A 13-week cash forecast is only as good as its last update. Built once, it's out of date within two weeks. Updated every week, with last week's forecast checked against the bank, it becomes the number the client trusts most.
The weekly update takes most firms 20 to 30 minutes per client once it's routine. This is how we run it in 13-Week Cash Forecast, and it works just as well in a spreadsheet. If you're starting from scratch, begin with How to build a 13-week cash flow forecast for a client.
The weekly update in six steps
1. Pull the inputs on the same day every week
Bank balances, all as of the same date, plus the AR and AP agings. Pick a day, usually Monday or Tuesday, and keep to it. Every week of the forecast then covers the same days, and the comparison with actual is clean.
2. Close last week
The new bank balances are last week's actual ending cash. Put actual receipts and payments next to what was forecast, category by category.
3. Explain the differences
Write a one-line note beside every category that's off by more than a dollar threshold. For a small business, $500 is a sensible starting point. Scale it up for larger clients the same way you'd scale materiality in a monthly review.
4. Roll forward
Drop the week that ended, add a new week 13, re-place collections and supplier payments from the new agings, and let recurring items carry on. If a customer keeps paying later than their average, update their days late.
5. Update the scenarios
Rerun each scenario on the new base case and check whether its lowest point or minimum-cash breach has moved.
6. Publish and keep the old version
Send the update with a short note, and keep every week's forecast exactly as it was published. Next week's comparison has to be against what you actually told the client, not a version edited afterwards.
Three kinds of difference
Every gap between forecast and actual falls into one of three groups, and each one is handled differently:
| Type | Example | What to do |
|---|---|---|
| Timing | A customer paid a week later than expected | Move it to the new week. The 13-week total doesn't change. |
| Permanent | A supplier bill was larger than the aging showed | Note it. Check whether it will recur. |
| Missed | A repair nobody told you about | Note it, and ask whether there are others like it coming. |
Timing differences are normal and mostly harmless. A pattern of them from the same customer is a sign their average days late needs updating. Missed items are the ones to worry about: a few of them in a row mean the forecast is missing a category.
A worked example
Here's the first weekly update for a fictional client, Fieldstone Joinery Ltd., on Monday, November 9. The forecast published a week earlier expected ending cash of $66,800 for the week of November 2. The bank shows $63,000.
| Category | Forecast | Actual | Difference | Type |
|---|---|---|---|---|
| Collections | $6,800 | $4,900 | −$1,900 | Timing |
| Payroll | −$14,500 | −$14,500 | $0 | |
| Suppliers | −$5,200 | −$5,950 | −$750 | Permanent |
| Rent | −$4,800 | −$4,800 | $0 | |
| Equipment repair (not forecast) | $0 | −$1,150 | −$1,150 | Missed |
| Ending cash | $66,800 | $63,000 | −$3,800 |
Three notes cover the whole $3,800:
- Collections −$1,900: "Hollis Renovations paid on Monday the 9th instead of Friday the 6th. Moved into this week."
- Suppliers −$750: "Rush order of hinges and drawer runners for the Corbett job. Not expected to recur."
- Equipment repair −$1,150: "Emergency repair to the panel saw. Are there other repairs or purchases coming that we should add?"
Rolled forward with the $1,900 moved into the week of November 9, the lowest point is still the week of November 30, now at $48,300: $8,300 above the $40,000 minimum, and $1,900 lower than last week because of the two permanent differences.
Measuring forecast accuracy
A single week's miss tells you little. Four weeks tell the client how far to trust the numbers. A simple measure is one minus the average percentage error in forecast ending cash, comparing each week's actual with the forecast made a week earlier:
| Week of | Forecast ending cash | Actual | Error |
|---|---|---|---|
| Oct 12 | $91,200 | $90,100 | 1.2% |
| Oct 19 | $78,400 | $80,600 | 2.7% |
| Oct 26 | $86,000 | $84,500 | 1.8% |
| Nov 2 | $66,800 | $63,000 | 6.0% |
| Average | 2.9% |
Fieldstone's forecast is about 97% accurate over four weeks. That's a useful line in the client update. It's also an early warning: if accuracy slips week after week, something in the method needs fixing, usually collections timing or a missing recurring item.
One-week accuracy flatters any forecast, because next week is the easiest week to predict. Every so often, also check how well the lowest point predicted four or more weeks ago held up.
Scenarios
The base case answers "what happens if things go as expected?" Scenarios answer the questions the client is actually asking. Keep it to two or three, each changing one thing:
- Late collections: the biggest customers pay later than usual.
- Financing a purchase: paying cash versus taking a loan or lease.
- A contract starting or ending: new work with up-front costs, or the loss of a large customer.
Fieldstone is considering a $46,000 CNC router, to be delivered the week of December 7. Its supplier offers financing with a $9,000 deposit and monthly payments from January. Here's the lowest point over the next five weeks in each case:
| Scenario | Lowest point | Week of | vs $40,000 minimum |
|---|---|---|---|
| Base case (no router) | $48,300 | Nov 30 | +$8,300 |
| Alder Homes and Corbett Kitchens pay 3 weeks late | $36,100 | Nov 30 | −$3,900 |
| Buy the router outright | $15,400 | Dec 7 | −$24,600 |
| Finance the router | $48,300 | Nov 30 | +$8,300 |
The decision almost makes itself. Buying outright takes cash $24,600 below the minimum. Financing keeps the lowest point where it is, with ending cash of $52,400 in the week the router arrives. Late payment from two customers is a smaller risk, but worth a reminder to chase Alder Homes early.
What to tell the client
Most clients read the first two lines of a forecast update, so put things in this order:
- The lowest point: how low cash gets, in which week, and how far above or below the minimum.
- What changed since last week: the actual vs forecast differences, in a sentence or two.
- Decisions: any scenario that needs the client to choose, with your recommendation.
- The detail: the full 13-week table and the assumptions, for clients who want to dig in.
Common mistakes
- Editing last week's forecast. Once it's published, leave it alone. Otherwise the comparison with actual means nothing.
- Bank balances from different days. Transfers between accounts end up counted twice or not at all.
- Explaining only the total. A $3,800 miss can hide a $10,000 collection surprise offset by a delayed bill. Compare by category.
- Too many scenarios. Five scenarios with small differences are harder to read than none. Keep the ones that drive a decision.
- Skipping a week. A forecast that's two weeks old has already lost the client's trust.
Frequently asked questions
What is forecast vs actual in a cash flow forecast?
Forecast vs actual compares what last week's cash forecast expected for each category, such as collections, payroll and supplier payments, with what actually went through the bank. Explaining the differences shows the client why their cash position changed and improves next week's forecast.
How do you measure cash forecast accuracy?
A simple measure is one minus the average percentage error in forecast ending cash over the last four weeks. If the forecast ending cash was off by 1%, 3%, 2% and 6% in four weeks, the average error is 3% and accuracy is about 97%.
What is a timing variance?
A timing variance is a receipt or payment that happened in a different week from the one forecast, such as a customer paying a week late. It doesn't change the total over the 13 weeks, it just moves. A permanent variance changes the total, such as a bill that was larger than expected.
What scenarios should a 13-week cash forecast include?
Keep it to two or three that answer real questions: what if the biggest customers pay late, what if the client buys or finances a planned purchase, and what if a large contract starts or ends. Show the lowest point and any minimum-cash breach for each.