DATA & SIGNALS
Early Warning Signals in SME Credit: The Threshold Table That Turns Data Into an Alert
An SME early warning threshold converts a transaction pattern into an alert at a defined number, measured against the borrower's own baseline rather than a portfolio average. The four that matter most are revenue compression, facility utilisation, counterparty concentration, and repayment date drift. Trazmo monitors these signals for regulated lenders in Pakistan and the GCC, and records the inputs behind every alert.
Every credit team says it watches for early warning signs of stress. Few can put a number on what actually triggers an alert. "Revenue looks soft" is an opinion. "Revenue dropped more than 25 percent for two straight months" is a threshold. Only the second one can be applied consistently, by anyone on the team, without waiting for someone's gut feeling.
Below are four signals that show up before an SME defaults, the threshold that should turn each one into an alert, and how much warning time it typically buys. Check your own review process against it. How many of these already have a real number attached?
The four-signal threshold table
| Signal | What shifts | Alert threshold | Typical lead time before a missed payment |
|---|---|---|---|
| Revenue compression | Monthly account inflows move away from the borrower's own established baseline, not a seasonal dip already priced into the file | Inflows fall more than 25% below the trailing six-month average, sustained across two consecutive months | 4 to 6 weeks |
| Facility utilisation | A borrower who rarely touched their overdraft or working capital limit starts drawing on it repeatedly | Utilisation crosses 80% of the limit on three or more occasions in a month, up from a baseline under 30% | 3 to 5 weeks |
| Counterparty concentration | Inflows that were spread across customers start depending on one or two payers | The largest single counterparty's share of monthly credits rises from under 30% to over 50% within a quarter | Variable; raises severity of the other three signals rather than firing alone |
| Repayment date drift | A borrower who reliably repaid early or on time starts repaying later in the cycle | Scheduled repayment date moves 10 or more days later than the established pattern, for two consecutive cycles | 2 to 4 weeks, the most immediate of the four |
None of these four signals are exotic. The data is already sitting in every lender's transaction history. What is usually missing is the threshold: the exact number where "worth watching" turns into "act now."
Why a threshold beats a review meeting
A monthly portfolio review runs on data that is already 15 to 30 days old. If a revenue drop started mid-month, the review meeting only sees it four weeks later. For a short-term working capital loan, that gap can be the whole window where a phone call still helps.
A threshold does not wait for a meeting. It just needs a number, compared against the borrower's own history, not a portfolio-wide average that hides the exact deviation you are trying to catch. A borrower who normally moves PKR 2M a month and one who normally moves PKR 200K should never be judged against the same fixed cutoff. Every threshold above is set relative to the borrower's own baseline. Stress for one file is a normal Tuesday for another.
That is the difference between a portfolio view that reports status and a monitoring layer that raises its hand. The first tells you what happened. The second tells you what is about to.
What four to six weeks of warning actually buys you
An early warning signal is only useful for what it lets you do next, before the only option left is a collections call.
At four to six weeks out, a relationship manager can still have a cash flow conversation the borrower can respond to. A voluntary restructuring can happen before a missed payment changes the relationship. A short top-up facility can bridge a gap before it compounds. Extra security can be requested while the borrower still has something to offer.
At zero warning, none of that is available. The payment is already missed, and the only tools left are collections and, eventually, legal action. Both cost more and recover less than acting four weeks earlier would have.
Match the signal to the portfolio
Not every signal matters the same way in every book. Revenue compression matters more for short-term working capital loans than long-term ones, where a two-month dip can be normal seasonality. Counterparty concentration matters more for service businesses running on a few client contracts than manufacturers with a wide buyer base. Facility utilisation only means something where overdraft is a common product to begin with.
Apply every threshold the same way across every loan type and the team will start ignoring the alerts within a quarter. Treat the table as a starting point. The real work is calibrating each threshold to the segment it covers, and rechecking that calibration as the portfolio changes. That calibration is the substance of an early warning system, and it is the part no vendor can hand you pre-tuned.
Build the habit before the book forces it
Put numbers on these four signals while the portfolio is still small enough that a miscalibrated threshold only costs a few unnecessary conversations. Wait until the book is large, and the same mistake either buries the team in noise or misses the file that actually mattered.
If you are setting early warning thresholds for your own book, start with the table above. Test it against six months of your own transaction data before it goes live, and adjust each number to your portfolio rather than a generic benchmark. More on how the monitoring layer fits the rest of the stack.