FORECAST.ETS function
Returns a future value based on existing (historical) values by using the AAA version of the Exponential Smoothing (ETS) algorithm
Compare five original sources using saved native Excel forecasts, intervals and diagnostics. Download the formulas to edit and recalculate in desktop Excel. The browser selects saved results.
Syntax
FORECAST.ETS(Target date, Values, Timeline, [Seasonality], [Data completion], [Aggregation])Square brackets indicate optional arguments.
Arguments
| Argument | Required | Meaning |
|---|---|---|
| Target date | Yes | The data point for which you want to predict a value.A10 |
| Values | Yes | The historical values for which you want to forecast the next points.B2:B9 |
| Timeline | Yes | The independent range or array of numeric dates or times with a constant step.A2:A9 |
| Seasonality | No | Optional. The seasonal length. Use 1 for automatic detection (default) or 0 for no seasonality.1 |
| Data completion | No | Optional. How to handle missing points. Use 1 to interpolate (default) or 0 to treat them as zero.1 |
| Aggregation | No | Optional. A value from 1 through 7 that specifies how to aggregate duplicate time stamps.1 |
Compare native Excel results
Five original sources, calculated in native Mac Excel on 2026-10-17. Changing the choices below selects saved measured outputs. Use the source builder below to prepare your own series, then recalculate the downloaded formulas in supported desktop Excel.
A linear rise includes the repeated component 3, -1, 5, -7. Excel detected period 2 automatically in this sample. The explicit four-step model fits these generated observations exactly.
Forecast at this target
Native observation · saved to the displayed numerical precision
=FORECAST.ETS(C2,$B$2:$B$33,$A$2:$A$33,4)Read the formula, one argument at a time
Explore the written function calls and their inputs. The calculation result belongs to the complete formula.
FORECAST.ETS · top-level call
=FORECAST.ETS(C2,$B$2:$B$33,$A$2:$A$33,4)A zero fitted error or radius in an exact generated pattern does not rule out future changes. The chart connects the saved values; this page does not fit an ETS model in the browser.
Eight future points
| Target (column C) | Forecast | 95% radius | Lower | Upper |
|---|---|---|---|---|
| C2: 33 | 39 | 0 | 39 | 39 |
| C3: 34 | 35.5 | 0 | 35.5 | 35.5 |
| C4: 35 | 42 | 0 | 42 | 42 |
| C5: 36 | 30.5 | 0 | 30.5 | 30.5 |
| C6: 37 | 41 | 0 | 41 | 41 |
| C7: 38 | 37.5 | 0 | 37.5 | 37.5 |
| C8: 39 | 44 | 0 | 44 | 44 |
| C9: 40 | 32.5 | 0 | 32.5 | 32.5 |
Model diagnostics
Automatic period measured for this source: 2. That result is separate from the explicit model choice.
| Statistic type (column I) | Meaning | Native value |
|---|---|---|
| 1 | Alpha: level | 0.751 |
| 2 | Beta: trend | 0.001 |
| 3 | Gamma: seasonality | 0.001 |
| 4 | MASE | 0 |
| 5 | SMAPE | 0 |
| 6 | MAE | 0 |
| 7 | RMSE | 0 |
| 8 | Timeline step | 1 |
Original source and cell addresses
| Time (A) | Observed (B) |
|---|---|
| A2: 1 | B2: 23 |
| A3: 2 | B3: 19.5 |
| A4: 3 | B4: 26 |
| A5: 4 | B5: 14.5 |
| A6: 5 | B6: 25 |
| A7: 6 | B7: 21.5 |
| A8: 7 | B8: 28 |
| A9: 8 | B9: 16.5 |
| A10: 9 | B10: 27 |
| A11: 10 | B11: 23.5 |
| A12: 11 | B12: 30 |
| A13: 12 | B13: 18.5 |
| A14: 13 | B14: 29 |
| A15: 14 | B15: 25.5 |
| A16: 15 | B16: 32 |
| A17: 16 | B17: 20.5 |
| A18: 17 | B18: 31 |
| A19: 18 | B19: 27.5 |
| A20: 19 | B20: 34 |
| A21: 20 | B21: 22.5 |
| A22: 21 | B22: 33 |
| A23: 22 | B23: 29.5 |
| A24: 23 | B24: 36 |
| A25: 24 | B25: 24.5 |
| A26: 25 | B26: 35 |
| A27: 26 | B27: 31.5 |
| A28: 27 | B28: 38 |
| A29: 28 | B29: 26.5 |
| A30: 29 | B30: 37 |
| A31: 30 | B31: 33.5 |
| A32: 31 | B32: 40 |
| A33: 32 | B33: 28.5 |
The formula workbook contains five source sheets and all 325 measured formula results. Desktop Excel can recalculate its formulas; the frozen workbook contains no formulas. This Mac's source-edit/save/reopen workflow and Google Sheets remain unverified. No source data is uploaded, and these native examples do not count as flat browser-engine sample passes.
Build a forecast workbook from your own series
Prepare the source here, then calculate the forecast in desktop Excel. These controls create genuine ETS formulas; this browser does not fit a model or show predicted results for your edits.
Use 8–100 rows without a header and at least eight unique positive numeric times. The smallest observed spacing defines the regular step; whole-step gaps may occupy up to 30% of at most 142 positions. Calendar dates and irregular monthly spacing are outside this numeric exercise.
Duplicate times are combined here before export. Count and counta are identical for numeric-only input. Fixed seasonality requires at least two periods in this bounded exercise. Missing slots remain absent from the observations; Excel performs the selected completion.
Applied: 24 original rows → 24 unique observations, 0 missing slots, step 1, 8 future targets.
Inspect applied observations and missing slots
| Numeric time | Prepared observed value | Source records |
|---|---|---|
| 1 | 12 | 1 |
| 2 | 14 | 1 |
| 3 | 16 | 1 |
| 4 | 18 | 1 |
| 5 | 20 | 1 |
| 6 | 22 | 1 |
| 7 | 24 | 1 |
| 8 | 26 | 1 |
| 9 | 28 | 1 |
| 10 | 30 | 1 |
| 11 | 32 | 1 |
| 12 | 34 | 1 |
| 13 | 36 | 1 |
| 14 | 38 | 1 |
| 15 | 40 | 1 |
| 16 | 42 | 1 |
| 17 | 44 | 1 |
| 18 | 46 | 1 |
| 19 | 48 | 1 |
| 20 | 50 | 1 |
| 21 | 52 | 1 |
| 22 | 54 | 1 |
| 23 | 56 | 1 |
| 24 | 58 | 1 |
Formulas for the applied source
Forecast A/B holds unique observations; C holds future targets. Controls B2/B3/B4 supplies seasonality, confidence and completion; B6 is aggregation 1 because exported times are already unique.
D2: =FORECAST.ETS(C2,$B$2:$B$25,$A$2:$A$25,Controls!$B$2,Controls!$B$4,Controls!$B$6)
E2: =FORECAST.ETS.CONFINT(C2,$B$2:$B$25,$A$2:$A$25,Controls!$B$3,Controls!$B$2,Controls!$B$4,Controls!$B$6)
K2: =FORECAST.ETS.SEASONALITY($B$2:$B$25,$A$2:$A$25,Controls!$B$4,Controls!$B$6)
J2: =FORECAST.ETS.STAT($B$2:$B$25,$A$2:$A$25,H2,Controls!$B$2,Controls!$B$4,Controls!$B$6)
The workbook retains original row order in Raw, prepared observations in Forecast, and settings in Controls. Forecasts, intervals and fitting diagnostics have no saved result caches. Open and recalculate in a supported desktop Excel version; lower/upper endpoints are point ± radius, not guaranteed future bounds. No source upload or browser persistence is used. Native source editing, saving and reopening on this Mac, and Google Sheets compatibility, remain unverified.
Reference and compatibility
Univer 1.0.4 / Apache-2.0. Calculation checks cover the supplied examples only. The browser worksheet does not implement every Excel feature, external service or connected data type.