GA4 acquisition analysis gets much easier once you ask one precise question first:
Are you trying to understand how people first discovered you, or what brought them back for a particular session?
Use User acquisition for the first question and Traffic acquisition for the second. Then compare complete periods, keep volume and outcome metrics together, and treat the AI’s output as a decision brief – not an automatic verdict on a channel.
This guide gives you a repeatable workflow for doing that in an AI workspace, including the exact reports to request, the mistakes to avoid, and how HireOtto can retrieve the evidence safely.
Table of Contents
User acquisition and Traffic acquisition answer different questions
The names sound similar, but the underlying scope changes what each row means.
Report | Row dimension | Best question | Typical measures |
|---|---|---|---|
User acquisition | First user channel, source/medium, or campaign | How did we acquire new users for the first time? | New users, engaged sessions, key events, revenue |
Traffic acquisition | Session channel, source/medium, or campaign | Which sources started sessions during this period? | Sessions, engaged sessions, engagement rate, key events, revenue |
Google defines User acquisition as user-scoped and Traffic acquisition as session-scoped. A person can first discover you through organic search, then return later through email, direct traffic, or paid social. User acquisition continues to associate that person with the original source; Traffic acquisition describes the source of each session.
That is why two reports can show different results for what appears to be the same channel. The numbers are not necessarily inconsistent. They may be answering different questions.
Use this rule:
Start with First user dimensions when the decision is about acquiring new users.
Start with Session dimensions when the decision is about traffic, engagement, campaign performance, or repeat visits.
Use unprefixed source, medium, and campaign dimensions only when the question is explicitly about attribution for key events. Do not mix them casually with user- or session-scoped dimensions.
Define the decision before pulling data
“Analyze our acquisition” is too broad. It encourages the AI to scan a large table and invent a story.
A useful request names the decision the report should support. For example:
Which session channels gained or lost meaningful traffic versus the previous period?
Which source/medium combinations brought engaged sessions and key events, not just volume?
Which campaigns have enough evidence to investigate, maintain, expand, or fix?
Are changes concentrated in one country, device category, or landing page?
Did new-user acquisition shift even though total sessions stayed flat?
Write the reporting contract before you request rows:
Property: the exact GA4 property ID
Scope: first user or session
Period: complete start and end dates
Comparison: an equal prior period, previous year, or another explicit range
Breakdown: channel group, source/medium, campaign, or a controlled combination
Measures: volume, engagement, business outcomes, and revenue where applicable
Filters: country, device, stream, campaign, or other necessary limits
Output: inline review, CSV, or both
This turns an open-ended “analysis” request into a reproducible report.
Use a layered acquisition analysis
Do not begin at campaign level. Start broad, then drill into the rows that explain the change.
1. Establish the property and measurement context
Confirm the property name and ID. If an organization has development, staging, regional, or client properties with similar names, a name alone is not enough.
Inspect the property’s reporting timezone, data streams, key events, Google Ads links, and custom definitions before interpreting performance. This does not prove the website tag fires correctly, but it prevents obvious context errors such as querying the wrong property or assuming a key event exists.
Also state the business timezone in the brief when it differs from the GA4 property timezone.
2. Run the channel-level report
For session performance, begin with sessionDefaultChannelGroup. A practical baseline is:
Sessions
Engaged sessions
Engagement rate
Key events
Session key-event rate, when available and compatible
Total revenue, when revenue is implemented and relevant
For new-user acquisition, begin with firstUserDefaultChannelGroup and metrics that support the acquisition decision, such as new users, engagement, key events, and revenue. Check the exact combination before running it; a metric’s existence does not guarantee compatibility with every dimension.
Compare two equal, complete periods. Keep both raw values next to the change. “Up 50%” means little when a channel moved from two key events to three.
3. Drill into source/medium
Channel groups are useful for orientation, but they can hide different behaviors.
Expand only the channels that deserve investigation. For session analysis, use sessionSourceMedium. For first-user analysis, use firstUserSourceMedium.
At this level, look for:
One source driving most of a channel’s gain or decline
Inconsistent medium naming that fragments one source across rows
Referral traffic that should be excluded or classified differently
(direct) / (none) growth that may reflect genuine direct demand, lost campaign parameters, redirects, consent effects, or other attribution gaps
(not set) values that require a data-quality check before a performance conclusion
Do not automatically combine rows because their names look similar. linkedin / paid_social, linkedin.com / referral, and linkedin / cpc may represent different tagging choices or genuinely different traffic.
4. Drill into campaign only after source/medium
Use the campaign breakdown to explain a source/medium change, not as the first table you open.
Review:
Session or first-user campaign, matching the chosen scope
Sessions or new users
Engaged sessions and engagement rate
Key events and rate-based outcomes
Revenue where trustworthy
Use campaign IDs where possible when names are duplicated or renamed. For manually tagged links, campaign fields depend on consistent UTM parameters. For Google Ads traffic, linked-account and auto-tagging issues can prevent expected Ads dimensions from appearing.
5. Segment only when it tests a hypothesis
Adding country, device, landing page, and date to every request creates a large table that is hard to interpret and more likely to hit reporting limits.
Add one diagnostic dimension when it tests a specific explanation:
Country: Did a channel change because delivery shifted into a different market?
Device category: Is weaker engagement concentrated on mobile?
Landing page: Did the campaign send traffic to a different or weaker destination?
Date: Was the change sustained or caused by one spike?
If the first segment does not explain the movement, return to the hypothesis rather than stacking more dimensions.
6. Convert findings into four decision queues
An acquisition report should end with actions, not observations.
Queue | Evidence pattern | Appropriate next step |
|---|---|---|
Protect | Meaningful volume and dependable downstream outcomes | Maintain coverage; monitor for tracking or delivery changes |
Investigate | Material movement with incomplete explanation | Drill into source/medium, campaign, landing page, country, or device |
Test | Promising engagement or outcomes with enough volume to learn | Propose a controlled budget, message, landing-page, or targeting test |
Fix measurement | (not set), suspicious referral/direct shifts, broken naming, missing key events, or incomplete links | Validate tagging, UTMs, redirects, consent, events, and platform links before optimizing spend |
The fifth valid result is insufficient evidence. Low-volume rows should stay unclassified until more data arrives or the analysis uses a longer period.
A copyable AI prompt
Use this as a review-first brief. Replace the placeholders and remove any metrics that are not relevant to your property.
For GA4 property PROPERTY_ID, analyze acquisition for the last 28 complete days (28daysAgo through yesterday) against the preceding 28 complete days (56daysAgo through 29daysAgo).
First confirm the property and inspect its timezone, data streams, key events, Google Ads links, and custom definitions. Search the property metadata for current acquisition dimensions and the metrics needed below. Check the exact field combinations before running reports.
Run two separate analyses:
1. Traffic acquisition by session default channel group, then drill into session source/medium and session campaign only for material movements.
2. User acquisition by first user default channel group, then drill into first user source/medium only where it changes the interpretation.
Include volume, engagement, key events, key-event rate, and revenue when those fields are implemented and compatible. Keep both period values and denominators visible. Do not rank tiny samples by rate alone. Flag (not set), (direct) / (none), fragmented UTMs, missing expected Ads dimensions, partial configuration reads, truncated results, and any thresholding or (other) rows.
Separate facts from hypotheses. Return four queues: Protect, Investigate, Test, and Fix measurement. Do not recommend changing spend from GA4 evidence alone; state what advertising-platform or CRM evidence is still needed. Return the top rows inline and a CSV for the detailed tables.This prompt deliberately asks for two separate reports. It prevents the AI from treating first-user acquisition and session performance as interchangeable.
How to read common acquisition patterns
Sessions are up, but engagement and key events are down
The channel may have gained lower-intent traffic, changed its landing-page mix, or suffered a tracking problem. Drill into source/medium and campaign, then compare landing pages and devices. Do not label the entire channel “bad” until you know where the mix changed.
New users are flat, but sessions are up
Existing users may be returning more often. That can be healthy for retention or remarketing even if top-of-funnel growth is unchanged. Compare User acquisition with Traffic acquisition instead of expecting them to move together.
Key-event rate improved while key events fell
The rate may have improved because traffic volume fell. Keep the numerator and denominator visible. A higher rate does not compensate automatically for fewer qualified outcomes.
Revenue is strong but key events look weak
Check what counts as a key event and whether ecommerce revenue is implemented consistently. Revenue and key events can describe different business behaviors. Do not infer a tracking error from the mismatch alone.
Direct traffic or (not set) increased
Treat this as a diagnostic signal, not proof of a channel shift. Review campaign parameters, redirect behavior, consent implementation, referral exclusions, Ads links, auto-tagging, and the timing of data processing.
A small channel has an extreme rate
Do not promote or cut it from the percentage alone. Set a minimum evidence threshold appropriate to the business, then review raw sessions, users, key events, and revenue.
Checks that keep the analysis honest
Use complete periods
Current-day data is partial. GA4 also processes intraday and daily data on different timelines, and reported values can change after daily processing. For routine comparisons, end at yesterday and record the property timezone.
Key-event attribution can continue changing after the event is recorded. Treat very recent conversion comparisons as provisional, especially when the decision is consequential.
Keep scope in the column names
Write “Session source / medium,” not just “Source / medium.” Write “First user campaign,” not just “Campaign.” The prefix is part of the meaning.
Inspect coverage and limits
Check total matching rows, collected rows, inline rows, export limits, offsets, and whether the result is truncated. A clean preview does not guarantee the CSV contains every matching row.
GA4 can also consolidate high-cardinality values into an (other) row. Privacy thresholds can withhold some data. Missing rows are not always zero.
Separate reporting evidence from causal claims
An acquisition report is observational. A channel associated with more key events did not necessarily cause the increase. Brand demand, campaign mix, seasonality, site changes, consent, and attribution rules can all affect the result.
Budget changes need additional evidence from the ad platform and, where relevant, CRM or revenue systems. GA4 alone does not establish lead quality, incrementality, or marginal return.
Treat naming problems as measurement work
An AI can identify fragmented UTMs and suspicious rows, but a person should decide the canonical taxonomy and remediation plan. Renaming future campaigns does not rewrite historical acquisition data.
Run this workflow in HireOtto
HireOtto brings GA4 evidence into Claude, ChatGPT, and other remote MCP-capable AI clients without requiring you to rebuild every question in the Analytics interface.
For this workflow, HireOtto can:
List the GA4 accounts and properties available to the connected Google login.
Inspect property details, streams, key events, Google Ads links, and custom definitions, while exposing partial-section failures instead of treating them as empty configuration.
Search the property’s current reporting metadata for session, first-user, engagement, key-event, campaign, and revenue fields.
Check whether the exact Core-report dimensions and metrics are compatible.
Run equal-period channel, source/medium, and campaign reports with filters and ordering.
Return a review-sized inline sample and a CSV, with coverage and truncation information.
HireOtto’s GA4 beta is read-only. It does not edit properties, repair UTMs, change key events, publish tag fixes, validate browser execution, judge CRM lead quality, or make budget changes. A human still needs to confirm the business question, review anomalies, validate tracking where necessary, and approve any action in the advertising or analytics stack.
See the GA4 tools reference and custom reporting guide for the exact reporting workflow, then use the GA4 quickstart and AI-client connection guide to get started.
Frequently asked questions
What is the difference between GA4 User acquisition and Traffic acquisition?
User acquisition uses first-user dimensions to explain how new users were originally acquired. Traffic acquisition uses session dimensions to explain what started each session, including visits from returning users.
Which GA4 acquisition report should marketers use most often?
Use Traffic acquisition for recurring channel and campaign reviews. Use User acquisition when the decision is specifically about acquiring new users. Many useful reviews need both reports, kept separate.
Can AI explain why a GA4 channel changed?
AI can identify where the change is concentrated and propose hypotheses. It cannot prove causation from an acquisition table alone. Use platform delivery data, landing-page evidence, tracking checks, and CRM outcomes to test the explanation.
Should I use current-day data?
Use realtime or current-day data for monitoring, not for stable period comparisons. Prefer complete periods for performance decisions and state when recent key-event attribution is still provisional.
Does a high key-event rate mean a channel deserves more budget?
Not by itself. Review the raw number of sessions and key events, the value and quality of those events, advertising cost, attribution differences, and whether the sample is large enough to act on.
Can HireOtto change my GA4 property?
No. HireOtto’s GA4 integration is read-only. It can inspect configuration and retrieve reports, but it cannot edit GA4 settings or publish tracking changes.
About Me
I’m Suyash – badminton junkie, ex‑GroupM ad‑ops grunt, first marketer at a B2B SaaS startup, and creator of Hiretto: MCP servers for performance marketers.
My mission: less clicking, more thinking.
Let’s build leverage together.