Cash Flow Operating System: A Weekly Playbook

Executive summary

A cash flow operating system is a weekly process, with named owners, that turns cash from something you report into something you manage. It has four parts: a direct-method 13-week forecast that is rebuilt every week, a standing weekly cash call that makes decisions, a record of how accurate each forecast line was, and governance rules (approvals, capex gates, minimum liquidity) that say who may commit cash and when.

The size of the prize is not a percentage. It is the amount of cash you currently learn about too late: the receipt that slipped, the supplier run that was larger than expected, the capex that was approved without anyone checking the week it would land. Each of those is a decision you could have made earlier.

The first move is small. Pick one weekly slot, one person who owns the forecast file, and one owner for each of the three big lines: receipts, supplier payments, and inventory purchases. Everything else in this playbook builds on that.

When to use this

Use this playbook when:

• Cash surprises you more than once a quarter, in either direction.
• Your forecast exists, but nobody changes behavior because of it.
• A lender, sponsor, or board member is asking for weekly liquidity reporting.
• Growth, a seasonal peak, or a restructuring is about to stretch working capital.
• Ownership of cash is split between finance, operations, and sales, and nobody can say who is accountable for the total.

Do not start here if the bank accounts are not reconciled or the AR and AP subledgers do not tie to the general ledger. A forecast built on balances you do not trust produces a confident wrong answer. Fix the reconciliations first, then return.

If you are already in a liquidity emergency, use Liquidity crisis triage first. This playbook is the steady-state system you build afterward, or before, so you never need the emergency one.

This guide deliberately does not repeat how to lay out the forecast spreadsheet. For the template and column structure, see the 13-week cash flow forecast template. Here the focus is the operating process around the forecast, and the mechanics that make it trustworthy.

Diagnostic questions

Answer these from your bank portal, ERP, and the last three forecasts.

1. Is there a forecast built from bank balances and expected cash movements (direct method), rather than derived from the income statement?
2. Is it rebuilt or rolled forward every week, on the same day, by the same person?
3. Can you say, for the last eight weeks, how far actual receipts were from forecast, line by line?
4. Was the error mostly in one direction, or random?
5. Who owns the receipts line by name? Supplier payments? Inventory purchases? Payroll and tax? Debt service?
6. Does a weekly meeting exist where someone can change a payment date, approve an exception, or escalate a customer, and are decisions written down?
7. Is there a written minimum liquidity level (cash plus undrawn committed credit) with actions attached?
8. Who can commit cash outside the payment run, and above what amount does a second approval apply?
9. Are capex requests tested against the weekly forecast before approval, or only against the annual budget?
10. Do collections promises feed the receipts line, and is their hit rate tracked?

If questions 1 to 3 fail, the forecast is the problem. If 5 and 6 fail, ownership is. If 7 to 9 fail, governance is. Most companies fail all three in different proportions, which is why this is a system and not a spreadsheet.

Framework: the four loops of a cash operating system

Think of the system as four loops that run on different clocks and are owned by different people.

Underneath the four loops sit three line owners, because most cash movement in an operating company is governed by three working capital processes:

• Receipts owner (credit and collections lead): owns the timing of customer cash. Works from the AR aging and promised dates. See How to reduce DSO and AR aging analysis and collections prioritization.
• Payables owner (AP manager or controller): owns the timing of supplier cash, including payment-run content and exceptions. See AP process design and DPO: extend payables without damaging suppliers.
• Inventory and purchasing owner (supply chain or operations lead): owns the cash committed when purchase orders are released, long before an invoice exists.

These three connect to the forecast owner and to the CFO. The point of the system is that nobody can say "the number is finance's problem." Each large line has a person whose name is next to it.

For where these three levers sit in the larger picture of working capital, see the working capital playbook.

Step-by-step: the weekly cash call

The weekly cash call is the heartbeat. Everything else is preparation for it or follow-up from it.

Format

• Length: 30 minutes, same day and time every week.
• Attendees: CFO (chair), forecast owner, receipts owner, payables owner, inventory and purchasing owner.
• Preparation (the day before): the forecast owner refreshes the 13-week forecast and the variance report and circulates both. Nobody sees the numbers for the first time in the meeting.
• Output: a written decision log, sent within an hour.

Agenda

1. Last week, actual versus forecast (5 minutes). Only lines where the variance exceeds the agreed threshold.
2. Why (5 minutes). Each line owner states the cause in one sentence: timing (it will arrive next week) or permanent (it will not come back). They are different decisions.
3. The next four weeks (10 minutes). Walk the forecast week by week. Highlight any week where projected liquidity falls below, or within a stated buffer of, the minimum.
4. Decisions (8 minutes). Pull levers deliberately: move a payment date within terms, release or hold a purchase order, escalate a named customer, draw on a credit line, defer a discretionary item. Each gets an owner and a date.
5. Weeks five to thirteen (2 minutes). Tax payments, annual insurance, bonuses, covenant test dates, seasonal inventory builds.

Rules that make the call work

• No new numbers in the meeting. Updates go into the file before the call.
• Every variance has a cause label from a short fixed list (customer paid late, early, dispute, short-pay, supplier timing, volume, price, one-off, forecast error). Fixed labels make patterns visible.
• Decisions are logged with an owner and a date. An unfinished decision is the first item next week.
• The line owners present. The CFO asks questions and decides.
• A quiet week takes ten minutes. It is never skipped.

Step-by-step: the direct-method 13-week forecast

The sheet layout is covered in the 13-week cash flow forecast template. This section covers the mechanics that decide whether the output is believable.

Why direct, not indirect

The direct method forecasts cash the way it moves: cash received from customers, minus cash paid to suppliers, employees, tax authorities, and lenders. The indirect method starts from net income and adjusts for non-cash items and working capital changes. It suits annual planning, but it hides the timing of individual movements, which is what a weekly operating forecast must manage.

Closing cash = Opening cash + Receipts − Disbursements
Liquidity    = Closing cash + Undrawn committed credit available to draw

State which version of liquidity you use, and use the same one with your lender. If availability depends on a borrowing base, forecast the borrowing base too.

Building each line

Receipts. Start from the AR aging by customer and invoice and assign an expected payment week, using in order of strength: a promised date (adjusted by that customer's record of keeping promises), the customer's own historical days-to-pay, and only then the due date. Add invoices not yet issued from scheduled billing and shipments, converted through your invoicing lag. Keep invoiced and open (high confidence) separate from not yet invoiced (lower confidence).

Supplier payments. Start from the AP aging. The forecast week is the week of the payment run on or after the due date, not the invoice date. Add approved invoices not yet entered, using goods received and open purchase orders.

Payroll, tax, debt service. These are predictable. Use actual debit dates, including variable pay on the day it is paid, tax in the week it is due (not the month), and interest and principal on contractual dates.

Capex and one-offs. Only commitments that passed the capex gate, with approved-but-unpaid shown separately from pipeline items.

Inventory purchases. If purchase orders are released well before the invoice, forecast supplier cash from PO release plus the supplier's terms. Otherwise the forecast discovers purchases weeks too late to change them.

A short illustrative roll-forward

Illustrative numbers in thousands of dollars, to show the arithmetic only. Not a benchmark, not a real company.

If minimum liquidity is 1,000 of cash, the lowest projected closing cash is 1,450, which leaves 450 of room. The cash call asks how much of that 450 is exposed to a late customer or a larger supplier run. The next section measures that from history.

Rolling, not re-creating

Each week, drop the week that ended, replace it with actuals, and add a new week 13. Save a dated copy of every week's forecast, because you cannot measure accuracy without the forecast as it stood at the time.

Forecast accuracy: variance by line and bias

A forecast is a set of claims about the future. Score it like one.

Variance = Actual − Forecast
Variance % = Variance ÷ Forecast × 100
Bias = Average of the signed variances over a window
Mean absolute error (MAE) = Average of the absolute variances over a window

• Bias says whether a line is systematically too optimistic or too cautious. It is fixable: change the assumption.
• MAE says how noisy the line is even if the average is right. Noise is managed with buffers, shorter horizons, and better information such as promised dates.

Track variance by line, never only on closing cash. Offsetting errors can leave the bottom line looking close while both lines are wrong.

Illustrative accuracy table

Illustrative numbers in thousands of dollars: six weeks of one company's receipts line.

Total variance is (440) on 7,300, or 6.0% below forecast. Bias is (73.3) per week and MAE is also 73.3, because every error has the same sign. When MAE equals the absolute bias, the whole error is bias: the method is consistently too optimistic.

Supplier payments over the same six weeks (illustrative): forecast 4,260, actual 4,400, variance +140 (3.3%), average +23.3 per week, over forecast in every week. Combined, cash ended 580 below the forecast: 440 from receipts and 140 from payments.

The receipts owner reviews the method: customer-specific days-to-pay instead of terms, promised dates haircut by promise-keep history. The payables owner checks whether bills reached the system late or the run released more than forecast. Neither fix needs a bigger model. Both need a named person looking at a table every week.

What to do with accuracy data

1. Fix bias first. If a line's bias holds one direction across eight weeks, change the assumption that creates it.
2. Size the buffer from noise. Use each line's MAE and worst single-week variance to set the headroom the call insists on above minimum liquidity. Derive it from your history, not from someone else's percentage.
3. Track accuracy by horizon. Forecasts are usually better in weeks one to four than nine to thirteen. Tell readers which weeks are firm.
4. Report accuracy beside the forecast. It is more honest, and more useful, than a forecast shown alone.

Cash governance: approvals, capex gates, minimum liquidity

Governance turns information into constraints. Keep it short and written down.

Approval limits

Write a one-page table of who can commit cash, by type and amount, with the dollar levels set from your own scale: normal-run supplier payments (payables owner, within AP policy), off-cycle payments (controller, then CFO above a limit), inventory purchase orders (purchasing owner within the approved plan, CFO beyond it), capex (per the gates below), multi-month contracts (budget holder, then CFO, then CEO or board), and intercompany transfers, loans, and guarantees (CFO, CEO or board).

Two rules protect the system: a second approver for any payment outside the normal run, and segregation between whoever changes bank details and whoever releases payment. See AP process design.

Capex gates

1. Gate 1, need and return: purpose, expected benefit, and the owner of the benefit.
2. Gate 2, cash fit: the forecast owner shows the weekly cash and liquidity effect in the planned payment weeks, including deposits and milestones, and the lowest projected liquidity week with and without the item.
3. Gate 3, approval and scheduling: the approver signs off the amount and timing, and the item enters the forecast as approved.
4. Gate 4, release: cash is released only against the milestone or delivery the approval described.

The Gate 2 test is strict: if the item takes projected liquidity below the minimum in any forecast week, it does not proceed on that schedule. The options are a different payment date, a deliberate financing decision, or a different decision. This is the one place the forecast should be able to say no.

Minimum liquidity

Minimum liquidity = Fixed obligations in the most exposed weeks
                  + Buffer for forecast noise
                  + Any lender-required minimum

Fixed obligations are the largest combination of payroll, tax, and debt service that lands in a short window. The buffer comes from your variance history. For any lender minimum, read your agreement for its exact definition; see Covenant compliance and headroom tracking.

Then attach actions to levels. Above minimum plus buffer: normal operation. Within the buffer: CFO review, hold discretionary items, escalate named receipts. Below minimum in any forecast week: a decision inside that week's call (financing, deferral, or recovery plan), and lender notice if your agreement requires it. Write the ladder in a calm quarter, not during a bad week, and agree in advance who tells the lender and by when.

Connecting AR, AP, and inventory owners to one number

The most common reason a cash forecast fails is not mathematics. It is that the three people who control most of the movement do not see each other's effect on the total. Three mechanisms connect them.

1. One shared file, three owned sections. Each line owner is responsible for their section and updates it before the call. The forecast owner consolidates and challenges. Nobody edits another owner's section.

2. A shared definition of "committed." Cash is committed when the company makes an irreversible decision, often before the invoice: purchase order release for inventory, the signed order for capex, the pay run approval for payroll. The forecast shows commitments as they are made.

3. Cross-owner trade-offs on the call. A seasonal purchase order lands in the week a large customer's receipts are expected late: purchasing and receipts decide together whether to phase it. A supplier offers an early payment discount: the payables owner and CFO compare its implied return with the cost of liquidity in the week the payment would land (see DPO: extend payables without damaging suppliers). Collections is about to hold a customer's credit: sales hears it on the call, not from the customer.

One handoff is worth building explicitly: promised dates from collections into the receipts line, and the promise-keep rate back out. Over a few weeks the forecast stops being a guess about customer behavior and becomes a measurement of it. For where this sits in the wider cash cycle, see the working capital playbook and rolling forecast vs. annual budget.

Templates

Decision log: Date | Item | Decision | Owner | Due | Status | Outcome (next call).

Variance report (one row per line per week): Week | Line | Forecast | Actual | Variance | Variance % | Cause label | Timing or permanent | Owner.

Accuracy tracker (one row per line: receipts, supplier payments, payroll and tax, debt service, capex): Window in weeks | Bias | MAE | Worst week | Direction of bias | Fix in progress.

Governance one-pager: approval limits; capex gate signers; minimum liquidity (definition, level, how derived, review date); action ladder; escalation triggers and who notifies whom.

Metrics

KPIs and how to set targets

Do not borrow a forecast-accuracy target from a blog or vendor. Accuracy depends on your customer mix, billing pattern, and seasonality. Set targets from your own history:

1. Measure for eight to thirteen weeks first, without a target.
2. Target bias before noise. Bring each line's bias close to zero so errors become random rather than systematic.
3. Set the noise target relative to your liquidity buffer. The aim is that typical error sits comfortably inside headroom.
4. Track headroom, not just accuracy. An accurate forecast of a thin margin is still a weak position.
5. Track process KPIs separately: the call happened, the file was refreshed on time, decisions were completed, every line has a named owner.

Common mistakes

1. A forecast built from the income statement. Symptom: smooth forecast, lumpy reality. Fix: forecast cash events directly.
2. Overwriting last week's forecast. Symptom: nobody can say how wrong it was. Fix: archive a dated copy.
3. Judging accuracy only on closing cash. Symptom: offsetting errors hide. Fix: score every line.
4. Using due dates for receipts. Symptom: receipts consistently below forecast. Fix: customer days-to-pay and promise-keep history.
5. Forecasting purchases at invoice. Symptom: inventory cash leaves the bank before it appears in the forecast. Fix: forecast from PO release.
6. A meeting with no levers. Symptom: every call is a status update. Fix: name the levers and log decisions.
7. No written minimum liquidity. Symptom: every dip becomes a debate. Fix: set the level and ladder in a calm week.
8. Capex approved against the annual budget only. Symptom: a large payment lands in a tight week. Fix: Gate 2 cash fit.
9. Treating the forecast as the output. Symptom: accurate file, no action. Fix: the call produces tracked decisions.

Worked example: setting up the system for an illustrative manufacturer

This example is illustrative: the company is invented, the numbers are rounded, nothing is a benchmark.

A manufacturer with a monthly cash report and no weekly process starts the system after two bad weeks. Minimum liquidity is provisionally set at 1,000 of cash. Over six weeks the team logs variance and finds the pattern in the accuracy tables above: receipts 440 below 7,300 of forecast (6.0%) and supplier payments 140 above 4,260 (3.3%), in the wrong direction every week.

The receipts owner finds the forecast used invoice due dates while several large customers paid after them, and re-forecasts on each customer's recent days-to-pay. The payables owner adds approved-but-unentered bills from goods-received data and freezes payment-run content by 2 p.m. the day before release.

The worst single receipts week was 100 below forecast and the worst supplier week 40 above. Those could coincide, so the buffer must cover at least 140. The CFO chooses 200 as a judgment, which puts the action ladder at 1,000 (minimum) and 1,200 (buffer zone), reviewed each quarter.

In week 8 a 600 equipment request arrives. At Gate 2 the forecast shows the lowest liquidity week falling to 850, below the 1,000 minimum, so the approver moves the second payment to a later milestone and logs the decision with an owner and date. The sequence is the point: measure, find the biased lines, name the owner, fix the method, set the buffer from data, and use the forecast to decide before it becomes a surprise.

AI prompt assistance

AI can prepare analysis and draft logs. It does not replace checking the numbers. Supply the data yourself, keep sensitive data in approved tools, and verify every figure against the ledger and bank.

1. Variance analysis by line.

I will paste forecast and actual cash movements by line for the last 8 weeks. For each line compute weekly variance, variance percent, bias (average signed variance) and mean absolute error. Show every step. Identify lines where all variances have the same sign. Do not explain causes; list questions to ask the line owner.

Check: recompute one line by hand and tie totals to the bank.

2. Draft the cash call pre-read.

Given this week's 13-week forecast and last week's variance report (pasted), write a half-page pre-read: weeks where projected liquidity is within our buffer of the minimum, the three largest variances with the cause labels I supplied, and open decisions from the last log. Add no numbers that are not in the data.

Check: verify every figure against the file and confirm no cause was invented.

Implementation guide: a 90-day setup plan

Days 1 to 30: Foundation

1. Reconcile bank accounts and confirm AR and AP tie to the general ledger. Owner: controller.
2. Name the owners for receipts, supplier payments, inventory commitments, payroll and tax, and debt service. Owner: CFO.
3. Build the first direct forecast from the AR aging, AP aging, payroll and tax calendars, debt schedule, and approved commitments. Owner: forecast owner.
4. Start the weekly call in week two, even with a rough forecast, and archive every forecast from day one.

Days 31 to 60: Measure

5. Track variance by line weekly with cause labels, and compute bias and MAE once eight weeks exist.
6. Fix the most biased line first, usually receipts, moving from due dates to days-to-pay and promise-keep history. Feed promised dates from the collections log (see How to reduce DSO).
7. Bring commitments forward: purchase orders and approved capex enter the forecast at release, not at invoice. Owner: purchasing owner.

Days 61 to 90: Govern

8. Write approval limits and the second-approver rule. Owner: CFO.
9. Launch capex gates with a Gate 2 cash view from the forecast owner.
10. Set minimum liquidity and the action ladder, and agree escalation triggers including who informs the lender.
11. Review the system at day 90: what decisions changed, which lines remain biased, and what the board should see.

The CFO owns the system and the minimum liquidity policy, the forecast owner owns the file and accuracy record, and each line owner owns their line. The CEO or board approves limits and sees a one-page summary: closing cash, liquidity, lowest projected week, headroom, and accuracy by line. Expect the first benefit in month one (earlier decisions), bias reductions in month two, and behavior change from governance in month three. Do not announce accuracy targets until history supports them.

Downloadable assets

The templates above are designed to be copied into a spreadsheet. The 13 week cash flow builder can help you structure the forecast itself.

Related resources

• 13-week cash flow forecast template: layout and columns for the forecast itself.
• Working capital playbook: how receivables, payables, and inventory move the cash cycle.
• How to reduce DSO and AR aging analysis: the receipts owner's tools.
• AP process design: approval, matching, and payment-run controls.
• Covenant compliance and headroom tracking: testing the forecast against credit agreement terms.
• Liquidity crisis triage: the emergency version of this system.
• EBITDA bridge and quality of earnings: linking cash to earnings, and how a buyer tests both.

Frequently asked questions

What is a cash flow operating system?

A cash flow operating system is a weekly process, with named owners, that forecasts cash directly, reviews it in a standing meeting, records forecast accuracy by line, and enforces written rules on who may commit cash. It replaces a one-off forecast spreadsheet with a routine that produces decisions, so cash is managed on purpose instead of reported after the fact.

How often should you hold a cash call?

Hold the cash call every week, at the same time, for about 30 minutes. Weekly matches the cadence of the 13-week forecast, payment runs, and payroll, so decisions can still change the outcome. In a quiet week the call can take ten minutes, but skipping it breaks the habit the system relies on.

How do you measure cash forecast accuracy?

Measure it by line, not only on closing cash. For each line and week, calculate actual minus forecast, then across a window compute the average signed variance (bias) and the average absolute variance (MAE). Bias shows whether the forecast is consistently optimistic or cautious. MAE shows how noisy the line is. Fix bias first.

How do you set minimum liquidity?

Build it from components you can name: the largest fixed obligations that land in a short window, a buffer sized from your own forecast variance history, and any lender-required minimum from your credit agreement. Then write an action ladder that says what happens as projected liquidity approaches and falls below the level.

Why use the direct method for a 13-week forecast?

The direct method forecasts actual cash events: receipts from customers and payments to suppliers, employees, tax authorities, and lenders. That shows the timing of each movement, which is what a weekly operating forecast must manage. The indirect method, which starts from net income, hides timing and suits annual planning rather than weekly decisions.

How this works in a System of Performance

A weekly cash call only compounds if what was decided, what was forecast, and what actually happened stay linked from one week to the next. In most companies those facts sit in bank portals, the ERP, collections notes, and personal spreadsheets, so every review starts by reconstructing the past. A Performis routine can take on a recurring job like this, with the scope and fit agreed with you: it works from the records your team supplies, prepares the work for your named reviewer, and keeps each approval on record. To see where your own receipts and payments diverge from plan, bring it to us as a launch partner.