Start the 13-Week Before You Need It

If a US company ever reaches a debtor-in-possession facility or a fight over consensual use of cash collateral, the instrument it will be judged on is already specified. It is a 13-week cash flow budget. The American Bankruptcy Institute describes DIP budgets as built in a weekly format across 13 weeks — one quarter of a 52-week year — and used by the DIP lender to determine "whether, and how much, to provide the debtor in new loans." Counsel guidance on negotiating those budgets puts the period at "usually 13 weeks."

You may never file, and nothing here is a prediction that you will. What matters is what the terminal case reveals about the instrument: when a US credit process gets serious enough to reach a courtroom, the document the lender sizes its loan against is a 13-week cash flow forecast — and the borrower is the one expected to produce it.

Which raises the only question that actually matters: will you be building it then, or will you already have it?

What the credit agreement already tells you

You do not have to speculate about how a lender's attention behaves when confidence erodes. It is written into the loan documents.

In US asset-based lending, field examinations are the lender's standard verification tool, covering the full working capital cycle — cash, accounts receivable, inventory, accounts payable — along with payroll tax compliance. They validate collateral and the borrowing base. Credit agreement clauses commonly put them at the borrower's expense and cap their number in the ordinary course, often at one or two per fiscal year.

Then read the default clause. Representative language: if an Event of Default has occurred and is continuing, "there shall be no limitation as to the number and frequency of such field examinations." Some agreements simultaneously drop the advance-notice requirement.

Your credit agreement already contains the collapse of the evaluation horizon. One or two a year becomes unlimited, the moment trust breaks.

Note carefully what a field exam does and does not do. It verifies collateral and the borrowing base. It does not evaluate your three-year plan. Which is precisely the gap the next section is about.

The asymmetry nobody names

There is an unstated premise in most turnaround presentations: that the lender is evaluating the plan. Longer forecasts, more detailed initiative tracking, more granular EBITDA bridges — the instinct is that more plan produces more confidence.

It does not, and the reason is structural. A management-prepared three-year plan cannot be falsified inside the decision window. Nothing in it resolves before the decision has to be made, so it cannot earn belief on its own evidence. Note the asymmetry in what the lender can do about that: it can verify your collateral directly, at your expense, on demand. To get any comfort on the plan it has to hire someone to form a second opinion — which is a purchased view, not a checked fact. So the plan is not what carries the decision. Collateral value, borrowing base, and liquidity are.

A 13-week forecast has the opposite property. It is short enough that its claims resolve inside the relationship. Week 3 arrives, and either the collections landed or they did not.

This makes it the only forecasting artifact in a credit process that can accumulate credibility — and credibility, not accuracy, is what gets priced.

What lenders are actually reading

They are not reading the ending cash balance. Everyone's ending cash balance shows the company surviving; that is what makes the number uninformative. They are reading three other things, whether or not they say so.

1. Variance history. In my experience the threshold sits around six to eight weeks of forecast-versus-actual, delivered on a fixed cadence as the weeks occurred. What carries the signal is not the spreadsheet — it is the sequence of contemporaneous deliveries, which is exactly what a company cannot assemble after the fact. A company producing its first 13-week during a covenant negotiation has no such record and therefore no evidence: the forecast is an assertion.

2. Whether the variance is decomposed. Most variance reporting produces one number: forecast versus actual, by line. That conflates three failures with completely different meanings.

One caution on the first row, because it is easy to misread. Timing variance is low severity as a signal about your model. It is not low severity as a liquidity event. Liquidity is a minimum, not an average, and a collection that slips two weeks past a test date is fatal regardless of how neatly it corrects afterward. Decomposed variance tells you the model is sound; the trough tells you whether you survive. Those are two different questions and the forecast has to answer both.

A company reporting 8% aggregate variance where 7 points are timing has a working model. A company reporting 3% variance driven entirely by omissions has a broken one that happened to net out. Undecomposed variance reporting cannot tell those two apart — which is the whole reason to decompose it yourself, before someone else has to.

This is not an abstract preference. In a DIP or consensual cash collateral posture, variance is the covenant. Budget covenants are negotiated on the quantum of permitted variance — 5%, 10% — and on the basis of testing, whether per line item, on selected line items, in aggregate, or on total net cash flow. A breach typically constitutes an Event of Default under the DIP credit agreement or a termination event under the cash collateral order, which can let the secured party cut off the debtor's consensual use of cash. Variance discipline is not a reporting nicety at that stage. It is the difference between having access to your own cash and not.

3. Whether the forecast is built on balances or on history. A forecast that projects collections by applying last quarter's ratio to projected revenue is extrapolation, and extrapolation fails at exactly the moment it matters — when the underlying pattern breaks. A forecast that rolls the actual AR aging forward (End = Beginning + Billings − Collections) is anchored to a balance that exists and can be verified against the subledger, and against the same AR detail a field examiner will test. In a stressed situation the second is the only defensible construction, because the whole premise of stress is that the historical pattern has stopped holding.

The bank data discipline that separates real forecasts from theater

One methodological point, because it is where most mid-market 13-weeks quietly fail.

Bank statement data is a tie-out oracle, not a construction source. It validates aggregate ending cash after named float. It does not carry line identity, and it must never be used to build the forecast's structure.

The failure mode is seductive. A company categorizes its bank transactions, builds forecast lines from those categories, and produces something that ties to the bank perfectly. It ties because it was derived from the bank. It has no predictive content; it is a reformatted history. And it will silently misclassify, because a wire out to a vendor and a wire out to a related party look identical in a bank feed.

Structure and per-line drivers come from the accrual ledger and the P&L. The bank validates the total. An advisor who cannot articulate that distinction is producing a categorized bank statement and calling it a forecast.

Consistency beats accuracy

Total forecast error decomposes into two parts that behave completely differently: systematic error, a persistent offset in one direction, and random error, week-to-week scatter with no stable sign.

Only one of them is correctable by the reader. A forecast that runs consistently 6% high is wrong every week, but a lender can subtract the constant and still use it. A forecast that is right on average while swinging plus or minus 20% week to week cannot be corrected by anyone, because there is nothing stable to correct for. Its good-looking mean is an artifact of errors cancelling, not of the model working.

Do not chase zero variance. Chase stable, explainable variance.

A company that reports "we run 5–7% optimistic on collections, because our forecast starts the payment clock at the invoice date while our customers start it at receipt of goods, and here is the constant" has handed the lender a usable instrument. A company reporting near-zero average variance built from offsetting errors has handed them noise.

Two honest qualifications. First, a stable offset only stays useful while its cause holds; change the billing practice and the constant is gone. Second, and more often missed: six to eight weeks is enough to establish cadence and discipline, but it is a thin sample from which to quote a bias constant. If weekly scatter runs on the order of 10–20%, the standard error on a mean bias estimated from seven observations is roughly 4 to 8 points — wide enough that a claimed 6% bias is not yet statistically distinguishable from zero. A defensible constant firms up over a quarter or more, not over six weeks. Say that out loud rather than overclaiming precision you have not earned; a credit officer responds better to a range with an honest error bar than to a point estimate that cannot survive a follow-up question.

Which is one more argument for starting early. A forecast begun six months before anyone asks arrives at the conversation with enough observations to state its own bias with a straight face. One begun that month cannot.

What starting early actually costs

Almost nothing, which is the frustrating part.

A rolling 13-week is a few hours a week once the structure exists: roll the window forward one week, refresh the balances, record the variance, decompose it. The build itself is a matter of days when the team is not distracted, the data room is not open, and nobody is negotiating anything.

Compare that to the alternative. The same model built during a liquidity event gets compressed into days, assembled by a finance team that is simultaneously fielding lender calls, and delivered to an audience that has every reason to read it skeptically and no variance history to read it against. It costs more, it is worse, and it arrives without the one property that made it worth having.

Started during a negotiation, it is a document. Started six months earlier, it is a track record.

The conclusion

When a credit relationship comes under pressure, management's task is not to argue that the plan is good. It is to demonstrate that the company is governable — that someone inside knows where the cash is going and can prove it before the fact rather than explain it after.

That demonstration has a specific form: a direct-method, rollforward-based 13-week cash flow forecast, refreshed on a fixed weekly cadence, with variance decomposed into timing, amount, and omission, tied out to bank actuals.

And it has a specific timing requirement, which is the whole of the argument: it has to be running before anyone asks for it. Not because trouble is coming. Because a track record only accrues forward — and the day it becomes valuable is the day it is already too late to start one.

This article is general information, not financial, legal, or investment advice, and is not a substitute for counsel on any specific credit agreement or bankruptcy matter. Illustrative percentages are used to demonstrate method and are not drawn from any specific engagement. Sources: American Bankruptcy Institute Journal, "Taking a Fresh Look at DIP Budgeting," on the 13-week DIP budget and its role in the lender's funding decision; King & Spalding, "What Is It? Understanding and Negotiating a DIP Budget," on budget length, permitted variance quantum and testing basis, and the consequences of a budget covenant breach in the DIP credit agreement and the cash collateral order; Secured Finance Network, The Secured Lender, "Field Exams: Evolving Tools, Unchanging Purpose," on the scope of US ABL field examinations; representative credit agreement field examination provisions regarding borrower expense, ordinary-course frequency caps, and unlimited frequency following an Event of Default.