A chart with a course plotted in pencil, dividers resting on it

What is FP&A?

September 16, 2026 · Article · 16 min read

Sriram ChidambaramFounder & Managing Partner

FP&A stands for Financial Planning and Analysis, which is an expansion that explains nothing. The function it names is specific: deciding what the company is going to do with its money, and then holding the company to that decision. What it produces, how it differs from the accounting you already pay for, who does the work, and the honest answer on whether to build it or buy it.

Summary

FP&A decides what the company will do with its money, then holds the company to that decision. Most have the first half in a slide and neither half in a cadence.
Accounting is a record of what happened. FP&A is a claim about what will. The two reward opposite instincts, which is why one person is rarely good at both.
We build it as three pillars that are one function: RevenueOS for the revenue, MIS for the information, Pricing for the margin. They only compound together.

A founder asked us recently what FP&A actually is, and then added the part that made the question worth answering. "I keep seeing it in job titles," he said, "and I cannot tell whether it is a thing I am missing or a word consultancies use for budgeting."

It is a fair suspicion. The expansion explains nothing: every company plans and every company analyses. So the term either dresses up something ordinary, or names something specific that most Indian companies below a few hundred crore do not have.

It is the second one.

What FP&A actually is

The one-sentence version: FP&A is the function that decides what the company is going to do with its money, and then holds the company to that decision.

Both halves matter and companies usually have neither. The deciding half is a plan detailed enough that somebody can be held to it. The holding half is a cadence: a monthly comparison of promise against outcome, run by somebody who did not deliver the thing being compared.

Planning is the first half, analysis is the second, and the word "and" is doing more work than either of them.

Why this gets confused with accounting

This is the confusion that sends most people searching. The two functions sit in the same department, often report to the same person, and live in the same three spreadsheets.

Bookkeeping

is the recording of transactions: an invoice raised, a payment received, a bank line matched. It answers one question. Did this happen, and is it in the right place.

Accounting and controllership

turns the ledger into statements. Controllership adds the controls that make them trustworthy: who may approve what, how revenue is recognised, what gets reconciled and by whom. Together they answer a different question. What happened, stated correctly, in a form an auditor will accept.

None of that is optional or junior. But notice what it does not do.

Where FP&A starts

Accounting closes the month and hands over statements. FP&A picks them up and asks a different family of questions. Why is this different from the number we said. What does that tell us about next quarter. If we hire six people in October, what does March look like.

Accounting is a record of what happened. FP&A is a claim about what will.

A record can be right or wrong. A claim can only be well reasoned or poorly reasoned, and events will test it. So accounting rewards caution and precision, while FP&A rewards putting a number down before you are certain and being scored on it.

Which is why the fractional CFO conversation is usually about which of the two is missing. In our experience most founders who say they need a CFO have adequate accounting and no FP&A.

What the function actually produces

An FP&A function is not a report. It produces four things, in order.

The plan

A plan is a set of numbers the company has agreed to be measured against. It is not a forecast and not an aspiration. It is a commitment with a date on it.

Revenue of ₹40 crore is not a plan. Sixty projects at ₹20 lakh average, delivered by four teams of a hundred and sixty people, won at eighteen percent conversion from a pipeline of three hundred and thirty opportunities: that is a plan. The second version can be wrong in a specific place. The first can only be missed.

The forecast

The forecast is the plan updated for what you now know, and it must never replace the plan, or nothing is left to be accountable to. Two lines run all year: what we said in April, and what we think now. The gap is the story of the year.

The variance

A variance is the difference between plan and actual, decomposed until it names a cause. "Revenue was ₹2.10 crore below plan" is not analysis, it is a fact anybody holding the P&L already has. Here is the same miss, decomposed.

One quarter's ₹2.10 crore revenue miss, split into the two decisions that caused it

Line	Plan	Actual	Variance	What it means
Projects delivered	60	55	(5)	Five deals did not land
Average project value	₹20 lakh	₹18 lakh	₹(2) lakh	Every deal that did land was discounted
Revenue	₹12.00 crore	₹9.90 crore	₹(2.10) crore	The headline miss
Volume effect			₹(1.00) crore	Five missing projects at the plan price
Price effect			₹(1.10) crore	Fifty-five delivered projects, ₹2 lakh less each

Source: Illustrative figures, not a client. Volume effect is (actual less plan projects) multiplied by plan price. Price effect is actual projects multiplied by (actual less plan price).

Now it is two findings. Five deals did not land, a pipeline question. Every deal that did land came in ten percent under price, a discounting question. That is two meetings with two different people. The single number was one meeting with nobody.

A variance is useful in exact proportion to how specifically it names the decision that caused it.

The decision

The fourth output gets skipped, which is why companies end up with all the reports and none of the benefit. The function earns its cost only when the review ends with somebody agreeing to act differently: a price floor raised, a hiring plan deferred, a segment stopped.

What it looks like before you can hire one

A company doing ₹5 crore to ₹20 crore usually has FP&A, in the sense that somebody is doing it. It is the founder, on a Sunday, in a spreadsheet only they can open. At that size their head holds more context than any system would. But three things are missing.

The plan has no structure under the revenue line. There is an annual number, usually set by working backwards from what the round requires. When it is missed nobody can say which assumption failed, because it has no assumptions. It has a total.

The forecast has never been scored. The company forecasts but nobody checks how wrong the last six were, so it never improves and nobody knows how far to trust it. In our experience very few Indian companies below ₹50 crore have measured their own forecast accuracy.

Nobody owns the comparison. Sales forecasts sales. Delivery reports . That is structure, not dishonesty: everybody reports their own weather. The comparison needs somebody with no stake in the answer.

A quick test. Ask what your revenue plan this quarter assumed about conversion rate and average deal size. If the answer takes more than a minute, you do not have a plan. You have a target.

What it looks like once the company is large enough

Somewhere between ₹30 crore and ₹100 crore, depending on how much the company sells and to how many kinds of buyer, the spreadsheet stops working. Not because it is wrong, but because it is the only copy of the company's reasoning and it lives with the busiest person in the building.

Then the function starts to look like one. Somebody owns the plan and the comparison. There is a calendar everybody knows: close by a date, variance by a date, review on a date. The board pack, the investor update and the internal review come from one set of numbers.

And one separation matters more than the rest. The person who closes the books is not the person who analyses them. Asked for both in one week, anybody drops the analysis, every month, because only the close has a statutory date.

Three analyses that pay for the function

Ordinary analyses any competent finance person can run. The point is that most companies never have. The figures below are illustrative and describe one imagined business: an Indian B2B services company doing ₹40 crore across four delivery teams. No client is being described.

Read the P&L as percentages, not rupees

The rupee column tells you what happened. The percentage column tells you whether it is a problem. A cost that grew twenty percent while revenue grew thirty has actually fallen. Founders read rupees. The percentages are where the drift lives.

A common-size P&L for an illustrative services business doing ₹40 crore

Line item	₹ crore	% of revenue	What the percentage is for
Revenue	40.00	100.0%	The denominator, and the line most reviews discuss
Delivery salaries	18.00	45.0%	Watch at scale, and never in rupees
Other direct delivery cost	3.20	8.0%	Travel, contractors, tooling charged to projects
Gross profit	18.80	47.0%	Compare against last year, not against a target
Sales and marketing	5.60	14.0%	Rises before revenue, so read it a quarter ahead
General and administrative	4.80	12.0%	Should fall as a percentage as you grow
Leadership cost	2.80	7.0%	Often the line nobody may question
EBITDA	5.60	14.0%	What is left, and what gets multiplied

Source: Illustrative figures, not a client. Each percentage is the line divided by revenue of ₹40.00 crore.

Run three years side by side and the table becomes a diagnosis. The lines that move are rarely the lines anybody was watching. In our experience the commonest two are an administrative percentage that never fell while the company doubled, and a delivery cost worsened by a discounted segment outgrowing the rest.

Find out how concentrated your spend is

Every company believes its costs are spread across a long list. They almost never are.

Where ₹13.60 crore of controllable spend sits, in the same illustrative business

Cost line	₹ crore	Share of controllable spend	Who approves it today
Digital advertising	3.10	22.8%	Whoever runs growth, by habit
Software and cloud subscriptions	2.40	17.6%	Nobody. Renewals happen by themselves
Contractors and freelancers	2.20	16.2%	Each project lead, separately
Office, travel and facilities	1.90	14.0%	Admin, against last year's number
Professional fees	1.40	10.3%	The founder, one invoice at a time
Everything else (forty-one line items)	2.60	19.1%	Various, largely unreviewed
Total controllable spend	13.60	100.0%

Source: Illustrative figures, not a client. Delivery salaries excluded. Share is the line divided by ₹13.60 crore of controllable spend.

Two things fall out at once. The top three lines are more than half the spend, so a serious cost conversation is about three things and not forty-six. And the approval column carries the real finding: the second-largest line has no owner, because renewals continue unless stopped.

The analysis is cheap. It needs one consistently coded ledger, which is where many companies discover their source layer is not clean enough to run it. That discovery is worth the exercise on its own.

Revenue per person, by team and never blended

Revenue per delivery person by team, same illustrative ₹40 crore business

Team	Revenue (₹ crore)	Delivery people	Revenue per person
Enterprise accounts	16.00	48	₹33.3 lakh
Mid-market	12.00	52	₹23.1 lakh
Public sector and PSU	7.00	38	₹18.4 lakh
New product line	5.00	22	₹22.7 lakh
All delivery teams	40.00	160	₹25.0 lakh

Source: Illustrative figures, not a client. Revenue per person is team revenue divided by team headcount.

The blended figure is ₹25.0 lakh and no team is at ₹25.0 lakh. The enterprise team produces close to twice what the public sector team does per person, which is no criticism of that team: long cycles, heavy documentation and payment terms in months are the nature of the work.

It is a capacity question. If the company adds twenty people next year, where those twenty go is worth more than almost any other decision it will take, and this table is the only version of the question that can be argued about in a room.

Who does this work

People do search for the role, so describe it by the judgement each level owns, which is what a founder needs when deciding who to hire or whose time to buy.

The analyst

An analyst builds the model, pulls the actuals, produces the variance and prepares the pack. What separates a good one is not Excel. It is refusing to publish a number they cannot trace, on the morning the pack is due. What they own: every figure is correct and has a source.

The manager

A manager owns a business area and has to explain it. They know why the number moved before the meeting starts, and which two of the eight variances are worth twenty minutes. The judgement is editorial: the work is deciding what not to raise. What they own: the cause, and that it is the right one.

The head of FP&A

The head owns the plan and the relationship in which it gets agreed, which is adversarial in a productive way. Sales wants an achievable target, the board an ambitious one, delivery the hiring approved before the revenue arrives. Somebody must hold a number that is honest rather than comfortable.

This person also decides what the company will not do: which analysis is not worth running, where the plan stops, which variance the review ignores. A function with no head accumulates reporting until nobody reads it. What they own: the plan is defensible, and the company is held to it.

Below a certain size all three are one person, usually the founder, normal enough under ₹30 crore. What fails is three people, none of whom owns the last item.

The monthly cycle, week by week

A day in the life of an analyst is not a useful unit for a founder. The month is, because it is the cycle the function is built around, and a broken cycle defeats good people.

Week one: the close

Accounting closes the books: accruals, provisions, reconciliations, revenue recognised against delivery. FP&A waits, and spends the time on what does not depend on the close: pipeline, headcount against plan, the two schedules that are always late.

The most valuable improvement available to most Indian companies is moving this from eighteen working days to seven. A close landing on the eighteenth pushes the review into the last week of the next month, so a decision about October gets taken in late November, when most of its value has gone.

Week two: the variance

The actuals arrive and the comparison is run, decomposed along the lines that matter here. The first pass always contains two or three wrong numbers that would otherwise have eaten an hour of the review. By Friday the function should be able to state, on one page, what happened and why.

Week three: the forecast

The remaining months are updated for what the closed month taught you. If conversion has run below plan twice in a row that is not noise, and the year gets rebuilt on the new rate rather than on hope. Companies flinch here, because rebuilding honestly means telling the board something.

Week four: the review

The review is the point of the cycle and the part most often run badly. A good one runs an hour, has an agenda made of decisions rather than sections, and ends with actions carrying a name and a date. A bad one is a walkthrough of the pack.

Twelve times a year, the same shape. The consistency is not bureaucracy. It is what makes the comparison possible.

What separates a good one from a spreadsheet operator

Modelling is table stakes. Almost everyone applying can build a three-statement model. The differences that matter are these.

They know where the numbers come from. Not conceptually, physically: which system, which field, which export, and who changed a definition in March without telling anyone.
They decide what to leave out. One more scenario, one more cut, one more tab produces a model nobody can use. The good ones prune more than they build.
They can explain the number to somebody who hates it. The sales head being told their forecast is optimistic will not be persuaded by a spreadsheet.
They will put a number on it before they are certain. A range so wide it is useless is declining the job in a way that cannot be criticised.
They go back and check. Scoring your own past forecasts, in writing, and changing method where you were wrong.

The last is the real dividing line. Every other skill there can be taught inside a year. Willingness to be measured on your own claims is a temperament, and the thing to interview for.

How SRF thinks about it: three pillars, one function

Most firms sell FP&A as a deliverable: a model, a dashboard, a monthly pack. That is why so many engagements end with a folder of files and no change in how the company behaves. A function installed as a report is read twice and ignored after.

It takes hold only when a number nobody previously trusted becomes a number somebody will decide on, which is far harder to deliver than a spreadsheet. So we build it as three pillars, and do not sell them separately.

RevenueOS: the revenue side

A revenue line is only as good as the machinery underneath it: pipeline coverage, conversion, deal length and capacity, planned as numbers and governed on a cadence. It is why startups fail on process rather than market. Without it your variance says revenue missed, not which of the four broke.

MIS: the information side

One set of numbers, agreed across the company, at the altitude each audience needs, arriving early enough to act on. It sounds administrative and everything else stands on it, because a plan compared against actuals the company disputes is a debate, not a review.

Pricing: the margin side

Revenue at a price nobody governed is worth less than the same revenue at list, and the difference goes straight past the bottom line. The floor, the discount authority matrix, the gap between the price you quote and the price you keep: all FP&A questions, all filed under sales.

Look again at the variance table. The five deals that did not land are a RevenueOS finding. The ₹2 lakh given away on every deal that did is a Pricing finding. That anybody could tell the two apart is an MIS finding. One table, three pillars, one miss.

The plan, the information and the margin are not three problems. They are one question asked at three points.

Fix the reporting and not the pricing and you get excellent visibility of a margin you are still giving away. Fix the pricing and not the reporting and you set a floor you cannot tell whether anybody holds.

Build it or buy it

Now the honest part, because the answer is genuinely not always to engage somebody.

Hire, and engage nobody, if the work is continuous and there is enough of it. Above roughly ₹50 crore, or below it with several product lines, this is a full-time job. An in-house head sits in every room and hears what gets said after the meeting ends. That context is most of the job, and no outsider will ever have it.

Hire, also, if what you are missing is discipline rather than design. Some companies know exactly what their plan should contain and simply have nobody running the cycle each month. That is a staffing problem with a staffing answer.

Engage somebody if the function has to be designed before it can be staffed. Building the first plan structure, close calendar, variance format and review agenda is a different skill from running them monthly, and a few months of work rather than a permanent role.

Engage somebody if what you need is the argument rather than the arithmetic. An external party can tell a founder the plan is not achievable in a way an employee whose appraisal that founder signs usually cannot. An uncomfortable reason to buy a service, and frequently the real one.

And do neither yet if your books close on the twentieth and your definitions are not agreed. FP&A on a source layer nobody trusts produces confident analysis of numbers that were never real, which is worse than none, because it gets acted on. Fix the close and definitions first.

Our FP&A practice exists for the middle two cases. We build the three pillars above, run them alongside your team until the cadence holds on its own, and hand over. Where the right answer is to hire instead, we say so.