Why the weekly cash report still takes three days
In most mid-market groups the weekly cash report is assembled by one person from a dozen bank portals and a ledger export. Here is where the time actually goes — and which parts of it are worth automating first.
We ask every finance team we meet the same opening question: how long does the weekly cash report take, end to end? The answer is almost never under a day, and it is usually closer to three.
Where the time goes
Broken down across the groups we have onboarded, the pattern is consistent.
- Collection — roughly 40% of the effort. Logging into each bank portal, downloading yesterday's statements, saving them somewhere with a consistent filename, and chasing the one bank whose file did not arrive.
- Normalisation — roughly 30%. Different date formats, different sign conventions, different reference fields. Every bank describes a direct debit differently, and the mapping lives in one analyst's head.
- Reconciliation — roughly 20%. Matching statement lines to the ledger, identifying what is genuinely unmatched versus what is simply timing.
- Commentary — roughly 10%. The part that is actually valuable, and the part that gets squeezed when the first three overrun.
That distribution is the whole argument. Nine tenths of the effort produces a number. One tenth produces an explanation. The explanation is what the board reads.
The hidden cost is not the hours
Three days a week is expensive, but it is not the real problem. The real problem is single-point ownership. When one analyst builds the file, only that analyst can explain it, and only that analyst can produce it. Holiday cover means either a worse report or a later one.
The second problem is reproducibility. Ask most teams to reproduce the cash report as it stood eleven weeks ago, with the assumptions that were in force at the time, and it cannot be done — the workbook has been overwritten forty times since.
What to automate first
If you are going to fix one part of this, fix collection and normalisation before anything else. They consume the most time, they require no judgement, and they are the part where errors are silent. A misclassified statement line does not announce itself; it just quietly moves cash between categories.
Reconciliation comes second. It benefits enormously from automation but still needs a human to look at exceptions, so the goal is not to eliminate the work but to reduce it to a short exception list.
Leave commentary to your team. It is the part of the report that requires knowing that the receipt from your largest customer slipped because their approver was away, and no system knows that.
What good looks like
A controller should be able to open the position on Monday morning, see a short list of exceptions rather than a blank workbook, write three paragraphs of commentary, and send it. That is a forty-minute job, not a three-day one — and the forty minutes are the part that was worth paying for all along.