Your chronic late payer needs a different dunning path than your first-time 45-day outlier. Segmentation isn't complexity — it's precision.
The Difference Between Segmentation and Complexity
When AR managers hear "customer segmentation," the mental model that often comes up is: more buckets to manage, more rules to maintain, more ways for the process to break. That instinct makes sense if you've ever tried to build custom dunning logic in a standard ERP or cobbled together segment-specific reminders in a spreadsheet. The operational overhead can outweigh the benefit.
But segmentation done right isn't about creating complexity — it's about making the system's behavior match what an experienced AR specialist would do if they had time to think about every account individually. The goal isn't 12 customer segments with 12 different sequences. The goal is to stop applying the same sequence logic to accounts with fundamentally different payment behavior profiles.
This article walks through a practical segmentation framework: what data inputs to use, how to define meaningful segments without overengineering, and how to assign dunning sequences that actually respond to the differences.
The Three Data Inputs That Actually Matter
Customer payment segmentation requires only three inputs from your AR transaction history, and most mid-market companies have all three available in their ERP:
Average days-to-pay over the last 12 months — not lifetime average, but recent history. A customer's average days-to-pay from 3 years ago is mostly irrelevant to how you should handle their invoice today. The last 12 months reflects current behavior and is sensitive enough to detect recent changes.
Payment timing trend over the last 3-4 invoices — is the customer's days-to-pay moving up, down, or stable? A customer with a 12-month average of net+33 but whose last four invoices came in at net+38, net+41, net+45, net+49 is trending in a direction that the average doesn't capture. Trend is more predictive of near-term behavior than the trailing average.
Invoice volume and relationship age — how many invoices has this customer had in the last 12 months, and how long have they been a customer? This tells you how much data you actually have, and whether you're looking at an established payment pattern or insufficient history to draw conclusions.
You don't need external credit data, social signals, or complex behavioral models to build useful segments. These three inputs from your own transaction records are sufficient for a segmentation that meaningfully improves dunning precision.
Five Segments That Cover Most Mid-Market Portfolios
After working through payment data across a range of mid-market B2B portfolios, five segments cover the meaningful behavioral variation you'll find in most collections environments:
Segment 1: Stable Payers (Within or Near Terms)
Profile: Average days-to-pay ≤ net+10, low variance, 3+ invoices of history. These customers pay on time or very close to it, consistently. They don't need dunning in the traditional sense. Over-communicating with this group is a relationship cost with no collection benefit. Recommended approach: a single pre-due reminder at net-3 (three days before due), and a light follow-up only if payment hasn't arrived by net+5. No escalation to phone unless net+20.
Segment 2: Reliable Late Payers (Consistently Outside Terms)
Profile: Average days-to-pay between net+11 and net+35, low variance, established history. These customers have their own payment schedule that doesn't match your stated terms — but they pay reliably within a predictable range. The mistake most AR teams make with this segment is running the standard aggressive sequence (escalation calls at net+30) on accounts that will pay at net+28 without any intervention. That's friction with zero collection benefit and real relationship cost. Recommended approach: first reminder at net+7, second at net+20. Only escalate to a call at net+40 if the invoice hasn't paid — which, for this segment, is outside their established pattern and warrants real attention.
Segment 3: New or Low-History Accounts
Profile: Fewer than 3 invoices of payment history, or first invoice. Without established payment behavior, you can't classify this customer's pattern. These accounts need more attentive follow-up — not because they're high-risk, but because early interactions establish the payment relationship and surface process friction quickly. Recommended approach: proactive pre-due outreach at net-5 confirming invoice receipt and flagging the due date. If unpaid at net+10, phone contact. This sequence establishes expectations and builds data faster than waiting to see where the invoice lands.
Segment 4: Deteriorating Accounts (Trend-Flagged)
Profile: Average days-to-pay has increased by 10+ days over the last 3 invoices, or current invoice is 15+ days beyond their established average. This is the segment that generic dunning systems miss most consistently — because they react to current aging, not behavioral trend. A customer who averaged net+28 for 18 months but whose last three invoices came in at net+38, net+45, net+52 is telling you something about their current cash position or internal AP process. Recommended approach: proactive outreach call before the invoice is due, not after. The goal of this call isn't collections pressure — it's information gathering. Is there a dispute? An internal approval delay? A change in their AP team? This conversation at net+0 is far more productive than the same conversation at net+60.
Segment 5: High-Risk or Previously Delinquent Accounts
Profile: Prior 60+ DPD invoice history, or current trend pointing toward 60 DPD based on payment deterioration. These accounts need an accelerated sequence with earlier escalation and lower tolerance for non-response at each step. Recommended approach: reminder at net+3, phone contact at net+10, formal demand communication at net+20. Consider requiring proactive credit risk review before extending new terms to this segment.
How to Assign Customers to Segments
The classification logic needs to run at the account level, not the invoice level. When a new invoice comes in for a customer, the segment assignment determines which dunning sequence is activated for that invoice. This means the segmentation needs to be recalculated periodically — quarterly works for most portfolios — so that customers whose behavior has changed get reclassified before their next invoice enters your queue.
The most common implementation failure we see is building segments as static categories rather than dynamic classifications. A customer who was in "Reliable Late Payer" for two years but whose last three invoices show a meaningful trend shift needs to move to "Deteriorating Account" before the next invoice arrives — not after it's already 45 days past due.
At Cashvyne, recalculation happens on every payment event and on a weekly scheduled refresh. The practical implication: when a customer pays an invoice 8 days faster than their historical average, the model notes that; when they pay 12 days slower on three consecutive invoices, the segment assignment updates accordingly. This keeps the dunning logic current without requiring manual intervention.
What Segmentation Doesn't Fix
We want to be clear about the limits here. Segmented dunning improves the distribution of your AR effort — more attention on the accounts that need it, less friction on the accounts that don't. It does not eliminate the customers who won't pay regardless of how the sequence is structured.
A customer in genuine financial distress isn't moved by a better-calibrated email sequence. A dispute that's been festering for three months needs a resolution process, not a dunning sequence. A customer relationship where the sales team has made informal payment term accommodations needs a sales-finance alignment conversation, not a re-segmentation.
Segmentation also doesn't substitute for a working credit approval process. If you're extending terms to customers who shouldn't have credit, the collections problem is upstream of anything dunning logic can address. The cleanest AR portfolio is one where the credit decision at account opening was sound — segmented dunning helps you manage the variance within a reasonably healthy portfolio, not rescue a credit policy problem.
That said, for mid-market AR teams managing 80-300 active accounts with limited headcount, behavioral segmentation is consistently the highest-leverage change to how dunning effort is allocated. The same two AR specialists, with the same total volume of open invoices, produce measurably better collection outcomes when their sequences are calibrated to payment behavior rather than applied uniformly.