Arcvue
Lender access—for direct lenders and commercial banks financing GovCon
The covenant that drifted was visible in the data the whole time. Nobody was looking.
Arcvue computes the covenant metrics a credit agreement already names—leverage, fixed-charge coverage, whatever was negotiated—against the borrower’s own ledger, recording whether each test ran on a closed month or on a projection. This page follows Arcvue reporting one quarter to a lender, through nine steps, for an invented borrower.
Specimen Systems LLC · an example 120-person contractor · all figures synthetic
The quarter opens with everyone agreeing about the number, which happens once every ninety days.
Nothing is inherited yet. Leverage is 1.19× against a maximum of 3.50×, the last package has been received and tested, and the borrower and the lender hold the same view of the credit.
That agreement is the high-water mark of the relationship and it degrades from here. Not because anyone behaves badly—because the borrower's business keeps happening and the lender's picture of it does not.
What this page counts. Not whether the covenant breaches—on these numbers it does not, at any point. The gap between when the data knew and when the lender did. That gap exists on performing credits too, which is exactly why it goes unexamined.
Synthetic. An invented borrower. Note what did NOT happen: no breach, no concealment, no late filing. The covenant held the entire time—and the lender still spent a third of a year unable to say so.
On day thirty-four a recompete is lost, and the ledger reflects it that night.
The agreement from stop 01 stops being true. A contract representing 20.0% of the year's revenue target does not renew—$6,868,800.00 of revenue and, once the direct labor and fringe that leave with it are taken out, $2,349,129.60 of earnings. The overhead and G&A pools do not leave; they reallocate onto everything that remains.
Leverage moves from 1.19× to 3.07×. It does not breach—the maximum is 3.50×—but the headroom that was 2.31× on day one is 0.43× on day thirty-four. That is the point: this is not a distress story, it is an ordinary quarter in a performing credit. And the ledger has the new number before anyone has gone home.
The lender will find out somewhere between day one hundred and thirty-five and day one hundred and fifty.
The number from stop 02 exists, is correct, and is sitting in a system. The quarterly covenant package arrives 45 to 60 days after quarter end—if it arrives on time—which puts the lender's knowledge somewhere between day 135 and day 150 of an event that happened on day 34.
| Day | What the ledger holds | What the lender holds |
|---|---|---|
| 1 | Leverage 1.19× | Leverage 1.19×—agreed |
| 34 | Leverage 3.07×, recorded that night | Leverage 1.19×—now wrong |
| 90 | Leverage 3.07×, quarter closes | Leverage 1.19×—still wrong |
| 135–150 | Leverage 3.07× | Leverage 3.07×—agreed again |
Screen Arcvue covenant computation, against the actual credit agreement thresholds. Read this one row by row. The four rows are moments in one quarter, not components of a sum. Basis the leverage figures are the deal page's and the loss scenario is the FP&A page's. Nothing was invented for this page. Synthetic sample.
There is no disclosure failure in this story, and that is what makes it worth telling.
The borrower did nothing wrong at any point.
The natural reading of stop 03 is that somebody should have called. But the borrower is not obliged to report intra-quarter, the covenant never breached, and a contractor who telephoned their lender about every performing-credit movement would be doing something nobody asks for.
The problem is the cadence, not the conduct. Quarterly reporting was designed when producing a covenant certificate genuinely took six weeks of somebody's work. That constraint is gone—the computation now happens nightly whether anybody looks or not—but the cadence it justified has outlived it.
And this is the performing case. The same 101-to-116-day blindness applies to a credit that is genuinely deteriorating, where the lender's options narrow with every week they do not know. The covenant that breaches was drifting in the data for a quarter first—and the drift, not the breach, was the actionable moment.
A covenant package is a photograph, and it arrives six weeks after it was taken.
Stop 04 established that nobody is at fault. This stop is about what the artifact actually is. A covenant certificate tests compliance as at a date—it says nothing about the eighty-nine other days, and by the time it lands the business has moved on for another six weeks.
Which means a lender managing a portfolio of these is holding, at any moment, a set of photographs of different ages, none of them current, and reconciling them against a mental model that is doing most of the real work. The better lenders already know which borrowers run on institutional discipline and which are managing by feel. What they lack is a way to see it rather than infer it.
The alternative is not more access. It is exactly the agreed access, continuously.
The photograph at stop 05 wants to become a live view of the metrics the credit agreement already names—leverage, fixed-charge coverage, whatever was negotiated—computed nightly and shared by the borrower's own choice.
Not a data dump, and not a window onto the general ledger. Scope is a property of the reader. Arcvue carries a read-only external role with pipeline, pricing, business development and the admin console denied outright, and an outside reader is confined to a single engagement's rows. A lender seat is scoped the same way: the metrics the agreement names, on the engagement they are on, and nothing else.
Thresholds are configured per agreement, the model computes against them, and every observation records whether it was tested on a closed month or on a projection—a forecast breach must never read as one that happened, and that distinction is the difference between a covenant conversation and a covenant problem. The lender reads the engagement they are on and nothing else.
Everything so far has been backward-looking, including the live version.
A current covenant metric is still a measurement of what has happened. The question a lender actually carries is whether the borrower can service the facility over the next quarter, and that is a forward question.
The thirteen-week position is built from observed behavior—actual collection patterns, payroll cadence, vendor payment history, the real debt service schedule—rather than from a projection the borrower prepared. It closes at $2,086,400.00 on these numbers. A forecast the borrower did not build for you is a different kind of evidence from one they did.
The metrics that monitor a credit are the ones that should have underwritten it.
Stops 06 and 07 describe a credit that already exists. The same figures decide whether it should—and GovCon underwriting needs a specific set that generic credit analysis does not produce: backlog against funded value, recompete concentration, indirect rate stability, the split between funded and unfunded revenue.
Those are exactly the things Arcvue computes in the ordinary course. A borrower who can produce them at origination is demonstrating the same discipline that will make them legible for the life of the facility—and the ones who cannot are telling you something too.
A borrower who can be seen is not a better borrower. They are a differently-priced risk.
And stop 01's agreement was only ever true for a day.
The quarter opened with both parties holding the same number, and this page has been about the eighty-nine days afterward when they did not. That gap is priced—into covenant headroom, into reporting requirements, into how quickly a lender moves when something looks wrong, and into the rate.
Closing it does not make anybody trust anybody. It removes the thing that trust was standing in for. A lender who can see leverage on any given Tuesday is not extending goodwill; they are extending credit against an observable position—and the borrower stops paying a premium for uncertainty that was never theirs.
Which is why this is worth the borrower's while. Live access sounds like a concession and is closer to the opposite: the 101 to 116 days of blindness at stop 03 were being charged for, and the borrower was paying. Arcvue closes those days, and the saving is the borrower's.