered it to yellow-zone customers. Green-zone customers were already paying on time, so the discount was wasted. Red-zone customers wouldn't take it anyway. That saved them money. Luna: That's a smart use of segmentation. So you're not just predicting late payments, you're optimizing your incentives based on who will actually respond. Lucas: Precisely. And this approach scales. A business with fifty customers can do it on a spreadsheet. A business with five hundred can use software. The principle is the same: use your own data to spot patterns before they become problems. Luna: Before we go further, I want to mention something that's been on my mind. We put a lot of work into these episodes — researching cases like this distributor, thinking through the numbers, making sure it's useful. And we keep the show ad-free because we think that's a better experience for you. Lucas: Yeah, that's a deliberate choice. No sponsor reads, no mid-roll interruptions, just the conversation. If you've gotten value from these episodes and want to support that approach, there's a simple way. It's buy me a coffee dot com slash fexingo. Luna: And honestly, even just knowing people are listening and finding it useful keeps us going. But if that's something you can do, it helps us keep this going without ads. Lucas: Alright, back to the distributor. So once they had the scoring in place, they started tracking something else: the correlation between payment speed and order frequency. They noticed that customers who paid late tended to order less frequently over the following quarter. Luna: Interesting. So late payment was actually a leading indicator of churn. Lucas: Exactly. And that changed how they thought about collections. Instead of just seeing a late invoice as a cash flow problem, they saw it as a relationship health signal. So they started using the payment propensity score as part of their account management process — if a customer slipped from green to yellow, the sales rep would reach out to see if there was an issue with the product or service, not just the payment. Luna: That's a whole different mindset. You're using financial data to drive customer retention, not just collections. Lucas: And it worked. They reduced churn by about 15 percent in that first year. So the payment prediction system paid for itself in multiple ways: faster cash, lower bad debt, and happier customers who felt understood. Luna: So for someone listening who wants to try this, what's the first step they should take this week? Lucas: Pull the last twelve months of invoices. For each customer, calculate the average number of days past invoice date that they paid. Then sort customers by that average. The top 20 percent are your red zone. Start there. Don't change anything yet — just observe. See if the pattern holds. Then decide on one intervention, like calling them five days before due instead of three. Measure the change. That's all it takes to start. Luna: It's really about turning raw data into a simple signal you can act on. And this episode gave us a clear example of how that works in practice. Lucas: Yeah. And we'll keep digging into other angles on cash flow in future episodes. Maybe next time, how to handle the opposite problem — customers who pay too early and mess up your cash flow forecasting. Luna: That's a real issue too. Alright, thanks for listening. Lucas: We'll catch you next time.