Thought Leadership
.png)
Jul 24, 2026
4 minutes
Two managers hold positions in the same borrower. Same credit agreement, same covenant package, same quarterly reporting obligations. One begins repositioning in month four. The other remains fully constructive in month eleven, when the borrower's auditors raise a going concern question and the conversation shifts from relationship management to workout.
The usual explanation is talent. The first manager had a sharper analyst, deeper sector knowledge, a more skeptical investment committee.
That explanation is sometimes correct, but it is incomplete. It overlooks a simpler possibility: the first manager did not reason better, they saw sooner. And what they saw was not in the financials.
By the time deterioration reaches a leverage ratio, it has usually been visible elsewhere for months. Credit metrics are a lagging summary of a business. Servicing behavior is a continuous read on it.
Consider what crosses a loan administration desk in the quarters before a credit goes sideways:
None of these are credit events. Each has a reasonable explanation, and the borrower will offer one. But they are behavioral, they are early, and they reach the servicing layer well before they reach the investment committee.
Managers who spot risk earlier tend to be the managers whose servicing layer is instrumented well enough to register them.
There is a structural constraint here that is easy to overlook: you cannot detect risk faster than you close your book.
If portfolio data is assembled at quarter end, quarter end is your resolution limit. Everything that occurred in the intervening ninety days is compressed into a single observation. A borrower who paid on day 1 in month one, day 12 in month two, and day 29 in month three appears identical, at quarter end, to a borrower who paid on day 1 every time. Both are current. One is signalling something.
This is why risk detection in many portfolios arrives in bursts. Deterioration does not occur in bursts. Observation does.
Managers who detect earlier have typically made portfolio state continuously available rather than periodically assembled. Not because continuous data is inherently better, but because the signals that matter are changes in pattern, and a pattern change is not visible at a sampling rate of four per year.
The second constraint is granularity.
Most portfolio reporting is built for the audiences that consume it. LPs, the investment committee, and the board want aggregates: weighted average yield, weighted average leverage, sector concentration, watchlist count. Those are the correct outputs for those readers.
They are the wrong inputs for risk detection.
Aggregation is a lossy compression of the portfolio, designed to discard the position level texture where early warning lives. A watchlist with three names identifies the credits that have already been escalated. It says nothing about the credits where something has begun to change but no one has yet built a case strong enough to escalate.
Early detection happens in that gap, between nothing to report and formally concerning. A manager operating only on aggregates has no instrumentation in that space. The information exists, but it sits in an operations analyst's inbox and in that analyst's memory, and it surfaces only if the same person connects two events six weeks apart and raises it with someone who has authority to act.
That is a process with a single point of failure, and the point of failure is a person's recall.
The third constraint concerns where the authoritative book actually lives.
In many private credit shops, it is not in the loan management system. It is in a spreadsheet maintained in parallel, because the system of record cannot accommodate the amendment, the multi currency tranche, the bespoke fee structure, or the non standard waterfall. The real numbers are kept somewhere the operations team can control them.
This is shadow booking, and its direct costs are well understood: reconciliation risk, key person dependency, limited audit trail.
Its second cost receives less attention and is the one relevant here. A book that must be reconstructed at period end cannot be monitored between periods. There is no continuous state to observe, only a recurring assembly exercise. Between assemblies, the portfolio is not slow to read. It is unreadable.
Managers in this position are not choosing to detect risk late. They made an infrastructure decision, often years earlier and often for defensible reasons, that makes early detection structurally unavailable.
The case for early detection is often made in general terms about protecting the portfolio. It is worth being more specific.
Early detection does not prevent deterioration. Borrowers deteriorate for reasons unrelated to their lender's data infrastructure. What early detection produces is optionality, and in credit, optionality is priced in time:
Amendment leverage. Negotiating in month four, while the borrower retains liquidity and alternatives, produces different economics than negotiating in month eleven, when it does not.
Exit windows. Secondary bids exist for a performing but drifting credit. For a credit already in technical default, the bid list is shorter and the pricing is worse.
Reserve and mark timing. Recognizing deterioration on your own schedule rather than in response to an event is the difference between a managed mark and a surprise.
LP credibility. An LP who hears about a problem from the manager, early, with a plan attached, responds differently than an LP who hears about it from a valuation committee.
Treat servicing data as risk data. Payment timing, draw behavior, reporting punctuality, and waiver frequency should be captured as structured fields rather than as email threads. This requires treating operations as a monitoring function rather than a back office function.
Monitor rate of change, not level. A borrower at 4.2x leverage is not interesting on its own. A borrower that has moved 0.3x in each of the last three quarters while reporting has slipped by a week each cycle is interesting at a leverage level no threshold would flag.
Maintain one book. A portfolio held in one place can be queried continuously. A portfolio distributed across five places can only be assembled.
Shorten the distance between operations and investment. The person who notices the fourth consecutive late payment and the person with authority to act on it should be the same person or one conversation apart, rather than separated by several handoffs and a monthly meeting.
Much of the variance in risk detection across private credit managers is not variance in credit judgment. It is variance in infrastructure, specifically in the servicing and administration layer that the industry has generally treated as a cost to be minimized rather than an intelligence layer to be built.
That treatment was reasonable when portfolios were smaller, structures were simpler, and quarterly was fast enough. It is harder to defend against more complex capital structures, more bespoke terms, and LPs who expect answers on a timeline that periodic assembly cannot support.
The managers pulling ahead are not the ones who hired better analysts. They are the ones who stopped treating loan administration as paperwork and started treating it as the earliest available read on portfolio health.
.png)
.png)
