Paste your time-ordered data and choose a method — linear trend, moving average, or single/double exponential smoothing. Get fitted values, point forecasts, accuracy metrics and an actual-vs-forecast chart, all client-side.
Paste your time-ordered series, choose a method, and press Calculate.
You have 12 weeks of demand and need a forecast for next week. Enter the lab to separate trend and seasonality from noise.
Enter your values in time order, one per line (oldest first). You need at least four points; more history gives more reliable forecasts and accuracy estimates.
Pick your forecast method: a straight-line trend, a moving average (which smooths out bumps), or exponential smoothing (which gives more weight to recent data). Then adjust the settings and tell the calculator how many periods ahead you want it to forecast.
See the fitted values, your forecasts with simple error ranges, accuracy scores (MAPE and MAD), and a chart comparing actual data to fit to forecast.
Time-ordered data — demand, defects, downtime, throughput — carries patterns that one-off statistics miss. Forecasting models capture the level and trend in your series and project them forward so you can plan capacity, stock and staffing with evidence rather than gut feel.
Time series forecasting uses the history of a measurement collected at regular intervals to predict its future values. The simplest models fit a trend line; exponential smoothing adapts to changing levels and trends by weighting recent data more heavily. Good forecasts come with an honest measure of accuracy and uncertainty, so you know how much to trust them and how far ahead they hold.
The tool offers several ways to forecast from past data. A trend line fits a straight line through time and extends it forward. A moving average smooths out the bumps by averaging recent points. Exponential smoothing does similar but weights recent data more heavily, and the Holt version adds a trend so the forecast can keep climbing or falling. Each method's accuracy is scored using the average forecast error (MAD and MAPE), so you can see which one fits your data best before trusting its projection.
A warehouse manager has 18 months of order data with a clear seasonal pattern — demand spikes every December and dips every February. A simple average would smooth this out and give a useless forecast.
A time series model fits the trend and the seasonal pattern. Forecast for next December: 1,340 units — 23% above the annual average, consistent with history.
What to do with this: Order stock in November. Time series turns 'we always run out in December' from an annual complaint into a planned, preventable event.
Forecasts drive stock levels, staffing rotas, maintenance windows and capacity decisions. A model that captures trend and reports its own accuracy lets you plan with a known margin of error, reducing both stock-outs and overstock, and flags when a process is drifting rather than steady.
Use time series forecasting whenever you have regularly spaced, time-ordered data and need to anticipate the next few periods — demand, throughput, defect counts, downtime or cycle time. Choose a trend or double-smoothing model when there is a clear up or down movement, and single smoothing or a moving average for a stable series.
In DMAIC, time series methods support the Measure and Control phases — establishing a baseline, detecting drift, and forecasting the impact of changes. They complement control charts, which flag special causes, by quantifying the underlying trend and projecting it forward.
Compare methods on the same series and keep the one with the lowest MAPE that makes sense for your data. Quote the prediction interval, not just the point forecast, so planners know the uncertainty. Monitor the live series with a control chart and re-fit the model as new data arrives, and investigate any large forecast errors as possible special causes or structural changes.
Combine forecasting with control charts, capability analysis and a DMAIC structure to plan, monitor and sustain process performance over time.
A trend model fits a straight (or curved) line to the data over time and projects it forward — simple and good for steadily rising or falling series. Exponential smoothing weights recent observations more heavily and adapts as the series changes; it handles level and trend (and seasonality in fuller versions) without assuming a fixed line. ARIMA is a more general class of models that captures autocorrelation explicitly and usually needs more data and tuning. This calculator covers trend, moving average and exponential smoothing; ARIMA is an advanced option coming soon.
Exponential smoothing forecasts the next value as a weighted average of the current observation and the previous forecast, with weights decaying exponentially into the past. Single exponential smoothing tracks the level with one parameter α (0–1): a higher α reacts faster to change, a lower α gives a smoother forecast. Double (Holt) exponential smoothing adds a trend term with a second parameter β, so it can follow series that are steadily increasing or decreasing rather than producing a flat forecast.
As a rule of thumb, forecast no further than a small fraction of the history you have — short horizons are far more reliable than long ones. Trend and double-exponential models can project several periods ahead but the uncertainty grows quickly, which is why the prediction intervals widen with the horizon. Single exponential smoothing and moving averages produce a flat forecast, so they are best for one or a few steps ahead on series without a trend.
MAD (mean absolute deviation) is the average size of the forecast errors in the original units — lower is better. MAPE (mean absolute percentage error) expresses the same errors as a percentage of the actual values, so you can compare accuracy across series of different scales; a MAPE under about 10% is usually considered very good and under 20% reasonable. Both are computed from the one-step fitted errors, so they describe how well the model tracks the historical data.
A good forecast turns history into a plan. Structured training shows you how to choose models, judge accuracy and act on forecasts with confidence.
More calculators for statistics, quality, and improvement.
Want to build forecasting and trend analysis into a full DMAIC project? The Black Belt course covers advanced statistics, hypothesis testing, and the full analytical toolkit.
View Black Belt →