Demand Variance Calculator
Demand variance answers two questions planners mix up all the time. The first is simple: how far did actual demand land from the forecast for one period? The second is broader: how widely does demand swing around its own average over time, which is the volatility that drives safety stock and capacity decisions?
This calculator handles both in one place. Use the forecast tab for a single period gap, or paste a series of demand figures into the statistical tab and it works out the mean, both variances, the standard deviation, and the coefficient of variation for you.
One period tells you about forecast bias. A series tells you about volatility. The two tabs measure different things, so do not swap one result for the other.
Quick Answer
Key takeaways
- The sign carries the meaning: this tool defines variance as actual minus forecast, so a positive result means demand came in above the forecast and a negative result means it fell short.
- One period measures bias, a series measures volatility. A single +15% miss says little on its own; the same direction month after month points to a biased forecast, not random noise.
- Variance itself is in squared units, which is why the standard deviation (46.57 units on the worked example) and the coefficient of variation (5.28%) are the numbers to quote and compare.
- Volatility has a direct cash cost: safety stock formulas multiply the standard deviation of demand, so a wider spread means more buffer inventory for the same service level.
- Quick Answer
- Which tab answers your question?
- The two formulas, in plain terms
- Worked example 1: a month that beat its forecast by 15%
- Worked example 2: twelve months of demand, measured for volatility
- Reading the coefficient of variation without fooling yourself
- From variance to safety stock: why the standard deviation pays rent
- What competing calculators miss
- Frequently Asked Questions
- Add This Calculator to Your Website
- References
Which tab answers your question?
Use the forecast tab for one period and the statistical tab for a series. Picking the wrong one gives you a correct number for the wrong question.
Forecast vs actual answers "how far off were we this period?" Enter the forecasted demand and the actual demand for the same period. You get the gap in units, the gap as a percentage of the forecast, and a verdict on what the miss means.
Statistical variance answers "how jumpy is demand in general?" Paste demand for consecutive periods, separated by commas or spaces. The tool computes the mean itself, then the population variance, sample variance, standard deviation, and coefficient of variation, with a per-value deviation table.
Read the verdict before the raw variance. Variance is a squared-unit figure that is hard to judge alone. The verdict translates it: a forecast verdict flags bias, and a volatility verdict (stable, moderately volatile, highly volatile) is driven by the coefficient of variation.
The two formulas, in plain terms
The forecast formula is a subtraction and a division. The statistical formula averages squared distances from the mean.
The population and sample versions differ only in the divisor, a distinction the OpenStax business statistics text draws the same way: divide by n when your values are the whole set you care about, and by n minus 1 when they are a sample used to estimate a larger population. With 12 months of your own sales history, the population figure describes those 12 months exactly; the sample figure is the better estimate if you treat those months as a sample of future demand. The gap between the two shrinks as you add periods, which is why the tool reports both side by side instead of picking for you.
One convention warning, because it causes real confusion in review meetings: this tool, like the Corporate Finance Institute variance formula, defines the gap as actual minus forecast and the percentage against the forecast. Some teams define it the other way round, as forecast minus actual, or divide by the actual. The same month can therefore appear as +15.00% in one report and -13.04% in another, with both arithmetics internally consistent. Agree the convention before you compare numbers across teams.
Worked example 1: a month that beat its forecast by 15%
The forecast tab is pre-filled with this case: a forecast of 10,000 units and actual demand of 11,500 units for the same month.
| Measure | Result | How it is reached |
|---|---|---|
| Forecast variance | +1,500.00 units | 11,500 minus 10,000 |
| Variance % | +15.00% | 1,500 / 10,000 x 100 |
| Verdict | Demand came in 15.00% above forecast | Positive variance, so demand beat the plan |
A single month at +15% is not yet a forecasting failure. Promotions, weather, and one-off orders all land in a single period. The verdict text makes the useful distinction: a persistent run in the same direction is bias, and bias is fixable by adjusting the forecast itself. Random scatter around the forecast is noise, and noise is handled with buffer stock, not with a new forecast. If the shortfall side is what costs you, the cost of understocking calculator puts a dollar figure on running out when demand beats the plan.
Worked example 2: twelve months of demand, measured for volatility
The statistical tab is pre-filled with a year of monthly demand for one product: 820, 860, 790, 910, 880, 940, 870, 900, 850, 920, 890, 960. The months average 882.50 units, and no month strays far from that average, which is exactly what the numbers below confirm.
| Value | Deviation from mean | Squared deviation |
|---|---|---|
| 820.00 | -62.50 | 3,906.25 |
| 860.00 | -22.50 | 506.25 |
| 790.00 | -92.50 | 8,556.25 |
| 910.00 | 27.50 | 756.25 |
| 880.00 | -2.50 | 6.25 |
| 940.00 | 57.50 | 3,306.25 |
| 870.00 | -12.50 | 156.25 |
| 900.00 | 17.50 | 306.25 |
| 850.00 | -32.50 | 1,056.25 |
| 920.00 | 37.50 | 1,406.25 |
| 890.00 | 7.50 | 56.25 |
| 960.00 | 77.50 | 6,006.25 |
The squared deviations add to 26,025.00. Divide by 12 for the population variance of 2,168.75, or by 11 for the sample variance of 2,365.91. The standard deviation, 46.57 units, is the figure in the same units as demand itself, so it reads naturally: a typical month lands within roughly 47 units of the 882.50 average. That average is also the starting point for lead time planning, and the average daily demand calculator converts the same history into a per-day figure when your lead times are quoted in days. Note the squaring effect in the table: March, the worst month at -92.50, contributes 8,556.25 on its own, a third of the total. Variance punishes big misses far more than small ones, which is the right behaviour for a risk measure and a reason to look at the deviation table rather than the headline alone.
Reading the coefficient of variation without fooling yourself
The coefficient of variation is the standard deviation expressed as a percentage of the mean, and it exists because variance alone cannot be compared across products.
A variance of 2,168.75 sounds large until you set it against a mean of 882.50: the CV of 5.28% shows this product is steady. The same variance on a product averaging 90 units a month would be a CV above 50%, a genuinely erratic item. The tool turns the CV into a plain verdict using fixed bands: 15% or under is stable, over 15% up to 30% is moderately volatile, and over 30% is highly volatile. For a deeper treatment of the ratio itself, including how it behaves when the mean is near zero, see the coefficient of variation calculator.
| Context | Typical CV range | What it usually means |
|---|---|---|
| This tool: stable | 15% or under | Forecasts and level production plans hold up well |
| This tool: moderately volatile | Over 15% to 30% | Buffer stock and flexible capacity start to matter |
| This tool: highly volatile | Over 30% | Wide safety stock, unreliable single-point forecasts |
| Retail and e-commerce (rough norm) | 5–15% | Promotions and seasonality drive most of the swing |
| Manufacturing (rough norm) | 3–10% | Contracted and scheduled demand keeps the spread low |
| Technology products (rough norm) | 8–22% | Launches and short life cycles widen the spread |
Treat the industry rows as orientation, not targets. A CV is only comparable when the period length matches: monthly demand is almost always smoother than daily demand for the same product, because daily spikes average out. Comparing your monthly CV against a benchmark built on weekly data will flatter or punish you for the calendar, not for your forecasting.
From variance to safety stock: why the standard deviation pays rent
Variance earns its keep when it feeds a buffer decision. The standard safety stock formula used in inventory practice, set out in the Institute for Supply Management guidance, multiplies a service-level factor (the Z-score) by the standard deviation of demand over lead time. Double the standard deviation and, at the same service level, you roughly double the buffer you must hold and pay to store.
That is the practical reading of the worked example. A standard deviation of 46.57 units a month is a modest buffer problem. A product with the same 882.50 mean and a standard deviation of 300 units would need several times the safety stock for the same fill rate, and its forecasts would deserve far less trust in a production plan. Run the statistical tab per product, rank by CV rather than by raw variance, and the items that deserve forecasting effort and buffer budget sort themselves quickly.
What competing calculators miss
The demand variance tools ranking for this term split into two camps, and each camp leaves half the job undone.
Miss 1: forecast-only tools stop at one subtraction. The typical forecast variance calculator takes two numbers and returns a gap in units and percent. That is useful for one period and useless for the volatility question behind safety stock, reorder points, and capacity planning. None of them touch a demand series at all.
Miss 2: statistics tools make you do their job first. The strongest statistical competitor requires you to work out the mean by hand before its variance calculation will run, then reports variance on its own, with no standard deviation in demand units and no coefficient of variation to compare products of different sizes. Here the mean is computed from your data, and population variance, sample variance, standard deviation, and CV arrive together with the deviation table that proves them.
Miss 3: a different formula wearing the same name. One ranking tool defines demand variance as the average of squared forecast errors across periods, which is a forecast accuracy measure, not the spread of demand. Both are legitimate, but they answer different questions and produce unrelated numbers. This page states both of its definitions in the open, shows the deviation table behind the statistical result, and flags the sign-convention trap that makes two honest reports disagree about the same month.
Note: this calculator is a planning aid. Industry CV ranges are rough orientation figures, not standards, and buffer decisions should also weigh lead time variability, service targets, and the cost of a stockout for the specific product.
Frequently Asked Questions
Demand variance has two related meanings, and this calculator covers both. In forecast mode it is the gap between actual demand and forecasted demand for one period, in units and as a percentage. In statistical mode it is the spread of a series of demand values around their mean, reported as variance, standard deviation, and coefficient of variation.
Because teams use opposite sign conventions. This tool defines variance as actual minus forecast, so demand above forecast is positive: 11,500 actual against a 10,000 forecast is +1,500 units, +15.00%. A team defining it as forecast minus actual reports the same month as negative. Check the definition before comparing figures across reports.
Use population variance when your values are the complete set you care about, such as all 12 months of last year. Use sample variance when those values are a sample used to estimate a larger population, such as future demand. The tool reports both: 2,168.75 population and 2,365.91 sample on the worked example, and the difference shrinks as you add values.
Statistical variance cannot be negative, because it averages squared deviations, and squares are never negative. A forecast variance can be negative: actual minus forecast is negative whenever demand falls short of the forecast, and the sign is the useful part, since it tells you which way the forecast missed.
Lower volatility is easier to plan around, and the tool grades the coefficient of variation as stable at 15% or under, moderately volatile over 15% up to 30%, and highly volatile over 30%. As rough orientation only, retail demand often runs at a 5 to 15% CV, manufacturing at 3 to 10%, and technology products at 8 to 22%. Your period length and product type matter more than any single benchmark.
Safety stock formulas multiply the standard deviation of demand over lead time by a service-level factor, so higher variance means more buffer inventory for the same service level. On the worked example the standard deviation is 46.57 units, a modest buffer. A product with the same mean and a standard deviation of 300 units would need several times the safety stock.
No, and it is worth being clear about that. The forecast mode covers one period, and the statistical mode measures the spread of demand itself, not the spread of forecast errors. MAPE and multi-period forecast-error measures pair each period forecast with its actual, which is a different dataset. Use this tool for the single-period gap and for demand volatility, and a forecast accuracy review for error tracking over time.
Add This Calculator to Your Website
References
- OpenStax, Rice University, Introductory Business Statistics 2e, "Measures of the Spread of the Data" - population and sample variance and standard deviation definitions used in the statistical mode. OpenStax: Measures of the Spread of the Data. Accessed October 2026.
- National Institute of Standards and Technology (NIST), "Section 8: Statistical Techniques" - standard deviation estimation and pooling methods behind the variance formulas. NIST: Section 8, Statistical Techniques. Accessed October 2026.
- Corporate Finance Institute, "Variance Formula Template" - the actual minus forecast convention and the equivalent percentage form, Variance % = Actual / Forecast minus 1. CFI: Variance Formula Template. Accessed October 2026.
- Institute for Supply Management, "Optimize Inventory with Safety Stock Formula" - how demand variance and lead time variance combine in safety stock decisions. ISM: Safety Stock Formula. Accessed October 2026.
- Rework, "Safety Stock Calculation: Mathematical Approach to Demand Uncertainty" - the combined demand and lead time variability formula and its data cleaning cautions. Rework: Safety Stock Calculation. Accessed October 2026.
- Statistics By Jim, "Mean Absolute Percentage Error" - what MAPE measures and why it needs paired forecasts and actuals, the boundary this tool discloses. Statistics By Jim: Mean Absolute Percentage Error. Accessed October 2026.