Learning Goal: Assess an organisation’s data maturity and prepare datasets for AI projects.

Complete Lesson Content:
AI is only as good as its data. Fashion companies sit on goldmines: sales history, customer profiles, product images, body measurements, supplier data. But often this data is siloed, inconsistent, or unstructured. Before launching an AI project, you must: audit available data, clean and label it, ensure diversity to avoid bias, and establish data governance. This lesson provides a practical framework using Excel and a data audit checklist.

Definitions:

  • Structured Data – organised in rows and columns (e.g., sales transactions).
  • Unstructured Data – images, PDFs, social media posts, which constitute most fashion data.
    Step‑by‑step Explanation:
  1. Identify the AI goal (e.g., predict next season’s bestselling colour).
  2. List all potential internal data sources (historic sales, fabric orders, social mentions).
  3. Evaluate data quality: completeness, accuracy, consistency.
  4. Clean data: remove duplicates, correct errors.
  5. Label data if needed (e.g., tag images with attributes).
  6. Ensure privacy and security compliance.

Best Practices: Start small with a pilot dataset that is clean and well‑understood.

Common Mistakes: Using data that is heavily biased (e.g., only past bestsellers) which reinforces existing patterns and misses emerging trends.

Real Industry Example: A denim brand built an AI defect‑detection system using 50,000 labelled images of jeans, requiring a huge data labelling effort but yielding 99% accuracy.

Mini Case Study: 

A startup wanted to build a size recommendation AI but only had customer‑reported sizes, which were often wrong. They pivoted to 3D body scan data and improved accuracy dramatically.

Practical Activity: Using a sample dataset (provided), identify data quality issues and suggest cleaning steps.

Assignment: Conduct a mini data audit for a hypothetical fashion brand (or your own if applicable) and write a one‑page Data Readiness Report.

Lesson Summary: Data strategy is the first step in any AI project; poor data leads to poor AI.

Key Takeaways:

  • Audit data before starting.
  • Clean, label, diversify.
  • Data governance is essential.
PreviousNext