What is PAR30 and PAR90, and Why Do They Matter?

Portfolio at Risk — PAR — is the single most-watched health metric in microfinance lending, and also one of the most frequently misunderstood. This guide explains what it actually measures, how PAR30 and PAR90 differ, how they connect to Bank of Tanzania's loan classification bands, and where the metric can quietly mislead you if you're not also watching the right supporting signals.

What does "Portfolio at Risk" actually measure?

Portfolio at Risk measures the proportion of a loan portfolio's outstanding balance that is exposed to delinquency — not the number of loans that are late, and not just the amount of the missed installment. This distinction matters more than it first appears. Counting overdue loans by number understates risk whenever a handful of large loans go bad, because ten small delinquent loans and one large delinquent loan can look identical on a "number of loans overdue" report while carrying wildly different exposure. Counting only the amount actually missed — the overdue installment, rather than the loan's full remaining balance — understates risk in the opposite direction: a borrower who has missed one installment on a large loan is treated as a small problem, when in fact the entire remaining balance of that loan is now in question.

PAR solves both distortions by taking a simple, conservative view: the moment a loan crosses a delinquency threshold (30 days past due, 90 days past due, or whichever threshold you're measuring), its entire outstanding balance — not just the missed piece — is counted as "at risk." This is why PAR is the standard metric used across the microfinance industry globally, and why regulators and rating agencies rely on it far more heavily than raw arrears counts.

PAR30 vs. PAR90 — what's the actual difference?

Both PAR30 and PAR90 use the exact same formula — outstanding balance of loans past a given number of days due, divided by total gross outstanding portfolio — the only difference is the day-count threshold, and that difference changes what the number is actually telling you operationally.

PAR30

PAR30 counts any loan more than 30 days past due. It is an early-warning signal. At 30 days, a loan is often still recoverable through normal collections activity — a reminder call, a field visit, a renegotiated payment date — so a rising PAR30 is primarily a signal to intervene, not yet a signal to write anything off. Institutions typically watch PAR30 weekly or even daily at the branch level precisely because it moves fast and gives the most lead time to act.

PAR90

PAR90 counts loans more than 90 days past due. By the time a loan reaches this threshold, normal collections methods have usually already been attempted and failed, which is why PAR90 functions as a much more severe signal — often treated internally as close to a "point of no return" for that specific loan. PAR90 is the figure most heavily used in provisioning and write-off decision-making, because a loan that has been delinquent for three months has a materially lower probability of full recovery than one at 31 days.

Worked example

The two metrics are best understood side by side on the same portfolio.

Total gross loan portfolio ............ TZS 500,000,000
Loans >30 days overdue (balance) ...... TZS  40,000,000
  → PAR30 = 40,000,000 / 500,000,000 = 8%

Of that, loans >90 days overdue ....... TZS  15,000,000
  → PAR90 = 15,000,000 / 500,000,000 = 3%

Notice that PAR90 is always a subset of PAR30 by definition — every loan more than 90 days overdue is, by construction, also more than 30 days overdue. Watching the gap between the two (8% vs 3% in this example) is itself informative: a wide gap suggests a lot of loans are stuck in the 31–90 day window, which is exactly where active collections intervention still has the best odds of working.

How PAR connects to BOT's 5-band loan classification

Bank of Tanzania's microfinance regulations classify every loan into one of five bands based on days past due, and this classification directly drives how much an institution must provision against potential losses. Per the classification schedule set out in the SACCOS-focused Microfinance (Savings and Credit Cooperative Societies) Regulations (GN No. 675): Current (0–30 days past due, no special provisioning); Especially Mentioned (31–90 days past due, minimum 10% provisioning); Substandard (91–180 days past due, minimum 30% provisioning); Doubtful (181–365 days past due, minimum 50% provisioning); Loss (over 365 days past due, 100% provisioning). This is the GN No. 675 schedule; institutions supervised under the Tier 1/Tier 2 regulations (GN No. 679) should confirm the exact bands and day-count thresholds against the current regulation text, which this article does not independently reproduce.

The conceptual link to PAR is direct: PAR30 roughly tracks everything that has moved out of the "Current" band, while PAR90 roughly tracks everything that has moved into "Substandard" or worse. This is exactly why regulators, boards, and rating agencies treat PAR30/PAR90 as a fast proxy for classification and provisioning trends without waiting for the full classification run to close — see the glossary entries for Loan Classification and Provisioning for the full definitions.

Why PAR alone can mislead you

PAR is powerful precisely because it's simple — but that simplicity means it can hide real problems if read in isolation.

Portfolio growth can mask rising PAR

Because PAR is a ratio, a rapidly growing denominator (total portfolio) can make a genuinely worsening numerator (delinquent balance) look flat or even improving. A branch that is disbursing aggressively can show a stable or falling PAR30 even while the absolute number of problem loans is climbing — the growth is diluting the ratio, not fixing the underlying delinquency. Always look at PAR alongside the absolute delinquent balance and portfolio growth rate together, never the ratio alone.

Restructured loans can artificially "cure" PAR

Refinancing or restructuring a delinquent loan resets its days-past-due counter, which can pull it back into the "Current" band and out of PAR30/PAR90 — but only the reported number changes if the borrower's actual repayment capacity hasn't improved. This is exactly why a transparent, disciplined restructuring policy matters as much as the PAR number itself; see Lending operations for how approval and restructuring controls should be structured to prevent PAR from being managed on paper rather than in reality.

Branch-level and product-level PAR tell a different story

A healthy blended portfolio PAR can conceal one branch or one loan product carrying a dramatically worse figure than the rest. Aggregate PAR is a starting point for a conversation, not the end of one — it should always be sliced by branch, loan officer, and product before being treated as a clean bill of health.

How to keep PAR30/PAR90 under control operationally

Controlling PAR is less about the metric itself and more about the discipline of what happens before day 30. A structured dunning ladder — SMS reminder, phone call, field visit, formal notice, and only then legal/recovery action, each triggered automatically by days-past-due rather than left to individual officer judgment — is the single highest-leverage operational control, because it guarantees consistent, timely contact instead of relying on memory. Capturing and following up on promise-to-pay commitments turns vague follow-up into a trackable queue rather than a guess. And officer intervention before day 30 — not after a loan has already crossed into PAR30 — is what actually keeps the PAR30 number itself low, since by definition nothing can be done to prevent a loan from entering PAR30 once it's already 31 days overdue. See Collections & delinquency for how this ladder should be configured in practice.

How Ustawi calculates and reports PAR

PAR30 and PAR90 aging in Ustawi is calculated automatically from each loan's live repayment schedule and current days-past-due, updated as repayments post — not from a manual end-of-month spreadsheet reconciliation. It's viewable broken down by branch, product, and loan officer (directly addressing the "branch-level PAR tells a different story" risk above), and the same underlying DPD data feeds straight into BOT 5-band classification and provisioning, so the two numbers never drift apart the way they can when they're calculated in separate, disconnected worksheets. See Regulatory reports & filings and the Regulatory Reports & Filings app for the full picture of how classification, provisioning, and portfolio disaggregation are handled end to end.

Key takeaways

• •PAR measures balance at risk, not number of loans or amount missed — always the full outstanding balance of a delinquent loan.
• •PAR30 is an early-warning signal; PAR90 is a severe, provisioning-driving signal — they're the same formula at different thresholds.
• •PAR connects directly to BOT's 5-band classification and the provisioning rates tied to each band.
• •PAR can be diluted by growth or artificially cured by restructuring — always read it alongside absolute balances and a transparent restructuring policy.
• •The best PAR control is operational discipline before day 30, not intervention after the fact.