Most PE-backed portfolio companies produce a budget once a year and call it a forecast. By March, it is a historical document. By June, it is fiction. By September, the fund's own portfolio forecast — the one presented to LPs — rests on assumptions the underlying business abandoned two quarters ago.
Deloitte's Planning, Budgeting and Forecasting survey found that approximately 75 percent of organisations consider their annual budget materially disconnected from actual business conditions by mid-year. The Association for Financial Professionals reports that while roughly half of large organisations have adopted rolling forecasts, adoption at the mid-market level remains significantly lower. For a PE fund managing five to twelve Greek SMEs, the consequence is structural: the fund's view of the portfolio is built on numbers that stopped moving months ago.
A rolling forecast is the alternative — a forward-looking projection that updates itself as new actuals arrive. It sits on Fortivis's current FP&A core: governed actuals, operational drivers, variance analysis, and rolling reforecasts. Watching these stand up inside companies that had never had one, the lesson is consistent: it is not a modelling exercise. It is an infrastructure project.
Why static budgets fail mid-market portfolio companies
The annual budget was designed for a different purpose. It is a control instrument — a spending envelope agreed at the start of the fiscal year against which actual performance is measured. As a planning tool, it has a fatal flaw: it assumes that the assumptions made in November will hold through the following December. In a stable, predictable business, that assumption is tolerable. In the Greek mid-market, where customer concentration is high, seasonality is pronounced, and external shocks — regulatory, macroeconomic, supply chain — arrive without notice, the assumption breaks within weeks.
The operational effect is predictable. The CFO knows the budget is stale but continues to report variances against it because there is no alternative baseline. The board reviews budget-versus-actual and asks why revenue is twelve percent below plan, when the real question — what does the next quarter actually look like, given what we now know — goes unanswered. The fund's operating partner receives a pack that compares the present to a past decision rather than projecting toward the future.
Gavriilidis et al.'s 2024 study of eighty Athens Exchange companies confirmed the direct link: operating cash flow management is positively associated with financial performance, while rising customer days and inventory days are negatively associated. The operational drivers matter. A forecast that does not model them cannot predict the outcomes they produce.
The architecture of a rolling forecast engine
The rolling forecast architecture has two parallel tracks: a thirteen-week cash forecast and a four-quarter margin projection. Both update continuously as new actuals arrive. Both are driver-based — structured around the operational inputs that produce financial outcomes, not around top-down percentage growth assumptions.
The thirteen-week cash forecast is not a Fortivis invention. It originated with restructuring advisors as a debtor-in-possession financing requirement and has become a PE governance standard. The version described here applies the same discipline to operating companies, then extends it with a margin projection layer over a longer horizon.
The thirteen-week cash forecast
The cash forecast models weekly inflows and outflows at the transaction-category level. Inflows decompose into collections by customer segment, with payment-behaviour assumptions calibrated to each segment's trailing DSO. Outflows decompose into payroll, supplier payments (mapped to payment terms by vendor category), tax obligations, debt service, and discretionary spend.
Each week, the forecast consumes the prior week's actual bank movements and recalculates the forward thirteen weeks. The actuals-versus-forecast comparison is not a scorecard — it is a calibration input. When collections from a particular customer segment consistently arrive three days later than modelled, the assumption adjusts. When a supplier category's payment terms shift, the outflow timing updates. The model learns from its own errors.
The critical design choice is granularity. A cash forecast built at the entity level — total inflows, total outflows — is easy to produce and nearly useless to act on. When the forecast shows a liquidity shortfall in week nine, the CFO needs to know which inflow category is compressing and which outflow category offers flexibility. That requires the same dimensional tagging at the cash level that the early warning wire requires at the P&L level.
The four-quarter margin projection
The margin projection operates on a different cadence and a different decomposition. Where the cash forecast tracks timing, the margin projection tracks structure — how revenue and cost interact across the dimensions that drive profitability.
Revenue decomposes by product line, customer segment, channel, and geography. Each combination carries its own volume trajectory, pricing assumption, and seasonality curve. Cost decomposes into fixed, variable, and step-function components, each mapped to the revenue driver that triggers it. The model expresses margin not as a single number but as a surface — the interaction of multiple drivers, each independently forecastable and independently verifiable against actuals.
Aberdeen Group research found that organisations using driver-based rolling forecasts report a twenty-four percent improvement in forecast accuracy versus those relying on financial extrapolation. The mechanism is straightforward: when each assumption is separately testable, errors are localised and correctable. When the forecast is a single top-down number, a miss tells you nothing about which assumption failed.
Operating the forecast cycle
Building the model is the smaller part of the work. Operating it — sustaining the data feeds, running the weekly update cycle, and embedding the output in the management rhythm — is where most mid-market forecast initiatives fail.
The operating cycle has four steps, repeated weekly for cash and monthly for margin.
Actuals ingestion. Bank feeds and ERP exports flow into the forecast engine. Cash actuals arrive daily; P&L actuals arrive at month-end close. Each feed is validated against the data contract — the same governance instrument that supports the alert architecture. A failed feed is flagged and escalated before it corrupts the forecast.
Variance measurement. The engine compares the prior period's forecast to the actuals that replaced it. The output is not a single variance number but a decomposition: which driver moved, by how much, and whether the movement is persistent or one-off. A collections shortfall caused by a single late-paying customer is a different signal from a systematic DSO drift across a segment.
Assumption update. The forecast's forward assumptions adjust based on the variance decomposition. This is the step that distinguishes a rolling forecast from a static model — the assumptions are living parameters, not locked inputs. The update is governed: each assumption change is logged, attributed, and visible in the forecast's audit trail.
Output distribution. The updated forecast feeds into the weekly cash position report (for the CFO and treasury) and the monthly margin outlook (for the board and the fund). The format is standardised: current-week actuals, thirteen-week forward cash, four-quarter forward margin, and the variance decomposition that explains what changed since the last cycle.
Measuring forecast accuracy over time
A forecast that is never measured against outcomes is an opinion with a spreadsheet. The accuracy measurement is built into the cycle from day one.
For the cash forecast, accuracy is measured as the mean absolute percentage error (MAPE) of the one-week-ahead, four-week-ahead, and thirteen-week-ahead predictions. The one-week-ahead MAPE establishes whether the data feeds and near-term assumptions are sound. The thirteen-week-ahead MAPE measures the model's structural validity — whether the driver relationships and seasonality curves hold over a full quarter.
In my experience, first-cycle MAPE for a thirteen-week-ahead cash forecast in a Greek mid-market company runs between eighteen and twenty-five percent — high, but expected while data contracts are being tightened. By the third cycle, with assumption calibration and data-quality improvement, the thirteen-week MAPE typically drops to eight to twelve percent. By the sixth, it stabilises in the five to nine percent range.
The margin projection follows a similar trajectory but on a longer curve — quarterly actuals provide fewer calibration points, so convergence takes two to three quarters rather than two to three months.
From static plan to living model
This is the natural extension of Fortivis's current FP&A core: turning rolling reforecasts into a living projection anchored in operational reality.
The EIF's 2024 European Small Business Finance Outlook found that one in four European SMEs reports severe access-to-finance challenges. For PE-backed portfolio companies, the forecast is not an internal convenience — it demonstrates financial visibility to lenders, co-investors, and the fund itself. A company that can project cash thirteen weeks forward with single-digit MAPE is a different credit and investment proposition from one that produces a stale annual budget.
The rolling number is not a better spreadsheet. It is a different operating discipline — one that treats the future as something to be continuously modelled, measured, and corrected rather than something to be predicted once and then compared against. The architecture is replicable. The discipline is what takes work.
Key terms
Rolling forecast
A forward-looking financial projection that updates continuously as new actuals arrive, extending the planning horizon by a fixed interval (typically thirteen weeks for cash, four quarters for margin) rather than anchoring to a fiscal year-end.
Thirteen-week cash forecast
A weekly cash-flow projection covering the next quarter, originally developed as a DIP financing governance tool and now a standard PE portfolio-monitoring instrument. Models inflows and outflows at the transaction-category level.
Driver-based forecasting
A forecasting methodology that builds projections from operational inputs (volume, price, conversion, churn, payment terms) rather than top-down financial extrapolation. Each assumption is independently testable and correctable.
Mean absolute percentage error (MAPE)
The standard accuracy metric for financial forecasts, calculated as the average of absolute percentage deviations between forecast and actual values. Lower MAPE indicates a more reliable model.
Data contract
A governance specification defining which data flows at which frequency with which dimensional tags. Shared infrastructure with the early warning wire.
Sources
- Association for Financial Professionals (2024). FP&A Benchmarking Survey. Approximately half of large organisations use rolling forecasts; mid-market adoption remains significantly lower.
- Deloitte (2023). Planning, Budgeting and Forecasting Survey, UK. 75% of organisations find their annual budget materially disconnected from business conditions by mid-year.
- Aberdeen Group. Driver-Based Planning and Rolling Forecasts. Organisations using driver-based rolling forecasts report 24% improvement in forecast accuracy versus financial extrapolation.
- Gavriilidis, K. et al. (2024). Cash flow management, performance and risk: evidence from Greece. EuroMed Journal of Business, 20(3). Panel study of 80 Athens Exchange companies confirming positive association between operating cash-flow management and performance.
- European Investment Fund (2024). European Small Business Finance Outlook, Working Paper 2024/101. One in four European SMEs reports severe access-to-finance challenges.
- Alvarez & Marsal. The 13-Week Cash Flow. Documents the 13-week model's origin in DIP financing and its adoption as a PE portfolio governance standard.
Maria Ntavou is an Analyst at Fortivis, where she develops analytical frameworks that identify operational inefficiencies and quantify improvement opportunities for portfolio companies. Prior to Fortivis, she worked in retail and logistics operations coordinating cross-functional initiatives and managing operational reporting. She holds an Integrated Master's Degree in Applied Mathematics and Physical Sciences from the National Technical University of Athens, specialising in Analysis and Applied Statistics.