How to do financial forecasting for your business

To forecast your finances, start from your history and project revenue first, because it drives everything else. Add fixed and variable costs, convert the result into a week-by-week cash forecast, and write down every assumption. Update it monthly and compare it with actual results, so you can see which assumptions were wrong.

You are about to commit to numbers: next year's budget, a lender's terms, a hiring plan. Your last forecast missed, and you are not sure whether the method or the assumptions failed. Knowing how to do financial forecasting well pays off in three places: you see a cash shortfall weeks ahead, you know when you can afford to hire or buy, and you can answer a lender or partner with numbers you can defend.

How is a forecast different from a budget?

A budget is the plan you intend to follow, and a forecast is your current best estimate of what will happen. They should differ, and the difference is information. How to create a financial plan for your business shows where each one fits.

How to do financial forecasting, step by step

Six steps take you from raw history to a forecast you can defend.

1. Gather your history

Pull monthly revenue, costs, and cash for at least two years if you have them, so seasonal patterns show up in more than one cycle. A newer business uses what it has and states the assumptions it cannot back with history. Check that the figures reconcile with your accounting records first.

2. Forecast revenue by driver

Tie revenue to something you can count: customers, units, contracts, or pipeline. For example, suppose you have 40 customers paying an average of $2,500 a month, and you expect to win three a month and lose one. By month six you have 52 customers and revenue of $130,000 a month (52 × $2,500). The drivers are the 40, the $2,500, the three, and the one, and each can be checked against reality.

3. Forecast costs

Separate fixed costs, such as rent and salaries, from variable costs, such as materials and commissions, which move with sales. Date every planned hire, price change, and purchase in the month it takes effect.

4. Convert the result to cash

Revenue is not cash. Shift each item to the week it will be collected or paid, and add taxes, loan payments, and equipment purchases. How to improve cash flow for your business explains the timing levers.

5. Write the assumptions down

List each assumption with its source and an owner: the close rate comes from the last four quarters, the price increase starts in April, and the supplier terms are net 30. A forecast without a written assumption list cannot be corrected, because nobody can tell what to change.

6. Build three cases

Run a base case, a low case, and a high case. Pick the low case from your own history, such as your worst quarter in the last two years or your largest customer paying 30 days late. If the low case runs out of cash, you have found your next decision while there is still time to make it.

If you are forecasting for a lender, the U.S. Small Business Administration's business plan guidance asks for monthly or quarterly projections for the first year and a five-year outlook.

Which forecasting method should you use?

Most owners should start with the driver-based method. The others are useful checks.

Whatever the method, ask your sales and operations leads what is changing before you finalize it. They know about the delayed order and the new competitor before the data does.

How often should you update a forecast?

Update the forecast monthly, once the books close, and the cash forecast weekly. Re-forecast between those dates when something material happens: a major contract is won or lost, a main cost moves, or results miss the forecast two months in a row.

A rolling forecast keeps the horizon constant. When a month closes, you replace it with actuals and add one more month at the end, so the forecast always looks 12 months ahead.

How do you know if your forecast is any good?

Measure the error. The mean absolute percentage error is the average gap between forecast and actual, expressed as a percentage of actual. Bias is the average of forecast minus actual, and it shows whether you lean optimistic.

For example, suppose you forecast revenue of $100,000, $110,000, and $120,000, and the actuals are $95,000, $112,000, and $100,000. The errors are 5.3%, 1.8%, and 20.0%, so the mean absolute percentage error is 9.0%. The bias is plus $7,667 a month, with two of three months over. The pattern says you are optimistic, and the 20% miss in month three is where to look for the assumption that failed.

Track accuracy by horizon, too. A forecast for next month is usually closer than one for month 12, and knowing the gap tells you how much weight each one can carry.

What mistakes make a forecast useless?

• Forecasting revenue with hope instead of drivers, so every month grows by the same percentage.
• Treating revenue as cash and ignoring when customers pay and bills fall due.
• Leaving out costs that arrive in lumps, such as insurance renewals, taxes, and annual software.
• Forecasting once a year, then never comparing it with results.
• Building the forecast alone, without the people who know about orders, hires, and prices.

What to do next

1. Export the last 24 months of revenue and costs by month, and check that they reconcile with your accounting records.
2. Write down the five assumptions that drive revenue, and ask sales and operations to challenge them.
3. Build the base case and the low case, and convert both to a 13-week cash forecast.
4. Put the monthly update on the calendar, and record the error each time, so the forecast improves.

For help with the model, see how Surfside Capital Advisors handles budgeting, forecasting, and scenario modeling as part of its FP&A (financial planning and analysis) work. The article on what FP&A is explains how the pieces fit.

Questions owners ask

Why forecast if the future is uncertain?

A forecast is not a prediction you will be held to. It shows what your assumptions imply, and it tells you when one of them stops holding. Without it, a surprise looks like bad luck. With it, a surprise points at a specific assumption you can fix.

Do you need forecasting software?

Not at first. A spreadsheet works while you have a handful of drivers and one person maintaining it. Consider software when several people edit the model, when you need scenarios side by side, or when data should flow in from your accounting system without retyping.