Same Score, Different Customer
Credit scores have decoupled from behaviour — and treatment built on them shows it
Do two customers with the same credit score behave the same way?
No. A credit score predicts default probability, coarsely and after the fact. It says nothing about whether a customer responds to a payment plan, prefers to self-serve, or needs a restructure. Score migration has also shifted what a given number means, so customers in the same band can need opposite treatment.
The credit score is the industry’s load-bearing segmentation tool, and it is carrying more weight than it can bear. Two findings make the point.
Score migration changed what the numbers mean
First, scores have migrated in ways that changed what a given number means. Federal Reserve researchers found that pandemic-era score migration alone added roughly five to six percentage points to measured subprime delinquency rates. Borrowers who moved up the score distribution carried their old risk with them into new bands.¹ TransUnion measured it directly: consumers who migrated to higher score tiers performed like borrowers about 25 points lower than their new score suggested.²
The middle of the distribution is meanwhile hollowing out. The 600–749 band shrank from 38.1% of consumers in 2021 to 33.8% in 2025, and the average FICO score has posted its first multi-point decline in over a decade, driven partly by a reporting event (resumed student-loan delinquency reporting) rather than by any change in behaviour.³ When 3.6 million borrowers default in two quarters and their scores drop 91 points on a policy change, the score is describing the policy, not the person.⁴
“The score is describing the policy, not the person.”
Credit score segmentation predicts default, not response
Second, and this is the deeper problem, the score was never designed to predict how a customer behaves under stress or treatment. It says nothing about whether this customer responds to a payment plan or ignores it, self-serves at 11 p.m. or answers a 2 p.m. call, needs a reminder or needs a restructure.
Peer-reviewed research on mortgages finds that 94% of defaults are triggered by cash-flow shocks and life events, and 70% trace to cash flow alone. Strategic default, even among underwater borrowers, is rare.⁵ If that pattern holds beyond mortgages, as we believe it does, most delinquency is an ability and timing problem. Yet score-band treatment waterfalls assume a willingness problem, and apply escalating pressure to customers who needed a different payment structure, not a sterner letter.
of mortgage defaults are triggered by cash-flow shocks and life events. Strategic default is rare.
Source: Ganong & Noel, Quarterly Journal of Economics, 2023
Same score, different outcome: the BNPL evidence
The natural experiment hiding in the BNPL data shows how much structure can matter. The Richmond Fed found that deep-subprime consumers, the same people showing double-digit card delinquency, repaid their BNPL instalments 96% of the time.⁶ Same consumers, same scores, radically different outcomes, because small, paycheque-aligned instalments fit their cash flow and the product’s friction was near zero. Autopay defaults, small balances and the wish to keep access all played a part; none of them rescues the score as a predictor of who pays.
Friction compounds the same way in the other direction. A confusing statement, a portal that requires a password reset, a payment plan that takes three calls to establish: each is small alone. Together, in our program experience, they convert recoverable customers into rolling balances.
Behavioural segmentation lifts cure rates in every risk band
Our own program data makes the same point. Working with a major North American bank, Symend tested behavioural segmentation against the bank’s legacy risk-based segmentation across a credit card portfolio, holding risk band constant. Behavioural segmentation produced higher cure rates in every band: 88.9% against 83.3% in the lowest-risk band, 86.8% against 80.8% in medium, 79.0% against 73.3% in medium-high, and 61.3% against 55.0% among accounts the risk model could not classify at all. Results were validated as statistically significant.
Symend’s Delinquency Archetypes, which SymendCure uses to tailor each customer’s journey, are built on two dimensions a score can’t see: capacity to pay and readiness to pay. You can see how that segmentation works in practice.
What this means for issuers
Two customers with the same score are not the same customer. Across banking and financial services portfolios, segmentation built on behavioural response, meaning how people actually engage, pay and react to treatment, is where cure-rate improvement now lives.
“Scores tell you whom to underwrite. They do not tell you how to treat.”
The bottom line
The score still has a job, but treatment isn’t it. The full paper, “The Averages Are Lying,” carries the complete argument, including the early-warning signals that precede delinquency and the economics of moving intervention upstream.
A 7-part series
The New Delinquency Playbook
Seven short posts drawn from two Symend research papers. The first, “The Averages Are Lying,” argues that consumer stress has become concentrated rather than widespread, that delinquency is a lagging indicator, and that credit scores have decoupled from behaviour, so issuers need to read behaviour early, re-segment on it, and move intervention upstream. The second, “AI Is an Engine,” argues that collections AI amplifies whatever operating philosophy it is given, so it only works when behavioural science steers it, outcome objectives replace deflection metrics, and credit decisions stay on the deterministic side of the line.
From “The Averages Are Lying”
- Part 1 of 7 Delinquency Is a Lagging Indicator The behaviour that predicts default shows up one to two years before the missed payment
- Part 2 of 7 The K-Shaped Consumer Why every aggregate dashboard is wrong about consumer stress
- Part 4 of 7 The Prevention Curve The economics of acting before the roll
From “AI Is an Engine”
- Part 5 of 7 The 95% Problem Collections is adopting AI faster than it is getting results
- Part 6 of 7 The Stigma Exception Most consumers dislike service AI. Debt may be different — if the engine is steered right
- Part 7 of 7 The Deterministic Line What AI must never decide in collections — and where it earns its keep
Frequently asked questions
What is credit score segmentation in collections?
Grouping delinquent customers by score band and assigning treatment by band. It’s simple to run, but it assumes customers in the same band will respond to outreach the same way.
Why isn’t a credit score enough to segment collections customers?
Scores predict default probability, not response to treatment. They can’t tell you whether someone needs a reminder, a payment plan or a different channel.
What is behavioural segmentation?
Segmentation built on how customers actually engage, pay and respond to outreach. Symend’s Delinquency Archetypes sort customers by their capacity and readiness to pay.
Why do most borrowers default?
Research on mortgages finds 94% of defaults are triggered by cash-flow shocks and life events. Strategic default, where a borrower could pay but chooses not to, is rare.
Does behavioural segmentation improve cure rates?
In a Symend test at a major North American bank, behavioural segmentation produced higher cure rates than risk-based segmentation in every risk band, with statistically significant results.
Endnotes
- Federal Reserve, FEDS Notes, “The Effects of Credit Score Migration on Subprime Auto Loan and Credit Card Delinquencies,” January 2024.
- TransUnion, “Score Migration Impact to the Credit Ecosystem,” July 2023 (Q3 2021 pandemic-era cohort; the effect may have faded since). VantageScore disputes “score inflation” framing (June 2023); the claim here is about migration cohorts’ performance, not a broken scale.
- FICO, inaugural FICO Score Credit Insights Report, September 2025.
- Federal Reserve Bank of New York, Liberty Street Economics, “Federal Student Loan Defaults Return After Pandemic Pause,” May 2026. Simultaneity measured at default, not lead time. Student-loan serious delinquency subsequently fell from 12.88% to 7.83% year over year (Q2 2026), partly mechanically as defaults left the delinquent stock.
- Ganong, P., and P. Noel, “Why Do Borrowers Default on Mortgages?” Quarterly Journal of Economics 138(2), 2023. Mortgage-specific; extension to other products is our inference.
- Federal Reserve Bank of Richmond, Economic Brief 26-05, 2026 (2021–2022 BNPL repayment data). Separately, 41% of BNPL users self-reported a late payment in 2025 surveys (up from 34%); self-report and measured instalment repayment are different measures.
