Personalising GA4 Sessions with Raw BigQuery Data

How do you get better data for your business?

Ditch the GA4 interface and instead use that excellent raw BigQuery data to build the metrics and dimensions that actually work for your business.

Transcript

1. Personalising GA4 sessions using raw BigQuery data David Westby
4. Problem 1: How to customise sessions and make them bespoke for your business.
5. Problem 2: How sessions are constructed… (This changed from UA to GA4!)
10. What is the first problem What is the second problem What is the solution Customising sessions
11. If you're looking solely at the standard default GA4 session dimensions & metrics you're doing it wrong.
16. Your business is not every business.
17. What is the first problem What is the second problem What is the solution GA4 Metrics are generic
18. • Engagement Rate • Average engagement time per session • Events per session
19. Events per session
21. Engagement Rate
23. If these are crap, what do we want to do then?
24. We want to make custom metrics that are directly aligned to the actions you’d want users to take.
25. Raw GA4 BigQuery data ✨ ✨
26. Start here: What is our objective?
28. Example: Payroll Software Provider
29. Start here: What actions do we want users to take?
30. Customised GA4 Session Metrics • Pricing & Solutions pages visits. • Demo Video plays. • Forms started, but not completed. • Webinar Page visits.
32. The key is creating bespoke metrics that actually matter to your business
33. Raw GA4 BigQuery data ✨ ✨
34. What is the first problem What is the second problem What is the solution Problematic GA4 Dimensions
38. GA4 makes some funky decisions but we don't have to be stuck with them.
39. Example : Default channel groupings are odd.
43. Useful?
44. No
45. The answer?
46. Raw GA4 BigQuery data ✨ ✨
47. What is the first problem What is the second problem What is the solution
48. Problem 2: How sessions are constructed… (This changed from UA to GA4!)
49. What is the first problem What is the second problem What is the solution Why sessions?
50. Sessions allow us to group a collection of actions together.
53. Example : User lands organically from Google
58. Summary Point : Sessions mean we group by metics / dimensions to summarise actions
59. What is the first problem What is the second problem What is the solution Session attribution changed from UA to GA4
60. Big Change : If a user visits from multiple traffic sources within the same “session”, the later visits (and traffic sources) don’t exist.
62. I really want to buy some refillable deodorant
65. refillable deodorant
66. refillable deodorant
67. Fussy deodorant
69. I love meeting Cliff for coffee
71. You stink son, get yourself some fussy deodrant
72. I need a wee
74. Fussy deodorant
75. Air Fryers are the best thing ever
76. I’m bored of you Cliff
78. Fussy deodorant
79. Fussy deodorant
80. Fussy deodorant
81. Fussy deodorant
83. What is the first problem What is the second problem What is the solution How session attribution changed
85. GA4 Sessions are pretty simple: A new session starts after 30 minutes of inactivity.
86. UA Sessions have a few more conditions: A new session starts after 30 minutes of inactivity…but also.
89. Example:
90. UA: 3 sessions
91. Universal Analytics:
92. GA4: One session
93. GA4: One session
94. The numbers look completely different
97. Let’s take a real world example
98. eCommerce Client: Comparison of how revenue compares between UA and GA4, broken down by default channel grouping.
100. GA4 Session Attribution UA Session Attribution
101. Revenue ($)
102. Summary Point : GA4 & UA sessions give traffic sources different numbers of sessions, conversions & revenue.
103. In this example, we can see using UA session attribution meant that:
104. Paid Search was responsible for $11.7K more in revenue
105. Direct, Organic Search & Email were responsible for $16.5K less in revenue
106. What is the first problem What is the second problem What is the solution Let’s talk opportunities
107. Raw GA4 BigQuery data ✨ ✨
108. With GA4 BigQuery data we can build sessions ourselves….
109. … with all the custom metrics, dimensions, and attributions we want.
110. Developer? A tool? Our new AI overlords? You?
111. I’m not going to be writing SQL now…
115. What is the first problem What is the second problem What is the solution Getting our data ready
119. What is the first problem What is the second problem What is the solution Building session dimensions
123. first_value( ) over (partition by session_key)
124. What is the first problem What is the second problem What is the solution Building session metrics
125. There are different types of aggregation we can use
128. sum(revenue) sum(pageviews) max(pricing_page_visit)
129. What is the first problem What is the second problem What is the solution Choosing Session Attribution
130. The default would be using session keys to create sessions
131. We could also recreate UA sessions if we fancy…
134. What is the first problem What is the second problem What is the solution Summary
135. Summary: Session Dimensions These are created by duplicating specific dimensions across the whole session
136. Summary: Session Metric These require some sort of aggregation across the session
137. Summary: Session Attribution You can do your own session attribution - UA, GA4 or other!
138. What is the first problem What is the second problem What is the solution How to do it?
139. There’s a maturity scale here…
140. 1. Copy & Paste SQL from PipedOut/ ChatGPT 2. Write more customised SQL. What dimensions need to exist at each level? 3. Start to build a pipeline If you’re big this is what you should do.
141. 1. Copy & Paste SQL from PipedOut/ ChatGPT 2. Write more customised SQL. 3. Start to build a pipeline If you’re big this is what you should do.
142. 1. Copy & Paste SQL from PipedOut/ ChatGPT 2. Write more customised SQL. 3. Start to build a pipeline
143. On the Piped Out website , we have tonnes of queries using the GA4 BigQuery data …
146. Or use us at Piped Out It’s what we do.
148. LinkedIn: dawestby Thanks Folks! pipedout.com/brighton-seo-oct-25