Your forecast isn’t wrong because its receipt assumptions are weak. It’s wrong because somebody typed the first number from the ledger and called the job started.
That first number is week zero: the dated cash position from which every receipt, payroll run, supplier payment, tax obligation, and debt payment gets built forward.
If week zero is fiction, a clean 13-week schedule merely gives the fiction nice borders.
The handoff is simple. Bookkeeping closes the period against evidence. Cash flow starts the operating forecast from that result. The point of contact is the reconciled bank balance, not the cash account that happens to sit in your general ledger.
The ledger has useful opinions. The bank has cleared money. Different jobs.
The Hierarchy
There are five cash numbers in your room. Treat them as interchangeable and your forecast misses by the same amount every week while every invoice and bill appears to be entered correctly.
| Cash number | What it establishes | What it cannot establish alone | What happens next |
|---|---|---|---|
| Bank feed | Transactions and cleared balance available from the institution on a cutoff date | Whether every ledger item has been matched correctly | Compare it with the bank statement and ledger |
| Reconciliation | The ledger balance equals the statement balance to the cent, with each difference explained | Future cash timing | Save the dated reconciliation as the close evidence |
| General ledger | The recorded cash account and its accounting entries | Whether recorded cash is cleared bank cash | Correct identified omissions, duplicates, and classifications |
| Cash-flow statement | How profit, working-capital movements, investing, and financing reconciled to cash for a period | Which coming week reaches its low point | Use it to understand the bridge, not to schedule Friday |
| Forecast opening balance | The reconciled bank figure carried into week zero | Whether later receipt and payment assumptions are sound | Build receipts and disbursements forward from it |
Recorded cash is the amount in your accounting ledger. Bank cash is the cleared balance the institution confirms. Available cash is what you can actually use after restrictions, settlement delays, frozen accounts, and segregated funds.
An undrawn credit line is potential liquidity, not cash. Listing it as cash is a charming way to make the sheet agree with your lender’s brochure.
IAS 7 keeps a related boundary. Cash equivalents are short-term, highly liquid investments convertible to known cash amounts with insignificant value-change risk, while noncash investing and financing transactions stay out of the cash-flow statement and get separate disclosure.
That’s accounting treatment, not permission to put unavailable money into week zero.
The Close
Your month-end close doesn’t begin with financial statements. It begins with the accounts that have an external source of truth: bank accounts and credit cards against statements, receivables against the aging report, payables against vendor statements, and petty cash against a physical count.
The bank part has one standard you cannot bend. The reconciled ledger balance equals the statement balance to the cent, and every difference has a specific, dated outstanding item. An outstanding check, a deposit in transit, a processor hold, a bank fee, a foreign-exchange translation, or a transaction posted across the reporting cutoff can all explain a difference. “Close enough” cannot. There’s no accepted percentage-of-balance tolerance for reconciliation.
Work your close in three passes:
- Pull the bank feed and statement for the cutoff date and identify every unmatched item. That produces the list, not the answer.
- Trace each difference to a dated cause. Record bank fees, correct duplicates, and leave genuine outstanding items visible. A plug entry isn’t a cause. It’s a trapdoor.
- Save the reconciliation once the unexplained difference is $0.00, then carry that reconciled bank figure into the weekly cash file as week zero.
That’s different from your trial balance. The trial balance lists every ending ledger account and has to net to zero, because double-entry bookkeeping posts equal debits and credits. The balance sheet also has to satisfy assets = liabilities + equity on the same date.
Both tests matter. Neither tells you whether cash is right.
A missed bank fee can preserve the balance. A duplicate payment-channel entry can preserve the balance. A transaction posted to the wrong account can preserve the balance.
Arithmetic completeness isn’t correctness. It’s the bouncer checking that everyone entered through a door.
Reconcile a low-volume account monthly against the statement cycle. Reconcile high-transaction accounts — payment processors and multi-location retail are the corpus examples — weekly or daily.
Monthly is the floor for an active operating account. Delay it through a quarter and the reconciliation becomes an investigation with a calendar problem attached.
The Forecast
Once week zero is real, your near-term forecast is a direct-method schedule. It lists expected receipts and disbursements by category and doesn’t begin with net income. For the conventional 13-week window, use named invoices and actual clearing dates wherever the information exists.
Build it in three moves:
- Reclass every open receivable by the date it’s expected to clear, not merely the stated term. The corpus example forecasts a Net 30 customer at 38 days, because that customer averaged 38 days over the last four quarters.
- Place payroll and payroll tax at fixed dates and amounts, then supplier runs, rent, subscriptions, debt service, and tax obligations when they’ll actually be released. A month is a bucket, not a sequence.
- Net weekly receipts against weekly disbursements, then show credit-line draws and term-loan repayment below operating cash. Financing can fund a gap. It mustn’t disguise the gap.
Beyond about 13 weeks, named-invoice precision weakens because many of the invoices don’t exist yet. Switch to the indirect method: start with forecast net income, add back noncash charges, and project changes in receivables, payables, and inventory. Run that longer view monthly. Exit cash at week 13 should match the corresponding month in it.
If it doesn’t, one version is using stale assumptions. Spreadsheets rarely admit this voluntarily.
Find the Broken Stage
Use the visible symptom to find your broken stage. Don’t rebuild every assumption because one cell looks impolite.
Opening forecast cash differs from bank cash → reconciliation is broken
Outstanding checks, deposits in transit, processor holds, bank fees, or cutoff items weren’t identified and dated. Reconcile the operating account to a fully explained $0.00 difference before you change any forward week.
The trial balance nets to zero but forecast cash is still wrong → the external tie-out is broken
Total debits equal total credits, but a wrong account, missed fee, duplicate entry, or uncleared item can still sit in cash. Compare the ledger with the statement. Don’t solve a bank discrepancy with a trial-balance report.
The forecast misses by roughly the same amount every week → week zero is broken
You probably started from recorded ledger cash instead of bank cash, so the original difference rolls through every later balance. Correct the opening reconciliation once rather than explaining the same variance thirteen times.
A forecast is repeatedly revised before a decision, then looks accurate afterward → the vintage is broken
The original forecast was overwritten. Preserve original forecast, latest estimate, and actual columns, then score the decision version against the actual result.
The 13-week exit balance doesn’t match the corresponding month in the longer view → the planning loop is broken
One horizon carries a stale collection, payroll, borrowing, or operating assumption. Reconcile the overlap before you treat either number as a decision number.
Month-end cash is positive, but payroll or a supplier run still can’t clear → sequencing is broken
Liquidity failure happens at the lowest available balance before an obligation settles, not at week-end or month-end. Record daily or intraday minima where large payments move the account.
The Scorecard
Don’t let the weekly update erase the forecast that supported a hire, an inventory purchase, or a borrowing decision. Keep three columns: original forecast, latest estimate, and actual. Original-to-actual measures forecast performance. Latest-estimate-to-actual measures short-term control. Original-to-latest shows how much ground the outlook covered on its way there.
Score those columns at 1 week, 4 weeks, and 13 weeks ahead. Those are measurement windows, not a ceremonial reporting ritual.
Measure your receipts and payments separately as well as ending cash. Errors can cancel in a net number while customer collections and supplier releases were both badly estimated.
Report signed error for persistent optimism or pessimism, absolute error for miss size, and ending-cash error for the liquidity consequence.
Each week you complete then feeds the next. Replace the forecast line with bank-categorized actual activity, classify the variance as a timing, amount, or assumption error, carry the correction through weeks two to twelve, add a new week thirteen, and keep the original version intact.
That’s the weekly loop.
The monthly loop isn’t separate work. A current reconciliation makes close faster. The close supplies a bank-supported opening balance.
The forecast exposes recurring timing errors, and those errors tell your next close where to look.
APQC’s survey of 2,300 organizations found a median close of six calendar days, with the top quartile at 4.8 days and the bottom quartile at 10 or more.
That’s a larger-company reference point, not a target to staple to a part-time bookkeeper. The useful measure is calendar days from period-end to a locked trial balance, tracked consistently.
A balanced trial balance proves the entries pair. A completed reconciliation proves cash has an external explanation. A cash-flow statement explains a period. A reconciled week-zero balance makes the next period forecastable.
Close against the bank. Start the forecast from that result. Don’t make week one pay for a mistake from last month.