Forecast vs actual variance analysis that improves your next forecast
Forecast vs actual compares the numbers your latest forecast predicted with what your books recorded once the month closed. The gap, measured in dollars and percent, shows how accurate the forecast was. Budgeter.ai tracks that error by line and month, explains the cause and adjusts the drivers of your rolling forecast.
Read-only connections. Encrypted data. Cancel anytime.
Scroll sideways for all columns
| Month | Forecast | Actual | Error | Error % |
|---|---|---|---|---|
| April | $318,000 | $301,400 | -$16,600 | -5.5% |
| May | $326,000 | $314,900 | -$11,100 | -3.5% |
| June | $331,000 | $333,800 | +$2,800 | +0.8% |
| July | $342,000 | $336,200 | -$5,800 | -1.7% |
| August | $349,000 | $352,600 | +$3,600 | +1.0% |
| September | $358,000 | $355,100 | -$2,900 | -0.8% |
Absolute percent error by month
6 month MAPE
2.2%
Last 3 months 1.2%, bias slightly low after April fix to close rates.
Example figures inside the product interface.
Why forecast accuracy deserves its own review
Every decision a growing company makes about hiring, spending or raising money rests on a forecast. If that forecast is consistently 8 percent too optimistic on revenue, the cash runway on the dashboard is too long, the next hire looks affordable when it is not, and the fundraising date arrives sooner than anyone planned. Forecast vs actual is the review that catches this before it becomes expensive.
The point is not to grade the finance team. A forecast is a set of assumptions about prices, volumes and dates. Comparing it with what happened tells you which assumption was wrong, so the next version can be a little less wrong. Done every month, this compounds. After half a year the forecast stops being a hopeful spreadsheet and starts behaving like an instrument you can trust.
This is a different question from the one answered by budget vs actual. That companion report measures results against the annual plan the board approved, and it shows whether the company kept its commitments. Forecast vs actual measures results against the latest forecast, and it shows whether your understanding of the business is getting better. Most finance leads run both after each close, in that order.
Splitting a miss into price, volume and timing
A 60 person services company forecasts three lines a month ahead. The numbers below show how the same error looks very different once you split it by cause.
Scroll sideways for all columns
| Line | Forecast | Actual | Error | Error % | Main cause |
|---|---|---|---|---|---|
| Project revenue | $410,000 | $372,000 | -$38,000 | -10.2% | Volume, 2 of 14 projects not started |
| Retainer revenue | $186,000 | $179,500 | -$6,500 | -3.6% | Price, one client renewed at a discount |
| Contractor costs | $94,000 | $71,300 | -$22,700 | -31.8% | Timing, invoices arrive next month |
| Payroll | $312,000 | $309,800 | -$2,200 | -0.7% | In line |
Price
The volume happened, but at a different rate. Discounts at renewal, a currency move or a supplier price increase sit here. Price errors tend to persist, so the fix is to change the rate assumption for every future month, not just the next one.
Volume
The rate was right, but you sold, hired or consumed more or fewer units. Projects that never started and deals that were lost belong here. Volume misses usually trace back to a conversion rate or a capacity limit, which is the driver to revise.
Timing
The amount is right but lands in another month. A late invoice or a contract signed on the third makes this month look poor and next month look great. Timing errors should net to zero over a quarter. If they do not, they were never timing.
In the example, the total miss on profit looks modest because a contractor timing gain hides a real volume problem in project revenue. Read at the total level, the month looks fine. Read by cause, the company learns that its project start dates are too optimistic and that contractor costs will catch up in September. That is exactly the kind of insight a single total variance line hides.
How Budgeter measures forecast accuracy
Budgeter keeps the history, does the arithmetic and points at the driver, so the review takes minutes instead of an afternoon of version hunting.
-
01
Freeze the forecast
Budgeter saves each forecast as a dated version, so the number you compare against is the one you actually relied on, not a later edit.
-
02
Load the actuals
After the month closes, actuals sync read-only from QuickBooks Online or Xero, with revenue from Stripe and payroll from your payroll system.
-
03
Measure the error
For every line it calculates the error in dollars and percent, the absolute percent error and the bias across recent months.
-
04
Feed the lesson back
You tag each miss as price, volume or timing, and the AI CFO suggests which driver of the rolling forecast to adjust next.
The hardest part of forecast vs actual in a spreadsheet is knowing which forecast to compare against. Files get overwritten, tabs get renamed and nobody remembers which version went to the board. Budgeter stores every forecast as a dated version tied to the scenario it came from, so the comparison is always against the number people actually used to decide.
Accuracy is shown at three levels. The company view shows MAPE and bias on revenue, operating costs and ending cash. The line view shows each account with its error history. The driver view shows the inputs behind a line, for example the pipeline close rate behind new revenue or start dates behind payroll, which is where the fix belongs.
The AI CFO reads the transactions and drivers behind a miss and proposes a cause in plain English, such as two projects slipping or a renewal discount. You confirm or change the tag. Over time those tags show patterns, like revenue misses that are almost always volume, which tells you where your assumptions are weakest.
On the Scale plan the rolling forecast recomputes automatically after each month close, so the corrected drivers flow straight into the next 12 or more months. The same engine powers scenario planning, so an improved close rate assumption also improves every what-if you run afterward.
The learning loop into your rolling forecast
Measuring error is only useful if it changes the next forecast. This is the monthly loop a disciplined finance lead follows, and the one Budgeter is built around.
If you are still setting up your first model, start with our guide to budgeting and forecasting software and come back once you have two or three closed months to compare.
- Compare against the frozen version Use the forecast as it stood at the start of the month. Comparing against a version edited mid month hides exactly the errors you want to see.
- Rank misses by dollar impact A 40 percent error on office snacks matters less than a 4 percent error on subscription revenue. Start the review with the lines that move cash.
- Tag every material miss with a cause Price, volume or timing. Without the tag you cannot see patterns, and without patterns you keep making the same mistake.
- Change the driver, not the output Overwriting next month revenue by hand fixes one cell. Lowering the close rate or moving the start date fixes every month that depends on it.
- Watch bias over three to six months One month proves little. A consistent lean in one direction is the clearest sign that an assumption needs to move.
- Record what you changed and why A short note per change lets you check next month whether the fix worked, and it gives the board a clear story about forecast quality.
Spreadsheet vs FP&A suite vs Budgeter
You can measure forecast accuracy with formulas and discipline. The question is how much of the work you want to keep doing by hand each month.
Scroll sideways for all columns
| Task | Spreadsheet | Enterprise FP&A suite | Budgeter.ai |
|---|---|---|---|
| Keeping forecast versions | Copies of files, easy to lose | Built in | Every version dated and saved |
| Error and MAPE by line | Formulas you build and maintain | Usually available | Calculated after each close |
| Bias over time | Rarely tracked | Varies by product | Shown per line and per driver |
| Cause of each miss | Written by hand | Analyst work | AI CFO suggestion you confirm |
| Updating the next forecast | Manual edits across tabs | Built in | Driver change flows to all months |
| Price and setup | Your time | Often 10,000 to 150,000 dollars a year | Public price, self-serve |
A spreadsheet is perfectly able to calculate forecast error. What it cannot do on its own is remember which version was live, pull fresh actuals without copy and paste, or point at the driver that caused a miss. Those gaps are why accuracy tracking is often the first thing to slip when the month gets busy.
Dedicated FP&A platforms do this well, but they typically cost 10,000 to 150,000 dollars a year, and products such as Datarails, Cube and Abacum only share a price after a demo. A fractional CFO who runs the review for you often costs 3,000 to 10,000 dollars a month. Budgeter lists every plan on the pricing page, starting at 149 dollars a month.
Forecasts and scenarios are estimates based on your data and assumptions, not financial, investment, tax or accounting advice.
Questions about forecast vs actual
Forecast versions and accuracy history are kept for every company. Automatic reforecast after month close is part of the Scale plan and above.
What is the difference between forecast vs actual and budget vs actual?
Budget vs actual compares results with the annual plan you approved, which stays fixed for the year. Forecast vs actual compares results with your most recent forecast, which you update as you learn. The first checks discipline against a commitment, the second checks how good your view of the future is.
How do you calculate forecast error?
Forecast error equals actual minus forecast. Percent error divides that difference by the actual and multiplies by 100. Absolute percent error drops the sign, so overshoots and undershoots both count as misses of the same size.
What is MAPE in simple terms?
MAPE, mean absolute percentage error, is the average size of your misses in percent over several months. A MAPE of 4 percent on revenue means that, on average, your forecast was 4 percent away from what happened, in either direction.
What is a good forecast accuracy for a growing company?
There is no universal number. Stable cost lines such as rent and payroll can usually be forecast within a few percent, while new revenue, hiring dates and project costs move more. Track your own trend and aim for it to improve, line by line.
Why track bias as well as MAPE?
MAPE tells you how big the misses are, bias tells you whether they lean one way. If revenue is too high in nine of twelve months, the error is not random. It points to an assumption, often close rates or start dates, that needs to change.
Find out how accurate your forecast really is
Start with a runway forecast from a few numbers, then connect your books and see forecast error and MAPE after your next close.
Forecasts and scenarios are estimates based on your data and assumptions, not financial, investment, tax or accounting advice.