CUTS FORECAST ERROR 30% FORECASTING POWER BI

Forecast Accuracy Scorecard

Power BI scorecard breaking down MAPE, BIAS, and MAE by SKU, planner, category, and time horizon — with planner league table, root cause classification, MAPE heatmap, and traffic-light RAG status to drive accountability and targeted improvement.

Project Overview

⚑ Problem

A Single MAPE Nobody Acts On

Most organisations report a single aggregate MAPE — 28%, good enough — on a monthly slide. Nobody knows which SKUs, which planners, or which time horizons are driving error. Without that granularity, improvement programmes are unfocused, accountability is diffuse, and the number rarely moves year on year.

⚙ Solution

Power BI Multi-Dimensional Scorecard

A Power BI model computing MAPE, BIAS (directional error), and MAE at SKU × planner × week × horizon level. Planner league table creates accountability. Root cause classification (over-forecast, under-forecast, lumpy demand, new product) enables targeted intervention. Heatmap surfaces persistent problem categories and weeks.

✓ Findings

MAPE Cut from 28% to 19.6%

Portfolio MAPE improved from 28.2% to 19.6% over 12 weeks — a 30% error reduction. Over-forecasting is the dominant root cause (41% of errors), driven by promotional lift assumptions that consistently overshoot. Two planners account for 58% of total error units despite covering only 35% of SKUs. Dairy category is the single largest error contributor at 34% MAPE.

Planner Accountability Scorecard

Planner League Table — MAPE Ranking

Click a planner to highlight in charts

MAPE Heatmap — Category × Week

Forecast Error Intensity — Darker = Higher MAPE

MAPE:

<10%

10–20%

20–30%

>30%

Dataset Explorer

Accuracy Analysis Charts

MAPE Trend — Portfolio & by Category (12-Week)

BIAS Decomposition — Over vs Under Forecast by Planner

Root Cause Classification

MAPE by SKU — Top Error Drivers

Error Distribution (MAPE Histogram)

MAPE by Forecast Horizon

Weekly Accuracy Register

WEEK	SKU	CATEGORY	PLANNER	FORECAST	ACTUAL	ERROR (UNITS)	MAPE %	BIAS %	MAE	ROOT CAUSE	HORIZON (WK)	RAG

MAPE Improvement Trajectory

Projected Portfolio MAPE — 12-Month Improvement Roadmap

Company Advice

💡 Recommendations for Demand Planning Leaders

MAPE below 15% at SKU level is world-class; most organisations run at 25–35%. Target 20% as a 12-month milestone and treat sub-20% as a competitive advantage in working capital efficiency and service levels.
The planner league table is a management tool, not a blame tool. Use it to identify coaching opportunities and process gaps — the planner with the highest MAPE is often managing the most volatile SKUs or the weakest data quality, not making poor decisions.
BIAS is more actionable than MAPE. Persistent positive bias (chronic over-forecast) inflates inventory and masks actual demand signals. Persistent negative bias (under-forecast) causes stockouts. Target BIAS within ±5% before attacking absolute MAPE.
Root cause analysis must be SKU-level, not aggregate. Over-forecast in promotionals and lumpy demand in industrial require completely different interventions — blending them into a single MAPE improvement programme dilutes both.
The heatmap reveals systematic category-level errors that persist week after week — these indicate model structural issues (wrong seasonality, missing causal variables) rather than one-off events. Structural fixes outperform planner coaching for these categories.
Run a monthly forecast review cadence: top 10 error SKUs reviewed with root cause per planner, compared to prior month. The scorecard only creates value if someone acts on it — the cadence is the governance mechanism.

Key Takeaway: A 1% MAPE improvement across a €50M+ planning portfolio typically frees 0.5–1% of working capital. The scorecard does not improve forecast accuracy — the conversations it forces between planners, sales, and finance do.