← AIAP Tool GuidesAll programsHomeSearch
AIAP Tool Guides·Foundations of Google Ads·2:05:07

Google Analytics 4 for Marketers: First-Party Measurement, Data-Driven Attribution, the Event Model and Funnel Exploration

Kaushik Sandeka Marketing strategist at Google (~3 years, UK/EMEA advertisers), ~8-9 years in digital marketing (CleverTap, Acko-type insurer retention); Growth School performance-marketing mentor

The short version

  1. GA is pitched as the free "source of truth" built on consented, first-party and modelled data; linking it to Google Ads is claimed to lift conversions 23% and cut cost per conversion 10% (0:08-0:24).
  2. Attribution: last click gives the striker all the credit; data-driven attribution (DDA) shares it by contribution - shown in Google Ads and GA model comparison (0:36-0:56).
  3. GA4 is event-based (events + user properties + IDs); live demo of reports, comparisons, retention, traffic acquisition filters, funnel exploration and segment overlap, ending with UTM parameters (0:56-2:04).

At a glance, three clicks deep

Skim here first: the closed row is the glance, open is the study card with the key points and timestamps, and the ↓ link drops to that concept's full write-up below.

01GA as source of truth: consented, first-party, modelled dataGA4 is designed around consent, first-party data and modelling to fill cookie gaps;›

GA4 is designed around consent, first-party data and modelling to fill cookie gaps; linked to Google Ads it feeds bidding and audiences.

Three levers (0:09-0:13)

Google study: linked accounts +23% conversions, -10% cost per conversion (0:14)

BigQuery and audience activation (0:21-0:23)

Predictive metrics (0:25-0:30)

↓ Full write-up of this concept

02Last click versus data-driven attributionDDA gives each touchpoint fractional credit by how much it raised the chance of conversion;›

DDA gives each touchpoint fractional credit by how much it raised the chance of conversion; last click gives the final click everything.

Definition and analogy (0:37-0:42)

"78%" of EMEA advertisers on DDA (0:45)

Google Ads demo (0:46-0:50)

GA model comparison (0:50-0:56)

↓ Full write-up of this concept

03Events, user properties and identityGA4 records what users do as events and who they are as user properties, deduplicating people across device…›

GA4 records what users do as events and who they are as user properties, deduplicating people across devices with user ID, device ID and Google signals.

Events vs properties (0:57-0:59)

gtag vs Tag Manager (1:00-1:02)

Three identifiers and dedup (1:05-1:07)

25 user properties (1:08)

↓ Full write-up of this concept

04Reading GA4 reports with comparisons and filtersAdd a comparison (paid vs all users, organic vs paid) and filters (channel contains "paid", source/medium,…›

Add a comparison (paid vs all users, organic vs paid) and filters (channel contains "paid", source/medium, campaign) to turn standard reports into answers.

Real-time and snapshot (1:19-1:21)

Comparisons and events (1:20-1:29)

Retention cohorts (1:31-1:34)

Acquisition filters (1:38-1:44)

↓ Full write-up of this concept

05Explore: funnel exploration and segment overlapFunnel exploration counts each step and the drop-off between them;›

Funnel exploration counts each step and the drop-off between them; breakdowns, filters and elapsed time show who drops and how long they take.

Funnel (1:51-1:57)

Segment overlap (1:58-2:01)

UTM parameters (2:01-2:03)

↓ Full write-up of this concept

The concepts in full

01

GA as source of truth: consented, first-party, modelled data

Third-party data is going away; whatever you collect yourself, with consent, is what you will measure with.

Three levers: respect consent and regional rules such as GDPR, ingest first-party data from web, app and offline (a store associate signing the shopper in on an iPad ties the store visit to the same identity), and use modelling where cookies are refused. Collect and enrich, derive insights, take action: a GA segment can be pushed back into Google Ads, and high-volume data exported to BigQuery. Predictive features flag anomalies and estimate purchase, churn and revenue probability; a marketer behind target can see a two-week revenue forecast and move budget early.

Why it matters

Explains why GA setup is a marketing job, not just an analytics one.

02

Last click versus data-driven attribution

Under last click the striker gets every goal; DDA credits the defenders and midfielders too.

A Myntra shopper sees a Gmail ad, a YouTube ad and a short video, then clicks a search ad; last click hands search 100%. DDA uses device, timing, engagement and path comparison to distribute credit. In Google Ads, type "attribution" in the search bar for conversion paths, model comparison and a "switch to DDA" list; in GA, Advertising > Attribution compares models by channel - in the demo paid search loses credit under DDA while Instagram and cross-network gain it. "If you want to see the true value, you should look at data-driven attribution."

Why it matters

Upper- and mid-funnel channels look worthless under last click.

03

Events, user properties and identity

Your job as a marketer is deciding which events get passed; engineers implement them.

Predefined events (view item, add to cart, purchase) cover the basics; a custom "price filter applied" event can reveal price sensitivity. User ID must be a masked internal ID, never an email or phone number. Up to 25 user properties per project (loyal or high-value customer, frequent flyer, level reached). Collection is via the global site tag (gtag) or, more commonly, Google Tag Manager; IP addresses are not stored and data retention and ad personalisation are configurable.

Why it matters

Bad event design is invisible until the report you need cannot be built.

04

Reading GA4 reports with comparisons and filters

Stickiness (DAU/MAU) only means something against the nature of the product.

Demo on a dummy web-only property: real-time overview, reports snapshot with insight cards, comparisons, 1/7/30-day active users, events by name (form start, form submit), Business objectives > Generate leads, cohort retention, and Traffic acquisition filtered to paid then to Google campaigns, showing a brand versus a generic campaign. Thresholding blanks small demographic cards; revenue is zero because no revenue event is passed. Use the home page if you cannot build a dedicated landing page.

Why it matters

The everyday workflow for answering "which channel brings quality users".

05

Explore: funnel exploration and segment overlap

how-to

Five hours between page view and form start suggests when to send a reminder.

Explore > Funnel exploration > edit steps (page view, form start, form submit) > apply, then add a breakdown (device, first-user medium), a filter (medium = cpc) or "show elapsed time". In the demo, direct visitors start the form one time in four versus one in ten from paid. Segment overlap shows most paid traffic is mobile. Finally, UTM parameters (utm_source, utm_campaign) are how GA attributes non-Google traffic: Google Ads auto-tags, every other platform must pass them - "a mandatory step".

Do it in this order
Why it matters

Finds the exact step where leads are lost.

Tools referenced

ToolCoverageMomentContext
Google Analyticsdemonstrated
Google Adsdemonstrated
Google Tag Managerexplained
BigQueryexplained
Looker Studiomentioned
Search Ads 360mentioned
Display & Video 360mentioned
CleverTapmentioned
MoEngagementioned

Action items

    Resources mentioned

    Resources
    • docLesson page on the platform

    Extraction notes

    This page was built from an auto-generated transcript, which garbles product and people's names. Those were corrected silently in everything above and logged here for transparency. The warnings flag claims that were true on the recording day but change fast.

    Transcript corrections applied

    The transcript saysThe trainer actually means
    DB360DV360 (Display & Video 360)
    more engageMoEngage
    lens cardLenskart
    DAO to MAODAU/MAU
    data ribbon / data givendata-driven

    True on recording day — verify before relying