Every finance director in a growing group knows the board meeting where the forecast misses again. The FP&A team reworks the model, adds a scenario and promises a tighter view next quarter.
Often the model was never the problem. The forecast starts from the actuals, and in a group with several entities those actuals carry errors that nobody in FP&A can see.
A forecast inherits every gap in the close
FP&A builds its forecast on consolidated results. Those results come from a close that pulls each entity into group numbers.
In a group that has grown by acquisition, each business arrived with its own ERP and chart of accounts. Group finance bridges the differences with mappings and spreadsheets, because the reporting date will not wait.
When a mapping slips, the actuals are wrong. The forecast built on them is wrong in the same way, however good the model on top.
More FP&A software speeds up the same error
The natural response to a missed forecast is better tooling. A planning platform, driver-based models or an AI forecasting tool all promise more accuracy.
Each of them reads the data it is given. If revenue is recognised differently in 2 subsidiaries, or intercompany balances do not match, the new tool reproduces that faster. The board sees a sharper chart of the same unreliable number.
Good FP&A consulting starts in the data model
Before changing the model or the software, read the data the forecast depends on. That means checking:
- how each entity’s accounts map to the group
- where intercompany balances are matched and eliminated
- which manual adjustments sit between the close and the forecast
- whether the main board measures mean the same thing in every business
That survey turns a vague sense that the numbers are off into a written list of risks. Each one can then be fixed, priced or accepted with the board’s knowledge.
It is how we approach FP&A consulting at Constancia. We diagnose the data model before we quote. The price is then fixed. The people who do the survey are the people who build the work, so nothing is lost in a handover.
A finance director can test the forecast’s foundation this month
There is useful work an FP&A team can do before any outside help is involved:
- trace 1 board measure from the forecast back to the entity ledgers, and note every manual step
- list the spreadsheets that sit between the close and the planning model
- compare how each entity defines revenue, margin and headcount
- record which adjustments are repeated every month
If the trace runs cleanly, the problem probably is the model, and tooling will help. If it breaks along the way, the forecast has been carrying that break for as long as the spreadsheet has existed.
The AI plan depends on the same foundation
Boards now ask finance for AI as well as a better forecast. Forecasting agents and variance commentary read the same consolidated data as the planning model.
Fixing the data model for FP&A is therefore also the first step of any finance AI plan. Doing it once, in the right order, saves the group paying for it twice.
—
Alex Townend is CEO of Constancia. Constancia is an EPM consultancy that makes multi-entity finance data AI-ready. We diagnose the data model before we quote. The price is then fixed.
Leave a Comment