System_Live
Mode: Autonomous
Methodology // Financial Planning & Analysis
Driver-based FP&A: modeled from value drivers up,
not GL roll-forwards.
A general-ledger roll-forward tells you what happened. A driver-based architecture tells you what will happen when a variable moves — and proves the math when the board asks. This is how we engineer forecasts that hold under scenarios and survive an audit.
Integrity: Audit-Ready
Archive_State: Encrypted
The Problem // GL Roll-Forward
A roll-forward records the past. It cannot model the future.
Most enterprise forecasts are still last period's actuals, grown by a percentage. The general ledger is a record of what happened — extrapolating it forward assumes next quarter resembles last quarter. The data on where FP&A actually stands:
34%
of finance teams fully integrate operational drivers — units, SKUs, headcount, product lines, marketing spend — into their forecasts.
Vena 2026 FP&A Impact Report (n=431)
52%
report revenue forecast variances greater than 6% — a gap a roll-forward has no levers to close.
Vena 2026 FP&A Impact Report
46%
of FP&A time is spent collecting and validating data — the highest figure recorded in five years.
FP&A Trends Survey 2025 (n=459)
4%
of teams can produce a forecast within a day; 29% still need more than ten business days.
FP&A Trends Survey 2025
3%
of companies have strategic, operational, and financial planning fully aligned and integrated.
Gartner, FP&A Transformation
The question a roll-forward can't answer:
if this lever moves, what happens to the number?
Definition // Driver-Based FP&A
Financials expressed as a function of the variables that produce them.
A driver-based model does not grow a revenue line; it computes it. Revenue is units × price, where units decompose into pipeline × win-rate × cycle-time. Each leaf of that tree is a real-world quantity someone in the business owns and can move. A single topline target can resolve into more than 50 underlying inputs.
Driver-tree structure per KPMG, “Driver-based planning: Elevating FP&A.”
| GL roll-forward | Driver-based architecture | |
|---|---|---|
| Built from | Last period's actuals × growth % | Operational drivers, first principles |
| Answers "what happened" | Yes | Yes |
| Answers "what happens if X moves" | No | Yes |
| Scenario cost | Re-model by hand | Change one input, the tree recomputes |
| Every figure traces to a driver | No | Yes |
// Why Audit-Ready Is Not Optional
Autonomous and audit-ready are one requirement, not two.
Speed without governance is a liability. A forecast the board cannot interrogate is a forecast the board cannot sign.
Audit-ready means every figure has lineage: each output traces back through the tree to a named driver, a data source, and an assumption an accountable human approved. When the model recomputes autonomously, that governance is engineered in, not bolted on — the agent acts within bounds you set, and the math stays interrogable, never a black box.
This is the wedge. Software vendors sell a tool you operate. We design, build, and operate the architecture — at $500M–$5B complexity — and the output is a system, not a recommendation.
Engineered Guarantees
How To // Build a Driver-Based Model
Five steps from topline to operating instrument.
A condensed view of the methodology we deploy. Each step maps to a phase of an engagement — and to the FP&A practice on our services page.
- 01Map the Drivers
Decompose the topline to its drivers.
Start at revenue, cost of goods, and operating expense and break each into the operational quantities that produce it. Stop when every leaf is a number a specific owner can influence. Cross-functional input is non-negotiable — sales, operations, and HR hold the real drivers, and shared buy-in on them is what builds executive confidence in the forecast.
- 02Quantify
Define the relationships, not the growth rate.
Assign each input its weight and the function that links it to its parent. Some are arithmetic — units × price; some are behavioural and need an impact weighting. This is where first-principles modelling replaces the growth-percentage shortcut.
- 03Wire to Source
Connect the tree to the systems of record.
Wire the drivers to ERP, CRM, and the data warehouse (Snowflake, Databricks, dbt) so they update from actuals instead of manual roll-ups. 51% of teams report only moderate or limited integration between planning tools and source systems — precisely the gap that keeps forecasts slow and stale.
- 04Govern
Engineer the audit trail in, not on.
Define who owns each driver, what bounds the autonomous layer may act within, and how every output is logged. Audit-readiness is a design step taken before the first forecast runs — never a review bolted on at the end.
- 05Operate
Recalibrate it as a living instrument.
Review drivers at least quarterly: add drivers, retire dead ones, refine weights as the business changes. A driver model is an operating system for the forecast, not a frozen workbook.
Scenarios Become a Parameter
Move one driver and the tree recomputes. Scenario planning stops being a manual rebuild and becomes a single input change.
The Cycle Compresses
The forecast cycle compresses from the ten-plus days most teams report toward minutes — and the 46% of time lost to data collection is reclaimed for advising the business.
The Number Is Defensible
Every figure traces to a driver an accountable human owns. The forecast survives the board, the credit committee, and the audit. (Illustrative target, not a guaranteed outcome.)
FAQ // Driver-Based FP&A
Questions executives ask before they engage.
- What is driver-based FP&A?
- Driver-based FP&A models the financial statements as a function of operational value drivers — units, price, headcount, win-rate — rather than extrapolating prior-period actuals. It lets you ask what happens to the number when a real-world variable moves, and trace every output back to its driver.
- How is it different from a GL roll-forward?
- A general-ledger roll-forward grows last period's actuals by a percentage; it can describe the past but has no levers to test the future. A driver-based model is built from the operational quantities up, so scenarios are a single input change and every figure has audit lineage.
- What does “audit-ready” mean in an autonomous model?
- Every output traces through the driver tree to a named source, a stated assumption, and an accountable human who approved it. When the model recomputes autonomously, those bounds and logs are engineered in — so the math stays interrogable rather than becoming a black box.
- How long does it take to build one?
- It depends on the depth of the driver tree and the state of the source systems. We start with a small number of high-impact branches to prove value, then extend — rather than attempting the full framework at once.
- Does this replace the finance team?
- No. It removes the manual roll-up work — the 46% of FP&A time spent collecting and validating data — and moves accountable humans to judgment: setting drivers, approving assumptions, and advising the business.
- What systems does it connect to?
- The driver tree wires into the systems of record — ERP, CRM, and the data warehouse (Snowflake, Databricks, dbt) — so drivers update from actuals rather than manual exports.
// Commence Architectural Phase
If your forecast can't answer a scenario, let's re-engineer it.
Tell us where you're focused. Every engagement begins with a direct conversation — no intake desk, no sales funnel.
- Coverage
- North America
- Europe
- Middle East
- Asia-Pacific