Lab 5 — GQF-QMF

AI-Powered Fashion Trend Prediction and Inventory Optimization

Zara Demand Forecasting Project

Team Code: GQF-QMF · AI Business World

Readiness Score: 38/100

Business Problem Clarity
0/15
AI Fit
10/10
Data Readiness
15/15
Workflow & HITL
8/20
Risk & Mitigation
0/15
Stakeholder Management
0/10
KPI & Monitoring
0/10
Governance
5/5

Failed Critical Tests

business problem clarityworkflow logicrisk awarenessstakeholder adoptionkpi logic

❌ Connect each KPI directly to the business problem. "What number proves the project worked?"

Business Problem

WHO: Zara's regional inventory managers and supply chain planners across all markets.
HOW OFTEN: Every 2-week fashion cycle (26 seasons per year), with daily demand fluctuations.
COST: Incorrect demand forecasts result in: excess inventory (€2.3B annually in markdowns), lost sales from stockouts (15% of demand unmet), increased storage costs (€180M/year), and accelerated garment disposal (environmental and brand risk).

CURRENT STATE: Zara relies on historical sales data and manager intuition to predict demand 6-8 weeks ahead of each fashion cycle.
DESIRED STATE: AI-driven demand forecasting that integrates social media trends, weather data, search analytics, and customer behavior to predict demand 12+ weeks ahead, enabling proactive manufacturing and distribution decisions.

Goals:
  • Reduce excess inventory by 20% (saving €460M in markdowns)
  • Reduce stockout rate from 15% to 5% (capturing €150M in lost sales)
  • Reduce storage costs by 15% (saving €27M/year)

AI Solution

AI Technology: Ensemble of time-series forecasting models (LSTM, Prophet, XGBoost) combined with NLP analysis of social media trends and fashion blogs.

Data Inputs: Social media trends (Instagram, TikTok), Google search data, weather forecasts (temperature, precipitation), historical POS sales, customer purchase history, seasonal fashion week data, and regional store performance.

Outputs: Demand forecasts per SKU per region/store for 12-week horizon, inventory recommendations (optimal stock levels, reorder triggers), risk alerts for emerging trends, and markdown recommendations for slow-moving inventory.

Process Integration: AI forecasts feed into Zara's existing supply chain planning system, where regional managers review and approve before manufacturing orders are placed.

Human-in-the-Loop:

Override Points: (1) Regional managers can adjust AI forecasts before manufacturing orders are finalized. (2) Supply chain director reviews and approves final production quantities. (3) If AI predicts a demand spike >30% above historical average, manual approval required. (4) Marketing team can input known upcoming campaigns that may affect demand.

Areas for Improvement

Risk & Mitigation (0/15)

Stakeholder & Change Management (0/10)

KPI & Monitoring (0/10)

Suggested Professor Questions

  1. Who exactly has this problem, how often does it happen, and what does it cost?
  2. Why does this need AI rather than a normal dashboard, form, or automation?
  3. What exact data does the system need, where does it come from, and who owns it?
  4. Where exactly does AI enter the real process?
  5. Where can a human stop, correct, or override the AI?
  6. What is the most realistic way this system could fail in week one?