Custom definitions, scopes and core metrics

0:00 Custom definitions and metrics

Here's a frustrating moment every G A four user meets. Your developer sends a perfect parameter called author on every page view. You open the reports, and it isn't there. Anywhere. Nothing's broken. You've just discovered that collection is not reporting. In this lecture you'll learn how custom definitions work, how to choose the right scope, how to control cardinality so the dreaded other row doesn't swallow your data, the core metric definitions you must be able to explain, and how to audit definitions with the Data A P I.

0:39 Why it matters

Why does this matter? Because registration isn't retroactive. Data collected before you register a custom dimension won't appear under that dimension later. Register three weeks after launch, and you've lost three weeks. And definition slots are limited, with higher limits on the paid three sixty tier. So registering the right parameters, with the right scope, on day one, is one of the highest-leverage things an analyst does. It's also one of the most commonly forgotten.

1:12 Scopes

Here's the core idea: scope. Every custom definition describes something at a particular level. Event scope describes a single event occurrence, like form id, video title or author. User scope describes the user across events, like membership tier or customer type, and comes from a user property. Item scope describes a product inside the e-commerce items array, like color or size band. Think of it like labels in a supermarket. A label on the receipt, a label on the loyalty card, and a label on each product. Put the label in the wrong place, and it describes the wrong thing.

1:55 The classic mistake

The classic scope mistake: registering customer type as an event-scoped dimension. Then it only appears on events where it was explicitly sent, not across the user's whole behavior. So your report of purchases by customer type shows mostly not set. If an attribute describes the person, send it as a user property and register it with user scope. If it describes the action, it's event scope. And for product attributes inside the items array, use item scope.

2:28 Custom and calculated metrics

Custom metrics and calculated metrics are the numeric side. A custom metric registers a numeric parameter, like quote amount or reading time in seconds, so it can be summed or averaged. A calculated metric is a formula over existing metrics, like revenue per engaged session. Use calculated metrics when stakeholders keep computing the same ratio by hand in spreadsheets. It reduces errors and makes the definition visible to everyone.

2:58 Cardinality and (other)

Now cardinality, the number of unique values a dimension has. High-cardinality dimensions, like full URLs with query strings, unique ids, timestamps or free text, can make G A four group less common values into a row labeled other in some reports. To reduce the risk, band continuous values, like order value band fifty to one hundred. Strip unneeded query parameters. Keep ids out of general dimensions. And use explorations or the BigQuery export for granular analysis, where the other row doesn't apply.

3:34 Core definitions

And you must be able to explain the core metrics in plain language. Active users: distinct users who had an engaged session or certain first-time events, which is G A four's default users in most reports. New users: first-time visitors. Sessions. Engaged sessions. Engagement rate. Average engagement time, when the page was in focus. Key events. Session key event rate, the share of sessions with at least one key event. And total revenue, net of refunds depending on setup. Always name the metric precisely, active users, not just users.

4:13 Example 1: Riyadh news site

First example, a simple one. A news site in Riyadh wants to know which authors drive subscriptions. They add author and content group as parameters on page view, and register both as event-scoped custom dimensions. They register subscriber status, free, trial or paid, as a user-scoped dimension from a user property. Then, in a free-form exploration, rows are authors, columns are subscriber status, values are engaged sessions and subscribe key events. Now the editor can see which authors attract engaged readers who later subscribe.

4:50 Example 2: an online academy

Second example, a business case. An online academy sends course slug, lesson type and content language as event parameters on page view and on a custom lesson complete event. It sends learner plan, free or pro, as a user property. On launch day it registers three event-scoped dimensions and one user-scoped dimension. Within a month, one exploration shows which courses have high completion but low upgrade rates. That becomes the input to the next pricing experiment.

5:23 Watch me do it, part 1

Watch me audit definitions with the Data A P I. In the lesson's Python script, I create a client using application default credentials, so no keys in code. I call get metadata for the property and filter dimensions whose names start with custom event or custom user. That's my list of registered definitions. Then I run a small report: the author dimension with engaged sessions for the last seven days, limited to ten rows. I print the rows.

5:57 Watch me do it, part 2

The output shows the registered list, and the report shows a handful of authors, but most engaged sessions fall under not set. That tells me the definition is registered, but the parameter isn't being sent on every page view, probably only on article templates. I check with the developer, and it turns out the new video pages use a different template. They add the parameter, and a week later not set drops to a sliver. The A P I turned a vague feeling into a specific fix.

6:35 Definitions register

Keep a definitions register next to your tracking plan: name, scope, parameter, created date, owner, and where it's used. Treat slots as scarce, and retire unused definitions deliberately. It's a two-minute habit that saves hours when someone asks, six months later, why a dimension exists and whether it's safe to remove.

6:57 Common mistakes

Common mistakes. Registering definitions weeks after launch and losing early data. The wrong scope, especially user attributes registered as event scope. Using exact monetary amounts or ids as dimensions and then wondering why reports show other. And comparing users in G A four with visitors in another tool without explaining the definitions.

7:20 Recap

Recap. Custom parameters must be registered to be reported, and registration isn't retroactive. Choose the right scope: event, user or item. Control cardinality by banding and keeping ids out. Explain metrics precisely, especially active users. Keep a definitions register. And use the Data A P I, or BigQuery, to audit what's actually arriving.

7:43 Try this now

Try this now. Create a definitions register for a site you know with at least four custom dimensions across event, user and item scopes, and justify each scope in one line. If you have API access, run the lesson's script to see what's registered and whether values are arriving, and investigate the biggest not set.