Budgets people believe

Every budget round meets two numbers that rarely agree.

• A top-down target says what the business needs. Klasna's board asks for 3,000,000 UAH of operating profit next year.
• Bottom-up detail says what it takes to deliver: each team's plan for its own lines. Klasna's teams send budgets that add up to 1,523,968 UAH.

Neither number wins by default. The target can be out of reach; the detail can be padded. A budget people believe joins the two through drivers, the same learners, lessons, prices and costs the forecast is built from.

The order matters. First, check the bottom-up lines against their drivers and take out the padding. Only then is the remaining gap honest, and you close it by asking what would have to be true: more lessons, a higher price, or a lower target.

The shortcuts all fail the same way: forcing the target onto every line, accepting the bottom-up total unchecked, or splitting the difference produce a number with no driver behind it, so nobody owns it.

• Top-down target: a result the owners or the board need, set from above. Klasna: 3 million UAH of operating profit next year.
• Bottom-up detail: the plan built by the people who run each driver. Klasna's marketing plans 260,000 UAH for October and expects 1,010 learners.
• Budget slack: protection built quietly into a line, costs set high or sales set low, so the budget is easy to beat. Support asks for four coordinators all year, whatever the learner count.
• Visible reserve: the same protection made honest. Finance holds one amount for genuine surprises, shown as its own line.

Slack and a reserve can protect against the same risks. The difference is visibility. Slack spread across ten lines is invisible: nobody can say how much uncertainty the budget holds, and each team spends its cushion. One reserve in one place shows the total, and it is released only when a real surprise happens.

The first two terms are the inputs of the round, the last two are two ways to handle uncertainty.

Budget slack is costs set high or sales set low so the budget is easy to beat. Nobody is quite lying: people get blamed for missing a budget and almost never for beating one, so they protect themselves with "just in case".

You find slack the same way you built the forecast: recompute each line from its driver and compare.

• Support asks for four coordinators in every month, 48 coordinator-months. The rule, one coordinator per started 400 active learners, gives 37 with the forecast's learners. 11 extra months × 38,000 UAH = 418,000 UAH.
• Marketing asks for 3,400,000 UAH, while its own month-by-month plan adds up to 3,090,000. Extra: 310,000 UAH.

Together that is 728,000 UAH, exactly the gap between the forecast's 2,251,968 UAH of operating profit and the submitted budget's 1,523,968.

Notice what the check doesn't need: a confrontation. You ask each team for its driver and its source; the numbers then show where the cushion sits.

Once the padding is out, the gap that remains is honest: 3,000,000 − 2,251,968 = 748,032 UAH at Klasna. Close it the way the forecast was built, through drivers, by asking what would have to be true.

• Volume: each extra lesson leaves 187.25 UAH and fixed costs don't move, so 748,032 ÷ 187.25 ≈ 3,995 more lessons, about 5% above the forecast's 80,128.
• Price: spread over all 80,128 lessons and net of the 2.5% fee, 748,032 ÷ (0.975 × 80,128) ≈ 9.57 UAH more per lesson.
• Target: the board accepts 2.25 million.

All three are honest answers: each names a driver, an owner who can say whether it is realistic, and a number that can be checked month by month.

The tempting fourth answer, the same percentage off every team's costs, is not. It hits a lean line as hard as a padded one, teaches teams to pad by that percentage next year, and says nothing about how the business will actually earn the extra 748,032.

When fixed costs stay where they are, every extra unit adds exactly its contribution margin to profit. So the volume needed to close a profit gap is:

Extra units = profit gap ÷ contribution margin per unit, rounded up, because 3,994 lessons would leave the gap a little short.

Klasna: 748,032 ÷ 187.25 = 3,994.8, so 3,995 more lessons a year. Against the forecast's 80,128 that is about 5% more, less than half of one busy May.

The number is only the start of the conversation. Whoever owns learners and lessons per learner has to say whether 5% more is realistic, and how: more learners from marketing, or more lessons from the same families. The second route is cheaper, because more learners can also bring more coordinators, a step cost the simple division ignores.

The same logic works in reverse: a planned loss of volume costs its lessons × 187.25, unless fixed costs move too.

A budget written in drivers gives you a fair way to judge spending: recompute each driver-based line at the volume that actually happened, then compare.

Klasna's support budget isn't "152,000 UAH in May". It is "one coordinator at 38,000 UAH per started 400 active learners". The plan expected 1,460 learners, so four coordinators. If 1,620 learners come, the rule allows five: 1,620 ÷ 400 = 4.05, five started groups, 190,000 UAH. Support spending 38,000 more than planned is the rule working, not overspending.

The reverse holds too. If only 1,150 learners come, the rule allows three coordinators; a team that still pays four is over budget even though it spent "exactly the plan".

This is what makes people stop padding. When growth-driven costs are never counted against a team, nobody needs a "just in case" coordinator. And three rules keep the whole budget honest: every line comes from a driver someone owns, owners are judged on their drivers, and uncertainty sits in one visible reserve held by finance.

A team that spends less than its cost budget looks efficient. But "less than the budget" only means "less than the team itself asked for". If the request was padded, beating it every year shows the padding, not the efficiency.

Klasna's support came in well under budget last year, and Leo congratulated them in front of everyone. This year support asks for four coordinators in every month, 48 coordinator-months, while the rule (one per started 400 learners) gives 37 with the forecast's learners. Praise for underspending taught exactly the habit that produces slack.

The honest yardstick is the driver at actual volume. How many learners did support really serve, and how many coordinators does that justify? How much marketing did the actual plan need? If spending sits below that, the team really did run lean. If it sits below an inflated request, nobody learned anything.

The same goes for sales budgets: beating a low sales target is easy if the target was set low on purpose.

Slack is a reasonable response to how budgets are usually run, so lecturing people about honesty doesn't remove it. Changing the round does. Teams copy whatever gets applause, and they protect themselves against whatever gets blamed.

Practices that make padding less likely:

• Ask for drivers, not totals. Each team sends its drivers and how it got them; a padded line then stands out next to its driver.
• Hold one visible reserve in finance for genuine surprises. Teams no longer need private cushions, and everyone sees how much uncertainty the budget carries.
• Judge costs at actual volume. If learners beat the plan and support needs a fifth coordinator, that is the rule working, not overspending.

Practices that make padding more likely:

• Praising the biggest underspend: it rewards whoever padded most.
• Cutting every request by a fixed percentage: next year everyone adds that percentage first, and honest teams lose.

A budget is a promise about drivers, not a pot of money each team defends.