How do I choose events and attributes before standardizing tracking?

Before standardizing tracking, choose events and attributes by working backward from the decisions you want to make with the data.

1. Start with business questions

Ask:

• What do we need to measure?
• What actions do we want to optimize?
• What decisions will this data support?

Examples:

• “How many users start and complete onboarding?”
• “Which features drive retention?”
• “Where do users drop off in checkout?”

If a question won’t change a decision, it’s probably not worth tracking yet.

2. Identify the key user journeys

Map the main flows in your product:

• Sign up / onboarding
• Activation
• Search / discovery
• Purchase / conversion
• Retention / engagement
• Support / cancellation

Then define the critical steps in each journey. Those are usually your core events.

3. Prefer action-based events

Track meaningful user actions, not every UI interaction.

Good event examples:

• Signed Up
• Created Project
• Started Checkout
• Completed Purchase
• Invited Teammate

Less useful:

• Clicked Button
• Hovered Card
• Page Viewed for every page without context

If you do track generic actions, make sure they’re tied to a clear purpose.

4. Define attributes that explain “who, what, and context”

For each event, choose attributes that help answer:

• Who did it? user type, plan, role, account
• What happened? item, feature, step, outcome
• In what context? device, channel, page, experiment, locale

Examples for Completed Purchase:

• plan_type
• currency
• price
• payment_method
• coupon_used
• experiment_variant

Only include attributes you expect to analyze or segment by.

5. Separate event-level and user-level data

• Event attributes: describe the specific action
• User/account attributes: describe the person/account overall

Example:

• Event: Upgraded Plan
• Event attributes: from_plan, to_plan, upgrade_source
• User attributes: industry, company_size, role

Avoid duplicating user-level fields on every event unless needed for analysis.

6. Keep the first version small

Standardization works best when you begin with a focused set:

• 10–20 core events
• A small, stable set of attributes
• Clear naming conventions

You can always add more later, but too many events upfront create inconsistency and maintenance problems.

7. Check each event/attribute against these filters

For every candidate, ask:

• Is it tied to a meaningful decision or metric?
• Will it be used more than once?
• Is it precise and unambiguous?
• Can we reliably collect it?
• Is it stable enough to standardize?
• Does it duplicate something else?

If the answer is mostly no, leave it out.

8. Document definitions before implementation

For each event and attribute, define:

• Name
• Purpose
• When it fires
• Required vs optional attributes
• Allowed values / format
• Owner

This prevents inconsistent tracking across teams.

9. Design for analysis, not just collection

Think ahead to how you’ll use the data:

• Funnel analysis
• Cohort analysis
• Segmentation
• Attribution
• Feature adoption

If an event can’t support a useful analysis, reconsider whether it belongs in the standard plan.

10. Use a naming convention

Pick a consistent pattern early:

• Verb + object: Project Created, Trial Started
• Past tense or present tense, but consistent
• Clear, human-readable names

For attributes:

• Use lowercase snake_case or another standard
• Avoid synonyms (plan, subscription_plan, tier — pick one)

---

Simple rule of thumb

Track events when:

• a user takes an important action
• the action contributes to a key funnel or outcome
• the event will be used for reporting or optimization

Track attributes when:

• they add meaningful segmentation or context
• they help explain variation in outcomes
• they’re stable and consistent enough to standardize

If you want, I can also give you a template for selecting events and attributes or a sample tracking plan spreadsheet structure.