Stratification analysis

A stratification table (strat) shows how a loan pool distributes across bands of a given characteristic. Strats are your primary tool for understanding pool composition. They transform a loan tape with thousands of rows into digestible summaries that reveal concentration, skew, and correlation patterns.

This page covers the essential stratifications to run, how to read them effectively, and how to identify concentration risks that affect deal structuring.

What stratifications reveal

A single loan tape might contain 10,000 loans with 50 fields each—500,000 data points. Stratification tables compress this complexity into actionable insights.

The basic structure

Every strat table has the same format:

Band	Count	UPB	% of Pool	Additional Metrics
Band 1	n	$X	Y%	WA rate, WA FICO, etc.
Band 2	n	$X	Y%	WA rate, WA FICO, etc.
…	…	…	…	…
Total	N	$X	100%	Pool WA

Count: Number of loans in each band UPB: Total unpaid principal balance in each band % of Pool: Concentration in that band (can be by count or by balance) Additional Metrics: Weighted averages that reveal correlation between variables

Why both count and balance matter

A band might contain 5% of loans by count but 15% by balance. This happens when that segment has larger-than-average loans.

Both perspectives matter:

Count-based: Frequency of a characteristic; operational exposure
Balance-based: Dollar exposure; economic risk

For credit analysis, balance-based percentages typically matter more. For servicing capacity planning, count-based percentages may be more relevant.

Essential stratifications

Run these six stratifications on every pool. They cover the key dimensions that affect credit risk, structural risk, and eligibility.

Balance stratification

Shows loan size distribution and concentration in large loans.

Current balance	Count	UPB ($M)	% Pool	Avg rate
$0-10,000	1,200	$8.4	16.8%	12.5%
$10,001-25,000	1,800	$31.5	63.0%	11.2%
$25,001-50,000	800	$28.0	16.0%	10.8%
$50,001+	200	$14.0	4.2%	10.2%
Total	4,000	$50.0	100%	11.1%

Illustrative data. Pricing for reference only; see pricing disclaimer.

What to look for:

Is the pool concentrated in a few large loans? The top band here is 4.2% of balance but only 5% of count—large loans exist but don’t dominate
Do larger loans have different pricing? Here, larger loans carry lower rates (better credit)—expected pattern
Does any single loan exceed concentration limits? Run a separate check for individual loan concentrations

Band design: Use bands relevant to the product. Consumer loans might use $5K bands; equipment might use $50K bands. Match bands to the typical transaction size.

Rate stratification (WAC analysis)

Shows interest rate distribution and correlation with credit.

Rate band	Count	UPB ($M)	% Pool	Avg FICO
6-8%	400	$12.0	24.0%	745
8-10%	1,200	$20.0	40.0%	710
10-12%	1,600	$14.0	28.0%	675
12%+	800	$4.0	8.0%	640
Total	4,000	$50.0	100%	698

Illustrative data.

What to look for:

WAC calculation: The weighted average coupon is 9.8% here. This is your yield driver
Credit-rate alignment: Higher-rate bands should correlate with lower credit. They do here—appropriate risk-based pricing
High-rate tail: The 8% in 12%+ rates represents higher-risk exposure. Is this within guidelines?
Misalignment signals: If the 12%+ band showed 720 average FICO, something is wrong—either pricing is inefficient or data is corrupted

Credit score stratification

Shows credit quality distribution—the most important predictor of default.

FICO band	Count	UPB ($M)	% Pool	Avg rate
750+	600	$10.5	21.0%	8.2%
700-749	1,200	$17.5	35.0%	9.5%
660-699	1,400	$15.0	30.0%	11.2%
620-659	600	$5.5	11.0%	13.5%
<620	200	$1.5	3.0%	15.8%
Total	4,000	$50.0	100%	10.3%

Illustrative data.

What to look for:

Credit concentration: This pool skews near-prime to prime. Only 3% is subprime (<620)
Guideline compliance: If guidelines say minimum FICO 620, the <620 band shouldn’t exist (or should be tagged for exclusion)
Distribution shape: A bimodal distribution (spikes at both high and low credit) might indicate two different products or channels mixed together
Credit migration: If you have both origination FICO and current FICO, the shift tells you about seasoning and economic conditions

Geographic stratification

Shows state concentration—critical for regional economic exposure.

State	Count	UPB ($M)	% Pool
California	800	$12.0	24.0%
Texas	600	$8.0	16.0%
Florida	400	$5.5	11.0%
New York	350	$5.0	10.0%
Other (46 states)	1,850	$19.5	39.0%
Total	4,000	$50.0	100%

What to look for:

Single-state concentration: California at 24% exceeds typical 15-20% limits. This affects borrowing base calculations
Natural vs. forced concentration: An originator based in California may naturally have high CA exposure. A nationwide originator with 40% in one state is making choices
Regulatory exposure: Some states have more borrower-friendly foreclosure/collection laws, affecting recovery
Economic correlation: Multiple high-concentration states in the same economic region (e.g., oil states) compound risk

Vintage stratification

Shows origination timing and pool seasoning.

Origination year	Count	UPB ($M)	% Pool	Avg seasoning (mo)
2024	2,500	$25.0	50.0%	6
2023	1,500	$15.0	30.0%	18
2022	750	$7.5	15.0%	30
2021	250	$2.5	5.0%	42
Total	5,000	$50.0	100%	15

What to look for:

Seasoning mix: This pool is young—50% originated in the last 6 months. Performance is unproven
Single vintage concentration: If 80% originated in one quarter, you have a concentrated bet on that quarter’s underwriting and economic conditions
Vintage trends: Run performance metrics (DQ, losses) by vintage to see if newer vintages perform differently
Growth rate implications: Rapid recent originations (2024 spike here) suggests growth mode—monitor for credit loosening

Delinquency stratification

Shows current payment status—the near-term risk snapshot.

DPD bucket	Count	UPB ($M)	% Pool
Current (0)	4,250	$42.5	85.0%
1-29 DPD	400	$4.0	8.0%
30-59 DPD	200	$2.0	4.0%
60-89 DPD	100	$1.0	2.0%
90+ DPD	50	$0.5	1.0%
Total	5,000	$50.0	100%

What to look for:

30+ DQ rate: This pool shows 7% (4% + 2% + 1%). Compare to benchmarks: concerning for prime auto (expect 2-3%), normal for subprime consumer (expect 6-10%)
Pipeline building: If 30-59 and 60-89 buckets are growing faster than historical rates, defaults are coming
Cure patterns: Compare to prior tapes. Are loans moving from 30 to 60 to 90 (rolling), or from 30 back to current (curing)?

Reading stratification tables

Numbers on a strat table only become insight when you know what patterns to look for.

Pattern 1: Skew

Is the distribution normal or concentrated?

Normal distribution: Loans spread across bands with most in middle ranges. The balance strat above shows this—most loans are $10-25K.

Concentrated distribution: One or two bands dominate. A vintage strat showing 80% of the pool in the last 3 months is concentrated—you’re betting on recent underwriting.

Bimodal distribution: Two distinct peaks. This often indicates two different products, channels, or credit tiers mixed in one pool. Consider analyzing separately.

Pattern 2: Outliers

Are there bands with unusual characteristics?

A rate band showing 2% of loans at >20% rate in a prime pool deserves investigation:

Is this a different product accidentally included?
Are these rates correct, or data errors?
Were these loans originated under different guidelines?

Outliers aren’t necessarily problems, but they require explanation.

Pattern 3: Correlation

Do characteristics align as expected?

Expected correlation: Lower FICO should correlate with higher rate. The rate and credit strats above show this pattern—appropriate risk-based pricing.

Unexpected correlation: If the highest-rate band shows the highest FICO, something is wrong. Either:

Pricing discipline is weak (not risk-adjusting)
Data is corrupted
There’s an explanation (e.g., high-rate band is all new originations before rate drops)

Always cross-reference strats to check correlation.

Pattern 4: Concentration breaches

Does any single band exceed typical limits?

Concentration type	Typical limit
Single state	15-25%
Single vintage month	5-10%
Single credit band	30-40%
Single balance band	40-50%

Breaches don’t kill deals but affect advance rates and trigger negotiations.

Concentration analysis and the HHI

Beyond individual band limits, overall pool diversification matters. The Herfindahl-Hirschman Index (HHI) quantifies concentration in a single number.

Why concentration matters

A $100M pool with 1,000 equal-sized loans has different risk than a $100M pool with 100 loans where the top 10 represent 60% of the balance. The second pool is far more exposed to idiosyncratic risk—individual borrower outcomes matter more.

Diversification reduces idiosyncratic risk. Concentration amplifies it.

Calculating HHI

Formula:

HHI = Sum of (each loan's % of pool)^2 x 10,000

Perfectly diversified example: 10,000 equal loans, each 0.01% of pool HHI = 10,000 x (0.0001)^2 x 10,000 = 1

Highly concentrated example: One loan is 50% of pool, rest distributed HHI approaches 2,500

HHI interpretation

HHI range	Interpretation	Typical asset classes
<100	Well diversified	Auto, consumer, large mortgage pools
100-1,000	Moderate concentration	Equipment, smaller consumer pools
>1,000	Concentrated	Middle market, specialty finance

Equipment and middle market pools routinely have HHI >1,000. This isn’t disqualifying—just requires different structuring (more credit enhancement, tighter covenants).

Applying concentration in practice

Single obligor limits: Typically 1-2% of pool. One borrower default shouldn’t materially impact the facility.

Top 10/20 obligor limits: Often 10-15% for top 10, 20-25% for top 20. Ensures minimum diversification.

Geographic limits: Single state typically 15-25%. Region (e.g., “Oil Belt states”) might have aggregate limits.

Excess concentration haircuts

When a pool exceeds concentration limits, lenders apply haircuts. The excess concentration amount is either excluded from the eligible pool or advanced at a lower rate.

Worked example:

Pool: $50M California concentration: 28% ($14M) Limit: 20% ($10M) Excess: 8% ($4M) Haircut on excess: 50%

Calculation:

Eligible pool (within limits): $50M - $4M = $46M at 85% advance = $39.1M
Or: Full pool at reduced effective advance
- Haircut reduction: $4M x 50% = $2M
- Effective borrowing base: ($50M x 85%) - $2M = $40.5M
- Effective advance rate: $40.5M / $50M = 81%

This is how concentration limits translate directly into funding capacity.

Building effective strat packages

Standard package contents

For a new pool diligence, prepare:

Strat	Primary insight	Include in package?
Balance	Loan size distribution	Always
Rate/WAC	Yield and pricing discipline	Always
Credit score	Default risk distribution	Always
Geography	Regional concentration	Always
Vintage	Seasoning and origination trends	Always
Delinquency	Current performance	Always
Term	WAL and amortization	Usually
LTV (secured)	Loss severity exposure	If applicable
Collateral type	Recovery characteristics	If applicable
Channel	Origination quality variance	Sometimes

Cross-tabulation

Single-dimension strats miss interaction effects. Cross-tabs reveal them.

Credit by Vintage:

Vintage	<660 FICO %	660-699 %	700+ %
2024	18%	35%	47%
2023	12%	32%	56%
2022	8%	28%	64%

This reveals credit loosening over time—2024 has more low-FICO loans than prior vintages. A single credit strat wouldn’t show this trend.

Geography by Delinquency:

State	Current %	30+ DQ %
California	87%	13%
Texas	92%	8%
Florida	83%	17%
Pool Avg	85%	15%

Florida is underperforming the pool. California is at pool average despite concentration concerns. This is actionable information.

Commentary requirements

Never present strat tables without commentary. Add one to two sentences below each table:

“Pool skews near-prime with only 3% subprime exposure. Credit-rate alignment is appropriate—higher risk borrowers pay higher rates. The <620 FICO band, while small, should be monitored as it exceeds the 2% subprime limit in Series A guidelines.”

Commentary should:

Summarize the key finding
Compare to benchmarks or limits
Note any action items or concerns

Key takeaways

Run the six essential strats on every pool: Balance, rate, credit, geography, vintage, and delinquency cover the key risk dimensions.
Look for patterns, not just numbers: Skew, outliers, correlation, and concentration tell you more than individual band percentages.
Cross-reference strats: Credit should correlate with rate. Vintage trends should appear in performance. Check that patterns are internally consistent.
Quantify concentration with HHI: A single number that enables comparison across pools and time.
Translate breaches into economics: Excess concentration means lower advance rates or exclusions. Understand the impact.
Add commentary to every table: Numbers without interpretation aren’t analysis—they’re just data.