How to run FP&A month to month
The monthly loop - close, compare, explain, communicate, decide, re-forecast - and which half of it should be automatic.
# The short version
FP&A is not a report you produce. It is a loop you run every month, and its value is not in the document - it is in what changes because of it.
- Close fast, then report immediately. Numbers that land on day fifteen are archaeology, not management information.
- Compare against two things: the plan and the prior year. They answer different questions and you need both.
- Look past the P&L. Margin can beat plan while cash quietly gets worse.
- Know who is reading. Management wants drivers and decisions; shareholders want trajectory and cash.
- End every review with a decision that has an owner and a date. Agreement is not an outcome.
- Re-forecast, then freeze again - and run exactly the same loop next month.
The point of FP&A is not to explain last month. It is to change next month.
- 1 Close Get the numbers in
- 2 Compare Plan and prior year
- 3 Explain Why it moved
- 4 Communicate To the right audience
- 5 Decide Owner and date
- 6 Re-forecast Then freeze again
# A system, not an analysis
Most finance teams can produce a good analysis when someone asks for one. Far fewer run a system: the monthly loop, plus the structure and the record that make this month comparable to the last. An analysis answers one question once. A system answers the same questions every month, whether or not anyone asked.
The difference matters because insight comes from repetition. A single month's variance is noise. The same variance three months running is a trend, and you can only see it if you measured the same way each time.
- The same structure every month, so lines stay comparable. That is what designing your statements properly buys you.
- A frozen plan to measure against, not one quietly edited to match reality.
- The same comparisons in the same order, so nobody re-argues methodology instead of the business.
- A fixed date, so decisions land while they still matter.
- Preparation cheap enough that a busy month cannot kill it.
If the repeating half takes three days of manual work, it gets skipped the first busy month - and then it was never a system. Automate the preparation or accept that it will not survive.
# The monthly loop
Close, compare, explain, communicate, decide, re-forecast. Six stages, and the first two are where most teams lose the value of the other four.
Close fast, then report instantly. A slow close and a slow report are two delays stacked on top of each other, and by the time the numbers arrive the month they describe is history.
Teams often optimize the wrong one. They spend an extra week perfecting the close, then another week building the report by hand. A close that is 98% right on day three beats one that is 100% right on day fifteen, because the first one can still change a decision.
A fast close is mostly about spotting gaps quickly. Seeing the month next to the plan and the prior year makes a missing invoice or an unbooked cost obvious in a way that scrolling a trial balance never does - then you either chase it, or you accrue for it and move on. That is what makes finishing on day three realistic rather than reckless.
It helps to be honest about which half of the month is mechanical and which half is the job.
| Stage | Prepared - should be automatic | Judged - this is the work |
|---|---|---|
| Close | Import actuals, map new accounts, consolidate entities | Reviewing the numbers for gaps - a missing invoice, a cost that has not landed - and booking the accruals or provisions that close them |
| Compare | Variance against plan and prior year, on every line | Which variances matter and which are noise |
| Explain | Bridges, KPI refresh, department splits | Why it happened - the part no system can produce |
| Communicate | The view itself, in the right shape per audience | What to say, and what to leave out |
| Decide | A record of what was agreed, carried into next month | Choosing to act, to accept it, or to call it noise |
| Re-forecast | Rolling the plan forward on its drivers | Which assumption was wrong, and what it should now be |
Everything in the middle column is the same work every month. If you are still doing it by hand, that is where your month is going - and it is not the half anyone values you for.
# Two comparisons, two questions
Variance against plan and variance against prior year are not two versions of the same number. They answer different questions, and reporting only one of them hides half the picture.
| Comparison | Answers | Fails when |
|---|---|---|
| vs plan | Did we do what we said we would? | The plan was weak - beating it proves nothing |
| vs prior year | Is the business actually growing? | Last year was unusual - one-offs distort the base |
| vs prior month | Is something changing right now? | Seasonality makes it meaningless on its own |
A department can beat a padded budget while going backwards against last year. Look at both or you will reward the wrong thing.
Decide up front how big a deviation you investigate - a percentage, an absolute amount, or both. Without a threshold every month becomes an argument about which lines deserve attention, and the loudest person picks.
- Explaining every variance. Most are noise, and treating them equally buries the two that matter.
- Comparing against a plan that has been edited since it was set. That is not a variance, it is a circular argument.
- Reporting percentages without the absolute. A 40% miss on a small line is usually not the story.
# Look past the P&L
The P&L tells you what happened. The balance sheet tells you whether it is sustainable, and it is usually where the early warning sits - a month can beat plan on profit and still be the month the business got harder to run.
- Are customers taking longer to pay? DSO creeping up is a revenue-quality signal long before it is a cash problem.
- Is inventory building faster than sales? That is cash going onto shelves.
- Are you paying suppliers later than you used to? Sometimes a negotiation, often a symptom.
- Is headcount growing faster than revenue? Fine deliberately, dangerous by accident.
- Is the margin change price, volume or mix? Three different problems with three different owners.
Track a handful of these every month, in the same place, with the same definition. A KPI that appears only when someone asks for it has no trend, and the trend is the whole value.
# Know who is reading
The same month produces different reports for different readers. Not different numbers - different levels of detail, and a different question answered.
| Reader | Wants to know | Leave out |
|---|---|---|
| Management | What moved, why, and what to do about it - by department, with owners | Anything they cannot act on this month |
| Shareholders and the board | Trajectory, cash runway, whether the plan still holds | Line-level detail; they will ask if they want it |
Detail is not rigor. A view nobody finishes reading has failed, however complete it is.
Send a dashboard, not a document. A document is a snapshot that is already wrong by the time it is opened, and it forks the moment two people annotate their own copy. A live view everyone opens is one version of the numbers, and it is still right next week.
Keep the layout still. Management should open the same view every month and know exactly where to look - same metrics, same order, same position on the page. Rearranging it is expensive: everyone has to re-learn where their number lives, and the ones who do not simply stop opening it.
- Lead with the three things that changed, not with the statements.
- Say what you recommend, not only what happened. Finance that only reports is a cost center.
- Write the commentary into the dashboard, beside the number it explains. A variance with no sentence next to it gets explained out loud again every month.
- Build a view per audience rather than one view with rows hidden. The board is asking a different question from your operations lead - give it different content, not less of the same.
- Add what they actually ask for. If the same question comes up twice, it belongs on their view permanently.
# End with a decision
This is where most FP&A quietly stops. The review happens, everyone agrees why margin fell, the meeting ends - and nothing about next month is different. The analysis was correct and worthless.
Every material variance ends one of three ways: we act, we accept it and change the forecast, or we decide it was noise. Say which, out loud, with a name and a date against it.
- Act - someone changes something in the business. Cut the spend, change the price, pause the hire.
- Accept - the assumption was wrong and the plan moves. That is a re-forecast, not a failure.
- Ignore - it was noise. Legitimate, as long as it is a decision rather than an omission.
Write the outcome next to the number, where it will be read again next month. A variance that was explained twice with no decision in between is a variance nobody owns.
- Ending a review with 'we will keep an eye on it'. That is the absence of a decision wearing a decision's clothes.
- Taking the action yourself because it is faster. The point is that the owner owns it.
- Re-explaining the same variance three months running. If it recurs, it is not a variance - it is the new normal, and the plan is wrong.
# Keep the system running
The loop only compounds if it survives contact with a busy quarter. Three things keep it alive.
Freeze the plan and re-forecast on a rhythm rather than editing the original - the discipline is set out under re-forecast, do not defend. Without a frozen baseline your variances are measured against a moving target, which is the same as not measuring them.
Push the explaining out to the people who own the numbers. Planning by department is what makes that possible: the sales lead explains sales, and finance stops being the only person who can account for a number they did not create.
And treat the review as a conversation rather than a delivery. The value of doing this every month is that you build the context to ask a good question - why did that customer slow down, is that hire actually landing - which is the difference between a finance function that reports and one operations actually wants in the room.
Run the same loop every month, even in the months when nothing interesting happened. Especially those - that is how you find out early when something has.