How to Build a Mixpanel Tracking Plan (With Naming Conventions That Last)

I can usually tell within ten minutes of opening a Mixpanel project whether it started with a tracking plan. Without one, the event list reads like a group chat: Sign Up, signup, sign_up_completed, User Signed Up (new). Four events, one action, and no one sure which to trust.

A tracking plan is the cheapest fix in analytics. It costs a few hours up front and saves months of cleanup later.

What a tracking plan actually is

Mixpanel describes a tracking plan as a structured list of the events and properties you’ll send for a feature: what triggers each one, which properties it carries, and which naming rules apply. Think of it as the contract between the people who build features and the people who analyze them.

It doesn’t need special software. Mixpanel itself says a shared spreadsheet or Notion page is a fine place to start.

Start from questions, not clicks

The biggest mistake I see is starting with “what can we track?” Start with “what do we need to know?”

For each business goal, write the questions you’ll ask, then the user actions that answer them. For a SaaS onboarding flow, that might be:

  • What % of signups reach their first key action?
  • Where in onboarding do people drop off?
  • Do users who invite a teammate retain better?

Now the events almost write themselves: sign_up_completed, onboarding_step_completed, teammate_invited.

Get the level of detail right

Mixpanel’s guidance is that events should be neither too broad nor too specific. Define them at the level you’ll actually analyze.

Two examples I use with clients, both drawn from Mixpanel’s own advice:

  • Too specific: separate events for every page (home_page_viewed, pricing_page_viewed). Better: one page_viewed event with a page name property.
  • Too broad: one button_clicked event with a button name property for everything from playing a song to logging out. Better: distinct events for distinct actions, each with its own useful properties.

The rule of thumb: events are actions, properties are details. Mixpanel also suggests keeping your schema to a manageable set of core events and enriching them with properties.

Naming conventions that last

Mixpanel is case-sensitive. sign_up_completed and Sign_Up_Completed are two different events, and that alone causes half the duplicates I find in audits.

The conventions I recommend, which line up with Mixpanel’s guidance:

  • Use snake_case for event names, property names and property values. Mixpanel recommends it, and it travels well into data warehouses and APIs.
  • Object + past-tense verb: checkout_completed, plan_upgraded, report_exported. Consistent structure makes the event list scannable.
  • No dynamic values in event names. Not purchase_2026_10_01. Put the variable in a property.
  • Don’t start custom properties with $ or mp_. Those prefixes belong to Mixpanel’s reserved and default properties.
  • Standardize shared properties such as plan_type or platform so they mean the same thing on every event.
  • Standardize values too. premium and Premium will show up as two separate segments.

What goes in each row

At minimum, every event in the plan needs:

ColumnExample
Event namecheckout_completed
TriggerOrder confirmation returned by payment API
Propertiesorder_id, revenue, currency, item_count, payment_method
Property types and allowed valuespayment_method: card, paypal, bank_transfer
SourceServer
OwnerGrowth team
StatusPlanned / Live / Deprecated

The “trigger” column matters more than people think. “When the user clicks Buy” and “when the payment succeeds” produce very different numbers.

Make it part of how features ship

A plan that lives in one analyst’s drive dies quickly. Mixpanel’s workflow advice, which I fully agree with:

  • Add a tracking section to the feature spec or ticket template.
  • Add “tracking plan complete” to your definition of done.
  • Verify events in staging using Mixpanel’s Events view before release, checking that required properties aren’t missing or malformed.
  • Hold a short monthly data review between product and analytics.

On plans that support it, Mixpanel also offers governance features: Data Standards to define naming rules and required properties at the project level, and Event Approval to catch events that weren’t planned.

Mistakes to avoid

  • Tracking everything “just in case.” More events means more noise, and event volume affects pricing.
  • Letting each platform name things differently. iOS, Android and web should share one plan.
  • Never updating the plan. If it doesn’t match production, people stop trusting it.
  • Skipping the owner column. When an event breaks, someone has to care.

Final thoughts

The best tracking plans I’ve seen weren’t the biggest. They were the ones the team actually used. Start small, name things consistently, and make the plan part of shipping.

Sources