How often should you update the budget? Less often than you think

The previous guide ended with a practical problem. A budget built in January rarely reaches the owner at the moment of a decision in September. One common response is to make the budget move: re-plan quarterly, or monthly, or every time something material changes, so the plan the owner consults is never far from what is actually happening.

This guide argues that the response is borrowed from a setting where it comes with safeguards, and applied to one where it usually does not. Large companies re-plan through the year, but their planning systems keep budget, actual and forecast as separate things and preserve every approved version of the plan. The problem is not rolling forecasting. It is allowing each forecast update to overwrite the approved budget without preserving the original baseline, which is what re-planning tends to look like in a spreadsheet.

The short answer

Review budget-to-actual monthly. Update the forecast whenever material information changes, normally at least monthly. Revise the approved budget only when the business deliberately changes its plan, and preserve both the original and revised versions.

A budget overwritten whenever reality changes stops being a useful baseline. Preserve the original budget, update the forecast, and create a dated new plan only when the business deliberately changes its intent. Then test each commitment against both the plan and the current forecast when it arrives. We call this continuous feasibility: keeping the plan visible while testing each new commitment against the current forecast. It is easy to confuse with continuous re-planning, because both involve looking at the numbers more often than once a year, but they are different disciplines.

Why variance needs a fixed point

The first guide in this series set out four things that are often run together: the budget is what the business intends, the forecast is what it currently expects, a scenario is what a proposed decision would change, and feasibility is whether that scenario fits.

The only thing that connects the budget to the forecast is variance. Variance is the difference between what was intended and what is now expected or has already happened. It is the signal that something in the original reasoning did not hold. That signal is only legible if one end of the comparison stays still. If the plan moves every time the forecast moves, there is no longer a difference to read. There are two numbers that agree, and no way to say what the business learned.

This is why a budget and a forecast are kept as separate things. The forecast has to change, because the business changes. The original budget stays fixed as the baseline. If the business deliberately changes its plan, preserve the revised budget as a dated version alongside it. A plan that is overwritten with each forecast update has been asked to do the forecast’s job, and once it does, the business has a forecast and no plan.

Where continuous re-planning fits

None of this means re-planning through the year is wrong where it developed. Rolling forecasts and frequent re-planning came out of corporate finance as a response to genuine problems with the annual budget at scale: it took months to build, it was stale by the second quarter, and the negotiation that produced it encouraged people to pad their numbers. Those criticisms are fair. In our reading, re-planning through the year is safe and useful when three conditions hold, and they are worth stating as a framework rather than as facts about every company.

The first is that spending is delegated. A company with dozens of budget holders uses the plan as permission to spend. Each department has an allocation and can commit against it without asking. When conditions change, the only way to move money from a stalled initiative to one that is working is to change the plan, because the plan is what authorises spending. Re-planning is a reallocation mechanism. Without it, money sits in the wrong place for the rest of the year.

The second is that commitment happens in tranches. Large companies release funds in phases, review at gates, and hold back what has not yet been committed. Revising the plan between phases changes what gets released next, which is a meaningful lever.

The third is version governance. When a large company re-plans, a new version is approved, the prior version is kept, and variance is reported against the version that was in force at the time. Planning systems are built around this: budget, actual and forecast are held separately, and each approved budget version is retained (Oracle’s documentation on rolling forecasts and on comparing budget versions describes the pattern). The plan moves, but a fixed point is preserved at every step. Nobody loses the ability to say what was intended, because every intention is filed.

Where those three conditions hold, re-planning is a reallocation mechanism with a preserved baseline. Where they do not, it is something else.

Why it does not transfer

Many owner-led businesses in the one to ten million range do not have all three in formal form. Some have managers with spending authority, staged commitments and a saved copy of every plan version, and for them the practice transfers with fewer problems. For the rest, the picture below is closer to what practitioners described to us.

Spending is often not delegated. The owner is the allocation. If a marketing campaign is not working and a second technician would be, the owner moves the money by deciding to. There is no department to reallocate from, no permission to withdraw, no budget holder to inform. The coordination problem that re-planning solves is largely absent, because one person already coordinates everything by being the person who decides.

Operating commitments often arrive as discrete decisions rather than through formal funding gates: a hire in March, a vehicle in May, a lease renewal in August, each one a conversation and a signature. There is no gate at which the release of funds is reconsidered. There is the decision, and then there is the bank.

And version governance is rarely formal. When a small business re-plans, the spreadsheet is usually overwritten. The prior version is sometimes saved under a new filename and often not. Nobody reports variance against the version in force in March, because by September nobody is sure which version that was. The fixed point that large-company governance preserves is simply lost.

So the practice, imported without its safeguards, solves a problem the business largely does not have, requires machinery the business usually does not have, and takes away the one thing it did have: a stable statement of intent to measure the year against. What is left is reaction with a spreadsheet attached. The plan matches what happened, and the business has lost the ability to be surprised.

What the owner actually needs

The owner’s problem was never that the plan was out of date. It was that the plan, the current forecast and the known commitments were not in the same place when the question arrived. That is a workflow problem, not a planning problem, and the second guide in this series describes it in detail.

The answer is to keep the plan still and bring it to the decision.

Hold the plan. It was built in January from reasoning that was sound at the time. It does not need to be right about September to be useful in September; it needs to be present, so the September decision can be checked against what was intended.

Move the forecast. The forecast is where the changed conditions live. Late-paying customers, a lost contract, a stronger quarter than expected: all of that belongs in the forecast, updated as often as the information changes, and it should disagree with the plan whenever the business has moved.

Test each commitment against both. When the owner asks whether the second salesperson is affordable, the answer has three parts. What did the plan intend for this line? What does the forecast now expect? And given both, does the commitment fit? Sometimes the plan says no and the forecast says yes, because the business grew faster than intended. Sometimes the reverse. Either way, the owner is making the decision with the intention and the expectation both visible, which is what neither the abandoned budget nor the overwritten one provides.

That is continuous feasibility. The plan is consulted continuously. It is not overwritten continuously.

What holding the plan requires

Holding the plan sounds passive. In practice it takes three things, and they are the practical half of this guide.

Keep the assumptions. A budget line is a claim about the future, and the number on its own does not say why it was chosen. Revenue of 220,000 in March might rest on a renewal, a seasonal pattern, or a new hire ramping. When March arrives and revenue is 180,000, the assumption is what tells you which part of the reasoning failed. Without it, variance is a number with no explanation attached. This is often the cheapest improvement available to a firm: a short note beside each material line, written when the line is set.

Keep a decision record. Each time the owner commits against the plan, capture the decision, the date, what the plan said, what the forecast said at the time, and the reason for going ahead. That is the whole entry. It does two jobs. During the year it is the list to revisit when the cash position moves, so the hire judged affordable in March is looked at again in May when a large customer pays late. At year end it is the evidence for the next plan, because it shows where the business deviated from intent and why.

The decision record also answers a question practitioners raised repeatedly in our conversations: why budgets are hard to maintain for clients who are not difficult. The decisions that move a business away from its plan are made in conversations the firm is not part of. Nothing about the monthly review recovers them. A record kept at the moment of decision is the input the firm never had.

Grade the plan at year end. With assumptions and a decision record beside it, the plan can be marked honestly. Not just where it missed, but why: the assumption was wrong, or the assumption was right and a decision overrode it. “We grew faster than planned” and “we underestimated payroll” both produce a variance on the payroll line, and they are different lessons for next year. This is what makes the next budget better than the last one, and it is the honest version of the claim that planning improves over time. The plan does not improve during the year. The answers do, because the forecast is current. The plan improves next year, because the evidence was kept.

What a tool would have to do

Most of this can be done by hand for a small number of clients. A firm with a larger book will look for software, and the practice above is a reasonable test of what the software has to do.

It would need to hold the plan as loaded, without rewriting it, and take a new version only when the firm deliberately supplies one. It would need to keep the forecast current from the bank and the ledger. It would need to answer the owner’s affordability question against both, at the moment the question is asked, and it would need to create the decision record as a by-product of answering rather than as a separate task. And at year end it would need to hand the firm the graded plan and the record together, so the next budget is built from evidence.

That is a smaller piece of software than a planning system. It does not continuously rewrite the plan. It preserves each version supplied by the firm, records when it became effective, and keeps the original baseline available for comparison. It reads the plan, keeps it beside a moving forecast, and shows up when a commitment is about to be made. The judgement about whether to proceed stays with the owner and the firm; the software supplies attention and a record.

Five questions for a firm

1. When a client asks whether something is affordable, is the answer given against the plan, the forecast, or the bank balance?
2. If the plan has been revised this year, can you still see what it said in January?
3. Is there a record of the decisions that moved a client away from the plan, with the reason?
4. When you built this year’s budget, what evidence from last year did you build it from?
5. If a material assumption in the budget fails, will you know which one it was?

Frequently asked questions

Is this an argument against rolling forecasts?

No. The forecast should roll. It should be updated whenever the information changes and it should look forward on a moving horizon. The argument is that the budget should not roll with it, because the budget’s job is to stay still.

What if the plan is clearly wrong by April?

Then the forecast will show it, and the variance between the two is exactly the information you want. If the firm decides the original intent no longer makes sense and re-plans, that is a deliberate act: a new version, dated, with the old one kept. The problem is not re-planning. It is re-planning by drift, where the plan is nudged monthly until it matches the forecast and nobody can say when the intent changed.

Doesn't the annual budget still go stale?

Its numbers do. Its role does not. A stale budget beside a current forecast tells the owner both what was intended and what has changed, which is more than a current budget with no fixed point behind it. This approach keeps the annual plan in the job it was built for.

Isn't the decision record just more admin for the client?

If it is a separate task, yes, and it will not survive a busy quarter. It only works if it is captured where the decision is made, which for most owners means at the moment they ask whether something is affordable. The record has to fall out of the question.

Do larger clients in the range need to re-plan through the year?

A business approaching ten million with a few managers who hold spending authority starts to look like the setting the practice was built for. Even then, the version discipline matters more than the frequency: if the plan moves, keep the prior version and report against it.

Where does the firm's variance review fit?

It stays. The monthly comparison of budget to actual is where the assumptions get tested. What changes is that the review now has the decision record beside it, so a variance can be traced to a decision rather than only to a line.

About the author

Nilanjan Raychaudhuri is the founder of AgentLink (Tublian LLC, Columbus, Ohio), which builds controller-layer software sold through CPA and fractional CFO firms. He has spent the past year interviewing practitioners at fractional CPA and CFO practices about how advisory work is actually delivered. Team page

The practitioner observations in this guide come from design-partner conversations and are used with permission where attributed. Corrections to nilanjan@agentlink.finance.