Exponential smoothing

A forecast that blends the latest actual with the previous forecast, so weights on older periods decline geometrically.

DefinitionA forecast that blends the latest actual with the previous forecast, so weights on older periods decline geometrically.

Exponential smoothing forecasts the next period as a weighted blend of the most recent actual and the most recent forecast. Every past period still counts, but its weight shrinks the further back it is. Extended versions add a trend term (Holt) and a seasonal term (Holt-Winters); Excel's FORECAST.ETS uses the triple, seasonal version.

How it is computed

Next forecast = alpha x latest actual + (1 - alpha) x latest forecast, with alpha between 0 and 1. A high alpha reacts quickly to change; a low alpha smooths more. Alpha is usually chosen to minimize past forecast error.

Example

With alpha of 0.3, a last forecast of $100,000 and an actual of $120,000, the next forecast is 0.3 x $120,000 + 0.7 x $100,000 = $36,000 + $70,000 = $106,000.

Where it goes wrong

Simple smoothing on a trending series lags behind every period, and an alpha tuned on a quiet year overreacts after a shock. The full guide is the Excel FORECAST function.

Articles that use this term

How-to guides · Finance and FP&A teams

Excel Data Analysis ToolPak: how to enable it, the 19 tools, and one worked run

The Analysis ToolPak is the free Excel add-in behind the Data Analysis button. This page gives the click path to enable it on Windows and Mac, a table of all 19 tools with a finance or sales use for each, one worked Descriptive Statistics run on ten invoices, and the worksheet functions that check every line of the output.

Oct 1, 20266 min read How-to guides · Sales teams

Excel FORECAST function: FORECAST.LINEAR and FORECAST.ETS, with examples

How Excel's forecasting functions work on sales data: FORECAST and FORECAST.LINEAR fit a straight line, worked by hand on eight quarters; FORECAST.ETS uses exponential smoothing with seasonality, run in Excel on 24 months with its confidence interval. Arguments, Forecast Sheet, which to use, and where each goes wrong.

Oct 1, 20267 min read Forecast and pipeline

Predictive analytics: what it is, with examples from finance and sales

Predictive analytics uses statistical models fitted on historical data to estimate a future or unknown value. This page sets it beside descriptive, diagnostic and prescriptive analytics, gives examples from finance and sales, describes the main methods in plain words, and tests a trend forecast on quarters it has not seen, against a naive baseline.

Oct 1, 20266 min read Forecast and pipeline · Sales teams

Sales projection: how to project next year's sales, bottom up and checked

How to build a sales projection for next year from three parts: existing customers at their net revenue retention, new business from qualified pipeline and win rates, and a top-down check against the growth trend. Worked on three segments, with the Excel formulas and the gap that has to be explained.

Oct 1, 20265 min read