Choose a method before you open a spreadsheet
Most budget processes go wrong in week one, when someone copies last year's tab, adds 8% to everything and calls it a plan. That is incremental budgeting, and it is not always wrong — but it should be a decision, not a default. Here are the three realistic options with their true costs.
| Method | Build effort | Accuracy on variable cost | Use when |
|---|---|---|---|
| Incremental (last year ± %) | 1–2 weeks | Low to moderate | Stable business, flat headcount, low growth |
| Zero-based (justify every line) | 6–10 weeks | High, but degrades fast after Q1 | Cost reset needed, post-merger, margin under pressure |
| Driver-based (volume × unit cost) | 3–5 weeks | High and re-forecastable monthly | Growth or seasonality; anything usage-billed |
| Rolling 12-month forecast | 2 weeks setup, 2 days monthly | Highest short-horizon | Volatile demand; boards that want quarterly re-cuts |
For most companies under 200 people, driver-based is the best value. You define ten to fifteen drivers — new customers, average order value, churn rate, support tickets per customer, cloud cost per active user, cost per hire — and let the spend fall out of them. When reality diverges in March you change a driver, not four hundred cells.
The six-week build calendar
- Week 1 — close the gap. Lock year-to-date actuals through the most recent closed month and produce a straight-line estimate for the remainder. You cannot budget forward from numbers you have not reconciled.
- Week 2 — agree drivers and assumptions. Inflation on salaries, price increases from key vendors, expected win rate, churn, and any contractual step-ups. Write them on one page and get them initialled.
- Week 3 — revenue first. Build the top line by segment and by month, not annually divided by twelve. Seasonality matters: a January-heavy renewal base and a July-heavy one need very different cash plans.
- Week 4 — headcount. Name every role, its start month and its fully-loaded cost. Start months are the single largest source of budget error.
- Week 5 — non-people spend and scenarios. Software, marketing, travel, professional fees, cloud. Then build base, downside and upside cases.
- Week 6 — review and sign-off. Two passes maximum: one with budget owners, one with the board or founder. Publish the final version read-only and archive the drafts.
Cost headcount properly
A £52,000 base salary in the UK is not a £52,000 line item. Add employer National Insurance at 15% above the secondary threshold (roughly £6,000), pension contribution at 3–6% (£1,600–£3,100), employer's liability and health cover (£600–£1,400), equipment amortised over three years (£500–£700 a year), software seats (£900–£1,600 a year at typical stack prices) and recruitment fees where applicable (15–22% of first-year salary). The realistic loaded figure lands between £64,000 and £70,000 — a multiple of about 1.24x to 1.35x.
Then apply the start-month discipline. A role budgeted from 1 January that actually starts on 1 April costs 75% of the annual figure, not 100%. If you plan eight hires, assume slippage: a common convention is to budget each hire one month later than the hiring manager requests, which typically recovers 6–8% of your people budget as genuine contingency rather than optimism.
Three scenarios, clearly labelled
Do not build eleven versions. Build three, and define what triggers each.
- Base (55–60% likely): current win rate held flat, churn at trailing twelve-month average, planned hires at agreed months.
- Downside (20–25%): revenue 15% below base, hiring frozen after Q1, discretionary marketing cut 40%. Note the month cash falls below your minimum reserve.
- Upside (20%): revenue 12% above base, two additional hires triggered only after two consecutive months of over-performance.
Each scenario needs a written trigger and a named decision-maker. "We will revisit if things change" is not a trigger; "if bookings are below 85% of plan for two consecutive months, hiring pauses and the CFO reforecasts" is.
Where zero-based budgeting earns its keep
- Surfaces zombie subscriptions and duplicate tooling immediately
- Resets a cost base that has drifted for three or four years
- Forces owners to articulate what a line actually buys
Where it costs more than it saves
- Six to ten weeks of senior time in your busiest quarter
- Encourages sandbagging once teams learn the exercise repeats
- Little benefit where 70%+ of spend is contractual payroll
Sanity checks before sign-off
- Does monthly revenue add to the annual figure, and does seasonality look like last year's shape?
- Is every hire named with a start month and a loaded cost?
- Does the cash line stay above your minimum reserve in the downside case?
- Is there a contingency line of 3–5% of controllable spend, unallocated?
- Can each budget owner explain their largest three lines without opening the model?
Once the numbers are locked, make sure the sales plan behind the top line is credible — see year-end targets and the December sprint — and check your software spend against the planning tools comparison before you renew anything in January.