Two big card issuers, the same quarter, the same economy — and a 2x gap in delinquency. The aggregate number that smooths that gap away is answering a different question than the one you're asking.

Pull the Q1 2026 numbers for two of the largest U.S. card books. JPMorgan's card 30+ day delinquency rate was about 2.17%. Synchrony's was 4.54% (JPM Q1 2026 earnings supplement; Synchrony Q1 2026 earnings, Apr 21, 2026). Same macro, same product, more than double the delinquency. Nothing about the interest-rate environment or the labor market explains that gap. Composition does. And the moment you average across a book that contains both kinds of borrower, you produce a number that describes almost none of your actual accounts.

That's the problem with reading a portfolio by its mean right now. The consumer credit market is K-shaped, and a single blended delinquency rate doesn't just lose detail — it can move in the opposite direction from your actual risk.

Why the mean lies on a bimodal book

FICO's March 2026 data shows the distribution, not just the average. A record 48.1% of consumers now score 750+ (up from 43.3% in 2019), with 24.8% in the 800–850 band — also a record. At the same time the bottom thickened: the 300–499 band grew to 3.6% from 3.2%, and 500–549 to 7.0% from 6.3%. The middle hollowed out — the 600–749 range shrank from 38.1% of consumers in 2021 to 33.8% in 2025 (FICO Credit Insights, Mar 24, 2026).

So the "average FICO of 714" sits in a thinning middle that fewer and fewer real borrowers occupy. The mass is piling up at the two ends. When you compute a portfolio delinquency rate over a book shaped like that, the average lands in the valley between the two humps — a value that's true arithmetically and misleading operationally.

Two specific failure modes follow, and you should be able to name both to your team:

Mix shift can mask deterioration. If your prime balances grow faster than your subprime balances, your blended delinquency rate can fall even while delinquency rises within every single segment. That's a Simpson's-paradox reversal, and it is not a hypothetical — it's the default behavior of a growing prime book sitting on top of a deteriorating subprime tail. The aggregate improves while your risk worsens. If your monitoring tripwire is the blended number, it will stay green through exactly the period you most need it red.

The mean is not the modal account, and loss lives in the tail. Charge-offs and loss-given-default are driven by the lower mode, which the average structurally under-weights. JPMorgan at 2.17% and Synchrony at 4.54% are experiencing the same economy; the gap is who they lend to. A blended industry number tells you about neither book.

The public data backs the divergence beyond any single issuer. The NY Fed's Q1 2026 Household Debt and Credit report (May 12, 2026) shows aggregate delinquency roughly flat — about 4.8% of balances in some stage of delinquency — while stating explicitly that the increase has been driven primarily by subprime borrowers, with prime borrowers showing only marginal deterioration. And the Fed's own charge-off and delinquency series shows credit-card 30+ DPD of 2.92% across all commercial banks versus 6.43% at banks outside the 100 largest (FRED, May 2026) — the same product, a 3.5-point spread, purely by institution mix.

What to read instead of the average

None of this is exotic. It's a monitoring stack that reads the distribution instead of collapsing it:

• Delinquency by score band and origination segment. Track each band's own trend line. A flat blended rate composed of a falling prime band and a rising subprime band is a different animal than a genuinely flat book, and only the disaggregated view tells them apart.
• Vintage / cohort analysis. Delinquency by months-on-book for each origination quarter isolates underwriting drift from seasoning and from mix. A bad vintage is invisible in a portfolio average for quarters.
• Roll rates by segment. Current → 30 → 60 → 90 transition rates move before stock delinquency does, and they move at the segment level first. This is your leading indicator.
• Distribution-aware alerting. Monitor the shape — the share of balances in each band, and migration of the modal mass — not just the mean. An alert that fires on mix shift is what keeps an improving blended number from lulling you.

This is also the honest answer to "what does continuous monitoring actually buy me." It is not just taking the same portfolio average more often. Point-in-time aggregates are quarter-end photographs; mix shift and segment-level roll-rate deterioration show up between the snapshots and in the lower mode first. Reading the book by segment, continuously, catches the turn before it's large enough to move the blended number — which, by construction, it lags.

The honest take

The title is provocative on purpose, but the aggregate isn't actually "lying." It's answering a different question than the one a risk manager asks. A blended delinquency rate is the right unit for portfolio-level capital adequacy and liquidity reporting. It's the wrong unit for "where is loss accumulating in my book." Both questions are legitimate; the error is using the capital-adequacy number to answer the loss-accumulation question.

And segmentation has real costs. Thin segments produce volatile rates and false alarms — you need minimum-cell-size rules and some smoothing, or you'll chase noise and drown in multiple comparisons. Cohort monitoring demands data infrastructure and governance that a monoline, single-band book may not need at all. If you lend to one tight slice of the score distribution, your average is probably fine, and building segment machinery is over-engineering. The case for reading the distribution is strongest exactly where the distribution is bimodal — which, right now, for most consumer books, it is.

Takeaway

Do one thing this quarter: take your headline portfolio delinquency rate and decompose it by score band and by origination vintage. If the blended line is flat or improving, check whether it's improving because prime balances are growing while subprime delinquency rises underneath. That single decomposition tells you whether your aggregate is reporting health or hiding a mix shift — and it's the difference between seeing the K and being flattened by it.

‍

‍