Most product analytics set-ups are a bit of a mess, and it's nobody's fault. Events get added in a hurry before a launch. Someone leaves and takes the naming logic with them. A redesign changes what a button does but the event keeps its old name. Two years on, you have 400 events, and when someone asks a simple question about sign-up conversion, nobody is sure which of the three "signup" events to use.
An audit sorts this out. I've run several, and the process is similar each time. You don't need special software. You need a few hours, a spreadsheet and a willingness to open the product and click around.
1. Pull the full event list
Export every event name with its volume over the last ninety days. In Mixpanel the Lexicon is the place to start. Paste it into a spreadsheet and add four columns: what it should mean, who owns it, whether it's still used, and notes.
You'll almost certainly find events with zero or tiny volume. They are candidates for hiding or dropping. You'll also find a handful that make up most of your traffic. Those need the most care.
2. Look at the names
Are they consistent? A healthy list follows one pattern, such as "Object Action" (Order Completed, Video Played) in one tense and one case. A messy one has "signup", "Sign Up Clicked", "user_registered" and "RegistrationDone" side by side.
Don't rename everything at once. Pick the convention, apply it to new events, and plan a clean-up for the old ones. Renaming live events breaks historical charts, so it needs thought.
3. Check the properties
Events are only as useful as the properties attached. For each key event, ask what you would want to split it by. Plan or tier. Platform. Source. Product category. Price. Whether the property is present on every occurrence, or only some.
Look for properties that are the same thing under different names ("plan" and "plan_name") and for properties that hold messy values ("Free", "free", "FREE", "free plan"). Those splits quietly ruin reports.
4. Test identity
This is where I find the worst problems. Identity is how the tool decides that the anonymous visitor on Monday and the logged-in user on Tuesday are the same person.
Check what happens at sign-up and login. Does the tool merge the anonymous history into the user? Does it do that on both web and app? What happens when someone logs in on a second device, or logs out and a different person logs in on a shared one? A small mistake here inflates user counts and breaks funnels.
5. Walk the product and compare
Open the product with the live event view beside you and go through the main journeys. Sign up. Search. Add to basket. Pay. Cancel. For each step, does an event fire? Does it fire once? Does it carry the right properties?
I usually find three kinds of problems. Missing events, where a step in the journey isn't tracked at all. Duplicate events, where something fires twice, often on page refresh or when a component re-renders. And wrong events, where the name says one thing and the code does another.
6. Reconcile with a source you trust
Pick a number you can check against your database or payment system, such as orders last week, and compare it with the count of the matching event. A small gap is normal because of ad blockers and dropped connections. A gap of 20 percent means something is broken. I'd rather find that out in an audit than from a board meeting.
7. Check for things that shouldn't be there
Look through property values for personal information that shouldn't be in an analytics tool: full email addresses in event properties, phone numbers, free-text fields where users might type anything. Fix this early. It is easier to stop collecting than to clean up.
8. Write down what you found
The output of the audit is a document, not a feeling. I use a simple table.
| Event | Meaning | Key properties | Issue | Fix | Priority |
|---|---|---|---|---|---|
| Order Completed | Payment confirmed | value, currency, items | Fires twice on refresh | Guard against reload | High |
| signup | Account created | method | Name clashes with "Sign Up Clicked" | Rename to Account Created | Medium |
Rank the problems by how much they distort decisions. Missing key events and broken identity go first. Naming tidy-ups come last.
9. Make it stick
An audit is wasted if the mess returns in a year. Agree a short tracking plan: the naming convention, the list of core events and properties, and a rule that new features don't ship without tracking agreed in advance. Put a person in charge of the plan.
And a small thing that helps more than it sounds: add event checks to your release checklist. Thirty seconds of testing before a release saves a month of cleaning later.