Driver-based forecasting for owner-operators

What a driver is, and what it is not

A driver is a countable operational quantity that most of the revenue or cost of a business moves with.

Occupancy is a driver. Billable hours, transactions per location per day and units delivered per route are drivers. Each is something the operator already knows and can discuss without the accounting file.

Revenue growth of 8% is not a driver. It is an outcome, and so is gross margin percentage. These are useful measures of what happened and useless as forecast inputs, because they contain no mechanism you can question.

The test is simple. If you can ask the operator what this will be next month and why, and get an answer that is not a guess about the financial result, you have a driver.

Why driver based forecasting beats growth percentages

A percentage forecast is unfalsifiable, which sounds abstract and is the practical problem.

If you forecast revenue up 6% and it comes in up 1%, there is nothing to learn. The assumption had no structure behind it, so the variance has no cause, and next cycle you pick a different percentage, equally unfounded.

If you forecast 82% occupancy at a given rate and get 74% occupancy at a slightly higher rate, you know what happened and which part of the assumption to change. The forecast has become a set of claims about the world, and claims can be checked.

Percentages also hide compensating errors. Volume down and price up can net to a revenue line that looks fine while the business underneath has changed materially.

Finding the two or three real drivers

Finding them is a question about physics rather than accounting: what has to happen in the world for money to arrive?

A property or accommodation business runs on units available, occupancy and average rate. A professional services business runs on billable headcount, utilisation and effective rate. A retail or food location runs on transactions per day, average ticket and days open. A distribution or field-service business runs on routes, stops per route and revenue per stop.

Two or three is the right number. Six drivers is not a more accurate forecast, it is a model nobody maintains. Pick the ones where a 10% error changes the answer, and let everything else be a rate applied to those. If a candidate driver moves 10% and your forecast barely changes, it is not a driver.

Building revenue from drivers

Revenue becomes an arithmetic statement rather than an assertion. For a property entity: units multiplied by occupancy multiplied by average rate multiplied by days. For a services entity: billable headcount multiplied by available hours multiplied by utilisation multiplied by realised rate. For a location: days open multiplied by transactions per day multiplied by average ticket.

Write it out that way, each factor in its own cell, and forecast each separately. This is where the value appears, because the operator can dispute a specific factor. Utilisation will not hold at 78% in July because two people are on leave is a conversation that improves the forecast. Revenue feels high is not.

Capacity constraints also become visible. If the arithmetic requires 105% utilisation to hit the number, the plan cannot happen.

Building costs from the same drivers

Costs split into three behaviours, and mapping them correctly is most of the work.

Fixed costs do not move with any driver: rent, insurance, base salaries, software. Forecast these from contracts and known changes, not as a percentage of revenue.

Variable costs move with a driver directly: cost of goods per unit, card processing per transaction, vehicle cost per route, subcontractor cost per billable hour. Each should reference the same driver cell the revenue line uses, so when occupancy changes both sides move.

Step costs are the ones people get wrong. They are fixed until a driver crosses a threshold, then they jump. One more van means one more driver and one more insurance policy; a fourth location needs a supervisor. Forecasting them as a percentage of revenue promises margin expansion that cannot occur.

Testing driver forecasts against actuals

A driver forecast is testable, so test it.

Each month, compare the forecast driver value against the actual, separately from the financial variance. Occupancy forecast 82%, actual 76%. Utilisation forecast 78%, actual 79%. Transactions per day forecast 240, actual 251.

Then decompose the revenue variance into driver and rate error. This is where you learn whether you are systematically optimistic on volume, conservative on price, or simply noisy. Most owners find a consistent bias on one driver, and correcting that single bias improves the forecast more than any structural change to the model.

Track this for two quarters and you will know which drivers you can forecast. The ones you cannot are candidates for a range rather than a point estimate.

A worked example: three entities, three driver sets

Consider a group of three operating entities under one owner.

Entity A operates twelve short-stay units. Its drivers are units, occupancy and average nightly rate: 12 units at 74% occupancy at 180 per night. Variable costs are cleaning per stay and platform commission per booking, both keyed to nights sold.

Entity B is a services business with nine billable staff. Its drivers are headcount, utilisation and realised rate: 9 staff, 1,750 available hours each, 68% utilisation, 145 per hour. Variable cost is subcontractor hours when utilisation would exceed 85%, which is a step cost.

Entity C runs four service routes: 4 routes, 22 stops per day, 96 per stop, 250 operating days. A fifth route adds a vehicle, a driver and insurance at once.

Three entities, three different driver sets, one group total. This is what gets missed when groups run one forecast template across every business. The template forces a common structure, the common structure is usually a percentage of last year, and the forecast stops describing any of the businesses.

The roll-up is then arithmetic, with intercompany charges eliminated. What is tempting and wrong is to forecast the group directly, because a group-level growth assumption is exactly the unfalsifiable number this approach exists to avoid, for the same reasons a static budget goes stale by spring.

Can QuickBooks or Xero do this?

No, and it is worth being clear why.

QuickBooks Online and Xero hold financial quantities, not operational ones. Occupancy, utilisation, stops per route and transactions per day are not in the chart of accounts, so there is nowhere for a driver to live and nothing for it to feed.

Both do the actuals side well, and both give you budget versus actual per entity, which is useful for grading a forecast after the fact. Class and location tracking gets you closer, letting you split revenue by site to derive a per-location average ticket, but that is a workaround rather than a driver model.

So most groups end up in a spreadsheet, which is fine as long as the drivers live in one clearly marked input block and every formula references those cells. Panko's synthesis of audits of 88 operational spreadsheets found 94% contained at least one error, with an average cell error rate of 5.2%, and hard-coded driver values scattered through a model are the most common way that happens.

The other constraint is time. AFP/APQC benchmarking found only 25% of FP&A time goes to value-added analysis, with 42% spent gathering data, and the FP&A Trends Survey 2024 of more than 2,400 practitioners put time on high-value work at 35%. A driver model is worth building because it moves effort from assembling numbers to arguing about a handful of assumptions, and it only pays off if the actuals arrive reliably each month.

Where to start

Take your largest entity and write one line: the arithmetic that produces its revenue. Then find the last twelve months of actual values for each factor in it. Most owners find one or two factors have never been measured, and finding that out is worth the exercise.

Forecast those factors for the next twelve months, build costs off the same cells, and compare against actuals monthly. Once one entity works the others take an afternoon each, because the hard part was never the spreadsheet, it was deciding what drives the business.

Every forecast here assumes last month is closed and correct. If that is the shaky part for your group, it is the part cruisr works on. Get in touch.