# Forecast Value Add (FVA) Worksheet

> Source: https://netray.co/tools/forecast-value-add-worksheet
> Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems.
> Section: Free Tools

## Short answer

Compare naive, statistical and final consensus forecasts against actuals by item to measure whether each step in your forecasting process adds or destroys accuracy.

## Category

supply-chain

## Audience

- Demand planner
- S&OP lead
- Supply chain analyst

## Tool

**Kind:** worksheet

**Mode:** editable

**Columns:**

#### Item / family

**Key:** item

**Type:** text

#### Actual

**Key:** actual

**Type:** number

#### Naive forecast

**Key:** naive

**Type:** number

#### Statistical forecast

**Key:** stat

**Type:** number

#### Final consensus forecast

**Key:** final

**Type:** number

#### Naive abs error

**Key:** naive_err

**Type:** number

**Expr:** abs(actual - naive)

**Format:** number

#### Stat abs error

**Key:** stat_err

**Type:** number

**Expr:** abs(actual - stat)

**Format:** number

#### Final abs error

**Key:** final_err

**Type:** number

**Expr:** abs(actual - final)

**Format:** number

#### FVA of stat model (units)

**Key:** fva_stat

**Type:** number

**Expr:** abs(actual - naive) - abs(actual - stat)

**Format:** number

**Help:** Positive means the statistical forecast beat naive.

#### FVA of overrides (units)

**Key:** fva_final

**Type:** number

**Expr:** abs(actual - stat) - abs(actual - final)

**Format:** number

**Help:** Positive means human overrides improved on the statistical forecast.

**Rows:**

**Item:** Family A

**Actual:** 1240

**Naive:** 1180

**Stat:** 1215

**Final:** 1300

**Item:** Family B

**Actual:** 860

**Naive:** 910

**Stat:** 890

**Final:** 850

**Item:** Family C

**Actual:** 410

**Naive:** 520

**Stat:** 470

**Final:** 600

**Item:** Family D

**Actual:** 5200

**Naive:** 4800

**Stat:** 5050

**Final:** 5100

**Item:** Family E

**Actual:** 180

**Naive:** 150

**Stat:** 170

**Final:** 220

**Item:** Family F

**Actual:** 640

**Naive:** 700

**Stat:** 660

**Final:** 650

**Totals:**

#### Total actual

**Key:** actual

**Agg:** sum

**Format:** number

#### Total naive error

**Key:** naive_err

**Agg:** sum

**Format:** number

#### Total stat error

**Key:** stat_err

**Agg:** sum

**Format:** number

#### Total final error

**Key:** final_err

**Agg:** sum

**Format:** number

#### Total FVA of stat model

**Key:** fva_stat

**Agg:** sum

**Format:** number

#### Total FVA of overrides

**Key:** fva_final

**Agg:** sum

**Format:** number

**Summary:**

#### Naive WMAPE

**Key:** naive_wmape

**Expr:** sum_naive_err / sum_actual * 100

**Format:** percent

#### Statistical WMAPE

**Key:** stat_wmape

**Expr:** sum_stat_err / sum_actual * 100

**Format:** percent

#### Final consensus WMAPE

**Key:** final_wmape

**Expr:** sum_final_err / sum_actual * 100

**Format:** percent

#### FVA of stat model (WMAPE points)

**Key:** fva_stat_pts

**Expr:** naive_wmape - stat_wmape

**Format:** decimal

**Primary:** true

**Description:** Positive means the model beat the naive forecast.

#### FVA of overrides (WMAPE points)

**Key:** fva_final_pts

**Expr:** stat_wmape - final_wmape

**Format:** decimal

**Primary:** true

**Description:** Negative means overrides made the forecast worse.

**Notes:** Use the same actuals, lag and bucket for all three forecasts. The naive forecast is usually the last period's actual (or the same period last year for seasonal items). Measure over many periods before judging a team; a single month is noise.

## Content

**Intro:** Forecast value add asks a blunt question: does each step in your forecasting process make the forecast better than doing less? This worksheet compares a naive forecast, the statistical forecast and the final consensus forecast against actuals, item by item.

**Sections:**

#### How to use the worksheet

**Body:** For each item or family, enter actual demand and three forecasts made at the same lag: the naive forecast, the statistical forecast and the final consensus after overrides. The sheet calculates the absolute error of each, the units of value added or lost by the model and by the overrides, and the WMAPE for each layer. Run it over several months of history, ideally by planner or by product group, then discuss the results at the demand review. Export to Excel to build a monthly trend of FVA.

#### The logic of FVA

**Body:** FVA compares the error of one step with the error of the step before it. The naive forecast is the free benchmark. If the statistical model cannot beat naive, the model needs different parameters or data. If the consensus forecast has higher error than the statistical forecast, the overrides are subtracting value. The measure is simple: error before minus error after. A positive result means the step added value. The idea is popularised by Michael Gilliland in the forecasting community, and it is now a standard way to run forecast process reviews.

**Bullets:**

- Naive: last period or same period last year.
- Statistical: baseline model output.
- Consensus: after sales, marketing and planner overrides.

#### Reading the results

**Body:** With the example rows, the statistical model beats naive by a few points of WMAPE, and the overrides slightly worsen the final forecast in some families because large upward adjustments missed. Do not conclude that overrides never work. Look at where they help: known one-off events, new products and contract changes are often better judged by people, while routine adjustments to stable items usually add noise. Target the review at the largest negative FVA items by volume.

#### Common mistakes

**Body:** Comparing forecasts made at different lags. Judging a person or team on a few weeks of data. Using FVA only to criticise, when its best use is to focus effort where it pays. Skipping the naive benchmark, which leaves no baseline for improvement. And ignoring bias: a layer can lower absolute error while adding a systematic lean, so check the tracking signal for each forecast version as well.

**Cta:** If you want each forecast version saved and scored automatically from your ERP every cycle, we can design a read-only FVA report that updates itself.

**Faq:**

#### What is forecast value add?

**Answer:** It measures whether a step in the forecasting process improves accuracy compared with the previous step. You compare error from a naive forecast, a statistical forecast and the final consensus. Positive FVA means the step reduced error, negative means it increased error.

#### What is a naive forecast?

**Answer:** It is the simplest possible forecast, usually that next period will equal the last actual, or equal the same period last year for seasonal items. It costs nothing to produce, so any model or process that cannot beat it is not adding value.

#### Why would overrides make a forecast worse?

**Answer:** Common reasons are optimism bias, sales targets leaking into the forecast, adjusting stable items out of habit and double counting events already in the baseline. Overrides tend to help most with specific, documented events that history cannot see.

#### How much history do I need for FVA?

**Answer:** At least several months across many items, and ideally a year. A single period is dominated by noise. Group results by planner, product family and size of adjustment to see where the overrides help or hurt, once you have enough observations.

## Search terms

forecast value add, fva forecast value added template, forecast value add worksheet, does forecast override add value, naive forecast benchmark, fva demand planning, forecast process efficiency, statistical forecast vs consensus

---
Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json