Sign in Start free
ANALYTICS & REPORTING

GA4 exploration spec

Use when you know the question you want answered but not how to build the GA4 exploration for it.

ga4-exploration-spec.md
Download .md
You are a GA4 practitioner. Turn my question into a build spec I can follow in the GA4 Explore interface.

Question I need answered: {{BUSINESS_QUESTION}}
Events and parameters available: {{EVENTS_AND_PARAMS}}
Conversions defined: {{CONVERSIONS}}
Date range: {{DATE_RANGE}}

Output exactly these sections:
1. Exploration type (Free form, Funnel, Path, Segment overlap, Cohort) and one sentence on why that one.
2. Configuration: Dimensions, Metrics, Rows, Columns, Values, Filters, Segments - as a labeled list, using exact GA4 field names.
3. Segment definitions written as conditions, including scope (user, session, or event).
4. What a correct result looks like, so I can tell if I built it wrong.
5. Limits and gotchas that apply to this specific report: cardinality, thresholding, sampling, (not set), session attribution, and any event that is not retroactive.
6. What this exploration cannot tell me, in plain English.

Rules:
- If a dimension or metric I need does not exist in GA4, say so and give the nearest workable substitute rather than inventing a field name.
- Do not suggest BigQuery unless the question genuinely cannot be answered in the interface, and if you do, say why in one sentence.
- No generic advice about setting up GA4. Only this report.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

  • {{BUSINESS_QUESTION}}
  • {{EVENTS_AND_PARAMS}}
  • {{CONVERSIONS}}
  • {{DATE_RANGE}}

Getting a better result

  1. Paste your actual event names from the DebugView or the events report so it stops guessing.
  2. Ask it for the equivalent BigQuery SQL as a second step once the exploration shape is agreed.
  3. Section 6 is the useful one - read it before you present the numbers to anyone.

Questions about this prompt

When do I use this rather than building the exploration and seeing what happens?

When you know the question but not which Explore type answers it, and getting the shape wrong costs an afternoon. It also earns its place before you present anything, because section six lists what the report cannot tell you, and that is far easier to absorb before the numbers than after someone quotes them.

What do I need in front of me?

The business question, your real event and parameter names taken from DebugView or the events report, your defined conversions, and the date range. Pasting actual event names is what stops it guessing at fields. GA4 naming is specific enough that a plausible guess is usually a field that does not exist.

What comes back?

An Explore type with a one line reason, a labelled configuration using exact GA4 field names, segment definitions with scope stated, a description of what a correct result looks like, and the gotchas that apply here: cardinality, thresholding, sampling, not set, session attribution. Section four is what tells you that you built it wrong.

What is the mistake that costs me here?

Forgetting that GA4 events are not retroactive, so a spec depending on a parameter nobody was collecting cannot be run against last quarter. The other trap is reaching for BigQuery by reflex. The prompt only proposes it when the interface genuinely cannot answer the question, and export access has its own lead time.