पाठशाला Pathshala · संचालन Sanchālan, Operations · Lesson 14 · Build

Forecasting and the annual operating plan

Build the year from drivers a manager can own: revenue from pipeline and conversion, cost from a hiring plan by month, cash from when customers actually pay. Then reforecast each quarter by changing drivers, never by rebuilding.

Pathshala, The Founder Library · 11 October 2026 · 7 min read

Aerial view of green fields and low hills at Holkarwadi in Maharashtra with a town beyond under a cloudy sky. — Photograph: Ankit Rainloure · Pexels

A plan built from a growth rate tells a founder what they hope. A plan built from drivers tells them what has to be true, who is responsible for making it true and what the company can afford if it is not. The second kind takes a fortnight to build once and an afternoon to update every quarter.

This lesson sets out the difference between a plan, a forecast and a budget; how to build revenue, headcount and cash from drivers; the three cases a board should see; and how to reforecast each quarter without redoing the work. It assumes the company already keeps [one honest model](/library/one-spreadsheet-model-every-founder-should-build) and closes its books every month.

Plan, forecast and budget are three different things

The annual operating plan is the year as the board approved it before it began: revenue, gross margin, headcount, operating costs and cash, by month. For an Indian company the year runs April to March, so the plan is built in January and February and approved at the board meeting in March. Once approved it does not change. The forecast is the company’s best current view of the same year: actuals for the months that have closed, revised expectations for the rest. It changes every quarter. The budget is the slice of the plan each team leader is allowed to spend, by line and month. Confusing the three is the commonest planning failure. A company that edits its plan every time reality moves loses the ability to say how far it is from what it promised.

Build revenue and headcount from drivers, not from a growth rate

A driver is a number a named person can influence and report on every week. Revenue for a business-to-business company is built from the funnel: leads by source, conversion at each stage, sales cycle, average contract value, and how many reps are fully ramped in each month. For a consumer business it is visitors, conversion, average order value and repeat rate. For a subscription business it is the movement of recurring revenue, which David Skok’s SaaS Metrics 2.0 breaks into new, expansion and churned, whose sum is net new recurring revenue. Each component has an owner: marketing owns leads, sales owns conversion and contract value, customer success owns churn and expansion.

Keep bookings, revenue and cash apart from the first row. Andreessen Horowitz’s 16 Startup Metrics warns that bookings and revenue are not interchangeable: a booking is the value of a contract, revenue is recognised as the service is delivered. A twelve-month contract signed in April is one booking, twelve months of revenue and, depending on the payment terms, one cash receipt or twelve. A plan that adds them together will promise the board cash the company does not yet have.

In most startups people are the largest cost, so the hiring plan is the cost plan. Write it as one row per role with a start month, a loaded monthly cost that includes the employer’s statutory contributions, benefits, equipment and recruiting fees, and the driver the role serves. A sales hire is tied to a quota and a ramp; a support hire to a ratio of tickets or customers; an engineer to the roadmap. A role that cannot name its driver is the first one to question.

Paul Graham’s essay Default Alive or Default Dead? puts the risk plainly: hiring too fast is by far the biggest killer of startups that raise money. The plan is where that mistake is either made or prevented. Two rules help. Put every hire in the month the person will actually start, not the month the role is approved, because a plan that assumes hires on day one overstates both cost and capacity. And tie the later hires of the year to a trigger rather than a date: the fifth account executive starts when the first four have reached quota, not in October.

Cash is the line that decides

Revenue and headcount are what the board discusses; cash is what decides whether the plan is possible. Build cash from the other two: collections according to the actual payment terms and the actual days customers take to pay, salaries and statutory dues in the month they are paid, GST paid on the dates it falls due, annual software contracts paid up front, deposits and equipment. Skok shows why this matters for any business that spends to acquire customers before they pay: the cash trough gets deeper as growth accelerates, and up-front annual payment is one of the few levers that fills it. Graham’s test applies to the plan itself: with costs as planned and revenue growing at its recent rate, does the company reach profitability on the cash it has, or does it need to raise? Either answer is acceptable. Not knowing is not.

Set the hires to fifteen a quarter and watch what a confident hiring plan does to cash at month twelve. Then set the collection lag to three months and see how much of a profitable-looking plan was really waiting in receivables. Each slider is a conversation with the person who owns that driver.

Three cases and the triggers between them

A board should see three versions of the year, each with the same structure and different drivers. The base case is the plan the company commits to. The stretch case shows what the company does with more: which hires it adds and what they produce. The downside case shows the company that grows at half the planned rate, and what it cuts, in which order, to keep twelve months of cash. Each case is useless without its trigger: the observable number, read on a set date, that moves the company from one case to another. “If new recurring revenue is below ₹18 lakh a month for two consecutive months by the end of Q1, we pause the second sales pod and hold four engineering roles” is a trigger. Sequoia gave its founders a session on forecasting and scenario planning in May 2022, when the market turned; the companies that already had a downside case and its triggers written down acted in a week.

Dark low storm clouds gathering over open land at Sanand in Gujarat. — The downside case is written in clear weather. The trigger says which cloud means it is time to act on it. Photograph: Ankit Rainloure · Pexels

Freeze the plan, roll the forecast, and change drivers rather than cells.

Reforecast every quarter without starting again

The reforecast is quick only if the plan was built so that it could be. Keep drivers on one sheet and the calculations on others, so that a reforecast changes perhaps twenty inputs and nothing else. Copy the frozen plan into a forecast file at the start of the year and never touch the original. At the end of each quarter, after the books close, do four things. Replace the closed months with actuals from the ledger. Update the drivers that have moved, each with a one-line reason signed by its owner. Extend the forecast so it always shows four quarters ahead, which is what makes it a rolling forecast. And produce one page: plan, forecast and the difference, for revenue, headcount, burn and cash, with the three largest reasons for the gap. The [monthly MIS](/library/monthly-close-and-mis-report) already compares actuals with plan; the quarterly reforecast is that comparison turned into a view of the future.

A worked plan

An illustrative Pune software company starts April with thirty people, ₹8 crore in the bank and ₹40 lakh of monthly revenue growing at four per cent a month, with customers paying a month after invoice. Its first draft plan adds six people a quarter at a loaded ₹1.5 lakh a month each. The plan builder shows a month-twelve revenue run-rate of about ₹7.4 crore, fifty-four people and ₹3.25 crore of cash at year end, with net burn of ₹46 lakh a month: seven months of runway at the point the next round must be raised. The board asks for the downside case. At two per cent growth the same hiring plan leaves ₹2.7 crore and under five months. So the company keeps the first two quarters of hiring as planned and splits the last two: in Q3 and in Q4 two hires are fixed and four happen only if new recurring revenue was at or above plan in the quarter before. With that trigger, even the downside case ends the year with ₹3.4 crore and more than eight months. At the Q1 reforecast growth is running at three per cent, below plan. The trigger holds back the conditional hires, and the forecast shows forty-six people, ₹3.7 crore of cash and about ten months of runway at year end, against under six months had the original plan simply been followed.

The planning calendar

Put the year in the calendar now. January: each driver owner proposes their drivers for the coming year with the evidence. February: finance builds the three cases, and founders cut or condition the hiring plan until even the downside case ends the year with at least six months of runway. March: the board approves the base case and the triggers; the plan is frozen. The second week after each quarter closes, in July, October and January: actuals in, drivers updated, the one-page reforecast to the board. Every month: the MIS compares actuals with plan and names the drivers that moved. A company that runs this calendar twice will find the third plan takes days, not weeks.

Nothing here is legal, tax or investment advice. The worked plan and the figure are illustrative; your own cost structure, payment terms and statutory dues decide the numbers.