Unit 6: Data strategy as product strategy

Learning objectives

Understand why data is the foundation of AI product advantage. Evaluate your product's data assets and gaps. Design data collection as a product feature, not an afterthought.

Video script

Reading material

The data audit

Before planning AI features, run a data audit. Answer these questions.

What data do you collect today? List every data type: user actions, content created, transactions, communications, preferences, feedback. Be thorough.

How is it structured? Is the data in a format that's useful for AI training and inference? Or is it scattered across systems, inconsistently formatted, and poorly labeled?

What data do you not collect that you should? Are there user interactions you're not tracking that would be valuable for AI features? Gaps in data collection are gaps in your AI capability.

What data quality issues exist? Duplicates, inconsistencies, missing values, outdated records. Data quality directly affects AI output quality. Bad data makes bad AI.

What data do your users create that could become a shared asset? When users provide feedback, corrections, labels, or preferences, that data can improve the AI for all users. This creates a flywheel effect where more users lead to better AI, which attracts more users.

Designing for data collection

The best data strategies don't feel like data collection to the user. They feel like product features.

Feedback loops. When the AI makes a suggestion, make it easy for users to indicate whether it was helpful. A simple thumbs up/down collects training data while also building user trust.

Implicit signals. Did the user accept the AI's suggestion, modify it, or reject it? Each action is a data point about quality. You don't need to ask for feedback explicitly if the product design captures behavioral signals.

Structured interactions. When users create content in your product, the structure of that content is data. Well-designed forms, templates, and workflows produce cleaner data than freeform input.

Collaborative labeling. Some products turn data labeling into a user feature. "Is this categorization correct?" or "Help us improve by reviewing this suggestion." When done well, users contribute data willingly because the act of reviewing is useful to them.

Practical exercise

Exercise: Data strategy assessment

Choose a product you know well. Conduct a mini data audit.

1. List the major data types the product collects.
2. Identify 2-3 AI features the product could build with this data.
3. Identify 1-2 data gaps that would need to be filled for additional AI capabilities.
4. Describe one product design change that would improve data collection while also improving the user experience (the flywheel principle).

Write up your assessment as a one-page data strategy brief.