Solutions / Cash-flow forecast

Cash-flow forecasting

A rolling 13-week cash forecast built from your invoices, bills and bank data, with each customer’s real payment behaviour applied instead of the payment terms on paper.

Finance Forecasting ML
13-week horizon
refreshed on a schedule
Actual payment behaviour
not contractual terms
Gaps flagged early
while there is still time to act
Who it is for

Finance teams who rebuild the cash forecast in a spreadsheet every Monday and know it is wrong by Wednesday.

Short answer

The forecast starts from your receivables, payables and bank balances, then applies how each customer actually pays — this one is reliably eleven days late, that one pays on time only when chased. You get an expected cash position by week, a range rather than a single line, and an early flag on the weeks where the balance goes where you would rather it did not.

The problem

The cash forecast assumes everyone pays on terms, and nobody pays on terms.

01

Contractual terms are a bad predictor

Net 30 means forty-two days for one customer and twenty-eight for another. A forecast built on terms is systematically optimistic, and the error is largest exactly when cash is tight.

02

The spreadsheet takes a day and ages in hours

By the time receivables are pulled, pasted and reconciled, three payments have landed and two invoices have been raised. The effort is high and the shelf life is short.

03

A single line hides the risk

One number per week says nothing about how wrong it might be. The useful question is not the expected balance but how bad the plausible bad case is, and when it arrives.

How it works

From trigger to result, step by step.

01

Connect the sources that hold the truth

Accounting system, bank feeds, invoices and bills, plus recurring commitments like payroll, rent and tax dates. Nothing useful can be built from a monthly export.

02

Learn how each customer really pays

Payment behaviour by customer and segment from your own history: typical delay, variance, how it shifts by invoice size, season and whether they were chased. This is where the accuracy comes from.

03

Forecast a range, not a line

Expected inflows and outflows by week with a plausible low and high band. The band is the point: it tells you which weeks are genuinely at risk and which merely look tight.

04

Flag the weeks that need a decision

Alerts when a projected balance falls under your threshold, with the biggest contributing invoices attached — so the response is chasing three specific customers, not a general worry.

05

Compare forecast to actual, weekly

Every week the previous forecast is scored against what happened. Accuracy is reported and the model corrects, so trust is earned with evidence instead of asserted.

Before / after

What changes on the ground.

Today, by hand
×A forecast rebuilt by hand every Monday
×Payment terms used as the prediction
×One line per week, no sense of risk
×Cash gaps noticed when they arrive
With the automation running
A rolling 13-week forecast refreshed automatically
Per-customer payment behaviour applied
A range with a plausible worst case
Alerts weeks before a projected gap
What you get

Delivered, not demoed.

Connections to your accounting system and bank feeds
A payment-behaviour model built from your own collection history
A rolling 13-week forecast with low, expected and high bands
Threshold alerts with the invoices driving each projected gap
Weekly forecast-versus-actual scoring so accuracy is visible
Built with

We build in your stack rather than moving you onto ours. The list below is what this solution most often connects to — other systems are a scoping question, not a blocker.

Python Prophet XGBoost PostgreSQL Metabase QuickBooks Xero 1C
Time to production6–10 weeks
Build priceFixed price
First stepFree mini-audit
Honest limits

When this is not the right solution.

·With fewer than a couple of years of payment history, there is not enough signal to model behaviour, and a rules-based forecast will do the same job for less.
·If most revenue comes from a handful of large contracts, the forecast is decided by three negotiations, and your CFO already models those better than any system.
·This forecasts cash; it is not advice on financing or investment decisions, and we will not present it as such.

Questions we get about this one

Materially better than terms-based forecasting, and we report the number weekly rather than claiming one upfront. Accuracy depends on your history and how concentrated your customer base is.

Bring us the process that hurts.

The mini-audit is free: we take your version of this process apart and tell you plainly whether automating it pays. If it does, you get a scope and a fixed price.