Working Capital and 13-Week Cash

A 13-week forecast is really a schedule of timing, and the timing has a cause. Working-capital theory, as laid out in Brealey, Myers, and Allen's Principles of Corporate Finance, describes the cash conversion cycle: the stretch of time between paying suppliers and collecting from customers. In the usual form it is days sales outstanding plus days inventory outstanding, less days payable outstanding. The longer that cycle, the longer cash sits tied up in receivables and inventory before it comes back.

That is why a business can be profitable on paper and still run tight on cash: a working-capital swing can move near-term liquidity more than a quarter of reported earnings does. Brealey, Myers, and Allen make the same point through the cash budget, where each period's ending cash follows from beginning cash plus receipts minus disbursements. The 13-week forecast is that cash budget at weekly resolution. It does not restate profit; it schedules when cash actually lands and when it leaves.

The timing is not guessed; it is read off two schedules. The AR aging buckets receivables by how overdue each balance is (current, 1 to 30 days, 31 to 60, and so on), and it is the evidence base for collection timing: you estimate when each bucket turns into cash from how that customer base has historically paid, not from hope. The AP aging buckets payables by how soon each is due, and together with known dated items such as payroll runs, tax payments, and debt service, it schedules what leaves the account and when. A disciplined forecast, in the restructuring tradition, counts only the receipts the aging can support and dates each disbursement to when it truly comes due.

These schedules chain together. Each week's ending cash equals the prior week's ending cash, plus the collections read off the AR aging, minus the disbursements read off the AP aging and the dated items. That arithmetic has one right answer for a given set of inputs. So a forecast is not decision-ready because it looks finished; it is decision-ready when it is validated, which here means three things. The weekly chain ties, so each week follows from the prior week and the flows. The trough week is correctly identified as the true low point across the 13 weeks. And the minimum cash, the lowest balance the forecast reaches, is a figure the treasury team would stand behind. Validation comes before scenarios, because a scenario is just the base with a few inputs changed, and a stress test on a base that does not tie tends to be confidently wrong.

The five moves from Module 0 turn concrete here. The first is to map the journey. You start with the AR and AP agings and a cash skeleton, and you are headed to a validated 13-week forecast with the trough identified and two scenarios that show what could move it. Naming the destination keeps the work honest: the goal is a low-point number the treasury team trusts and a short list of levers to protect it, not a bigger spreadsheet.

Mapping also means anchoring to a worked example. If you have a prior quarter's forecast that was validated and trusted, put it in the folder as the template, so the model fills a known shape rather than inventing a layout. A forecast the model patterns on a proven example is easier to check than one it designs from scratch, and it keeps the output in a form your reviewer already knows how to read.

The second move is the split, and a cash forecast is where it does the most work. Chaining a cash forecast is deterministic work: for a given beginning balance, a set of collections, and a set of disbursements, there is one correct ending balance for each week and one correct trough. So have the model put that arithmetic into a formula or a small block of code, computed once and validated once, rather than typing figures into prose where small drift creeps in and a second, conflicting version of the numbers appears.

Then you validate the chain once. Only after the base ties does the model do its non-deterministic work: narrating scenarios on the validated base, such as a slower month of collections or a payment that slips a week, and naming the levers that protect the minimum-cash line. The sequence is build, validate, narrate, and it holds in that order. A scenario narrated before the base is validated tends to sound convincing and rest on sand.

The third move is fuel. The lab folder holds exactly what the forecast needs and nothing that dilutes it: the 13-week cash skeleton to validate, the AR aging behind the collections, and the AP aging behind the disbursements. A larger pile of treasury schedules would bury the timing that actually drives the trough and raise the chance the model anchors on something irrelevant. Minimum context is the fuel: give the model the inputs the chain depends on, and let the skeleton show the shape of the answer.

Concretely, a debt-covenant schedule or a full general-ledger export would pull the model toward line items that do not move weekly cash timing, and it could anchor the trough story on an accrual that carries no cash date. The skeleton and the two agings hold the timing that drives the low point, so they are the fuel; anything past them is ballast that dilutes the model's attention without changing the answer.

The last piece makes the workflow repeatable. Package the steps as a reusable skill or saved prompt, the standing instructions that tell the model to build and hold the chain, wait for your tie-out, and only then narrate the two scenarios. Written down once, it runs the same way each quarter instead of being re-improvised, and a teammate can run it and get the same shape of answer. Tool tiers such as Claude and its skill packs make this packaging convenient, but the discipline matters more than the tool.

The other half of the scaffolding is a validation checklist: the short, fixed list you walk before calling a forecast validated. Tie each week's chain, confirm the trough is the true low point, and check that each scenario moved the right cells rather than deleting an outflow the business still owes. You will run that checklist against your own model in Part Five. A skill that builds fast and a checklist that catches the break are what let you trust a forecast you produced in twenty minutes.

Download the folder and run the deterministic-split pattern in whatever AI you use. The folder holds Meridian's 13-week cash skeleton with beginning cash, collections, disbursements, and ending cash, plus the AR aging behind the collections and the AP aging behind the disbursements. Have the model build and hold the weekly chain, come back and validate that it ties and that the trough is right, and only then let the model narrate the two scenarios. The company and every number are fictional.

The heart of validation is the tie-out of the chain. Walk each week and confirm that ending cash equals beginning cash plus collections minus disbursements, and that each week begins where the prior week ended. Then find the lowest ending balance across the 13 weeks and confirm that week is the trough week and its balance is the minimum cash. A single week that does not chain is enough to send the base back before any scenario is trusted. Work the checklist below against your model before you would ever call the forecast validated.

Walk two links of the chain to see how one dated outflow reshapes the quarter. Week 7 begins with $3,028,000, takes in $4,545,000 of collections, and pays out $3,872,000, so it ends at $3,028,000 plus $4,545,000 minus $3,872,000, which is $3,701,000. That ending balance becomes week 8's beginning cash, which is how each link hands off to the next.

Week 8 is where the quarter turns. It begins at $3,701,000 and collects a fairly normal $4,305,000, but its disbursements jump to $6,026,000. That figure is a stack rather than one bill: roughly $3.9 million of ordinary operating outflows, the roughly $900,000 biweekly payroll that lands because week 8 is an even week, and the $1,200,000 quarterly estimated tax payment on top. Chaining it through, $3,701,000 plus $4,305,000 minus $6,026,000 leaves $1,980,000, a single-week drop of about $1,721,000 and the largest step-down anywhere in the forecast.

The lesson sits in the gap between the two weeks. The tax payment does not fall in the trough week itself; the trough is week 12 at $1,525,000, four weeks later. What the week-8 payment does is push cash out of its roughly $3.0 to $4.6 million range and into the tight $1.5 to $2.2 million range it then holds for the rest of the quarter, where the later even-week payrolls in weeks 10 and 12 carve out the actual low. Read in isolation, week 8 looks like the scare; read as a link in the chain, it is the step that sets up the trough weeks on. That gap is exactly why a validated chain earns its keep: the outflow that drives the low point and the week that records it are often not the same week, and only a chain that ties lets you connect them.

The framing below is what you would put in front of the CFO. It leads with the trough and the minimum-cash figure, names the swing factors that drive it, and ties each to what the treasury team can do.

Where it gets tight. Cash starts the quarter at $4,250,000 and draws down to a low of $1,525,000 in week 12, which is the trough. The path there is not a smooth decline; it steps down on the even weeks as the roughly $900,000 biweekly payroll clears, with a deeper step in week 8 when the $1,200,000 quarterly estimated tax payment lands on top of that week's payroll.

What it means. The week-12 low is comfortably positive, so the base case does not signal a shortfall. It does show that the cushion narrows to about $1.5 million at the trough, driven by known, dated outflows rather than by a collections surprise. That makes the trough a planning question, not an alarm: the team knows when it is tight and why.

With the base validated, the model narrates two scenarios on top of it, and each one points to a liquidity lever with a concrete action.

Scenario one: a 10 percent collections slip. If collections come in 10 percent light across the quarter, the receipts feeding each week shrink and the minimum-cash line falls below the $1,525,000 base trough, with the tightest weeks pulled around the same late-quarter window. The lever is to accelerate collections. The action it implies: work the AR aging from the top, calling the largest past-due accounts and offering a modest early-payment incentive to pull receipts into the weeks approaching week 12.

Scenario two: a one-week disbursement delay. Shifting a discretionary outflow out of the trough week into a later, higher-cash week lifts the minimum-cash line, because the payment moves rather than disappears. The lever is to delay a discretionary payment. The action it implies: identify a non-critical outflow scheduled in or near week 12, such as a deferrable vendor payment or a capital outlay, and move it to a week where the balance has more room, using the AP aging to confirm the payment has slack before its due date.

The lab is clean by design: the collections and disbursements are dated, the outflows are seeded, and the chain ties exactly. Real cash forecasts are messier, and three failure modes show up often enough to plan for.

A concentrated receivable slips. Collection timing is an estimate off the AR aging, not a certainty, and receipts are often concentrated in a handful of large accounts. If one customer that represents a third of a week's collections pays ten days late, the trough can move more than a smooth collections assumption suggests, and it can slide into a week the base case treated as safe.

An outflow lands that the AP aging never carried. The seeded disbursements cover payroll, taxes, and open payables, but a live quarter can add a supplier that switches to cash on delivery after a credit review, an insurance true-up, or a surprise tax notice. Because these do not sit in the aging, the real low point can be deeper than the forecast shows, which is why treasury tends to hold a cushion rather than plan to the last dollar at the trough.

The model narrates a confident scenario on a base that does not quite tie. A single mis-keyed disbursement or a broken beginning balance is enough to make one week stop chaining, and the AI will still produce a fluent trough story on top of it. The narrative reads well and hides the break, which is why the chain tie-out and the trough check are not optional. The workflow does not remove the treasurer's judgment; it clears the arithmetic so the judgment has more room.

Rate your own forecast against the rubric below. Marked against the rubric without flattering the result, the score shows which steps of the forecast build are secure and which need another pass. Your scores roll up to the workflow maturity dashboard on the course hub.