Insight
Most cash forecasts are a spreadsheet rolled forward on last quarter’s assumptions. Here is what changes when projections have to track real collection behaviour, payment timing and demand swings at the same time.
Antoine Herlin · May 2026 · 7 min read
The short version
Most cashflow projection software is a templated spreadsheet with a database behind it: you enter assumptions, it rolls them forward, and the output is only as good as the guesses you typed in. The tools that actually help finance leaders in 2026 do three things differently: they learn collection and payment timing from your own transaction history instead of using fixed DSO/DPO assumptions, they produce probabilistic projections with explicit confidence ranges rather than a single base case, and they expose a plain-language rationale for every movement so treasury can challenge, override or trust the number.
Part 1
Cash is the one number a business cannot fake its way through. Yet the 13-week cash forecast — the workhorse of treasury and FP&A — is still built, in most companies, in a spreadsheet that someone rebuilds by hand every Monday. The category has plenty of software, but most of it just digitises the same manual logic. The result is a projection that looks precise and is routinely wrong by the second week.
Three failure modes show up in almost every cash forecasting review:
A single DSO and DPO number per customer or supplier cannot capture who actually pays early, who pays 40 days late, and how that shifts with seasonality. Real collection behaviour is a distribution, not an average.
One line on the chart hides what treasury most needs: how low the balance could plausibly go, when a covenant or minimum-cash threshold is at risk, and which scenario warrants drawing on a facility versus waiting.
Cash inflow is downstream of sales and shipments, but most cash tools treat the revenue line as a manual input. When the demand or sales forecast moves, the cash projection only catches up whenever someone remembers to retype it.
Part 2
The shift since 2023 is not “a model replaced the spreadsheet.” It is that the projection can now learn directly from the raw ledger — invoices, payments, payroll runs, tax dates, bank transactions — instead of relying on the assumptions a human types into a template.
The cashflow platforms that perform well in 2026 share three traits:
Instead of one DSO assumption, the model learns a payment-timing distribution per customer segment from historical AR and AP behaviour. A customer who reliably pays on day 45 and one who pays anywhere from 20 to 90 days are modelled differently — which is exactly where the static spreadsheet goes wrong.
Rather than a single projected balance, leaders see a P10/P50/P90 cash position per week, plus the probability of breaching a minimum-cash or covenant threshold. Decisions about drawing on a revolver, delaying a payment run or timing an investment start matching actual risk instead of a point estimate.
For each week, the system can say: “projected inflow is down €120k because two large invoices shifted to next month based on this customer’s payment pattern, and a quarterly tax payment lands Thursday.” That narrative is what lets treasury trust, challenge or override the projection deliberately.
Part 3
The category is crowded and the language is converging. Every tool now says “AI-powered cash forecasting” on the homepage. These five questions reliably separate real AI-native platforms from spreadsheets with a dashboard:
1. Does it learn payment timing from my ledger, or do I still enter DSO/DPO assumptions?
If you are still typing in average collection days, the software is automating data entry, not forecasting. Real timing behaviour is learned per customer and supplier.
2. Does it return a probabilistic range and threshold-breach risk, or a single projected balance?
A point projection forces treasury to size liquidity buffers by gut. Calibrated ranges let facility, payment-timing and investment decisions match actual risk.
3. For any week, can it explain in plain language why the projection moved?
Projections that cannot be explained get overridden in a side spreadsheet — and then nobody trusts the system at all.
4. Does the cash forecast update automatically when the revenue or demand forecast changes?
Cash is downstream of sales and shipments. If the revenue line is a manual paste, your cash projection is always one stale assumption behind.
5. Is it back-tested on my historical cash position, with held-out weeks, before go-live?
Vendor accuracy claims on demo data are marketing. Out-of-sample accuracy on your own bank balances is signal.
Part 4
Two shifts are already visible in the cash platforms gaining ground with finance and treasury teams in 2026:
From weekly rebuild to continuous re-projection. The Monday-morning spreadsheet refresh is being replaced by a forecast that updates the moment a large invoice is raised, a payment clears or the sales forecast shifts. The weekly meeting becomes a review of exceptions, not a recomputation.
From isolated cash model to one connected forecast. The same platform that predicts demand and revenue increasingly predicts the cash those numbers turn into — collections, payments and the resulting balance — in a single view. Cash forecasting stops being a treasury silo and becomes part of how the whole business steers.
The throughline is the same in every category: leaders do not want a more precise-looking number, they want a number they can defend. Software that produces a calibrated, traceable, explainable cash projection will keep winning. Software that produces a confident-looking single line will keep getting rebuilt in a spreadsheet.
LucidForecast learns collection and payment timing from your own ledger and links it to your demand and revenue forecast, producing a calibrated cash projection you can defend week by week. A 30-minute discovery call is enough to see where your current cash forecast is leaking accuracy.
Book a 30-min discoveryMore research on forecasting: visit the blog, read about sales forecasting software or demand forecasting software.