Going beyond "what happened" in SERP analytics (brightonSEO San Diego 2024)

Transcript

1. Going beyond “what happened” in SERP analytics Ray Grieselhuber DemandSphere Speakerdeck.com/raygrieselhuber @raygrieselhuber
2. Hi, I’m @raygrieselhuber Speakerdeck.com/raygrieselhuber @raygrieselhuber Founder & CEO of DemandSphere 18+ years experience in data systems, engineering, and enterprise SEO
3. Speakerdeck.com/raygrieselhuber @raygrieselhuber The Modern SERP
4. Speakerdeck.com/raygrieselhuber @raygrieselhuber No, Google is not dying (lol) https://sparktoro.com/blog/is-google-losing-search-market-share/
5. Speakerdeck.com/raygrieselhuber @raygrieselhuber We’re going to talk about Google & LLM / Chat Search
6. Speakerdeck.com/raygrieselhuber @raygrieselhuber First, let’s look at Google (this works for Bing too)
7. SERPs are the most undervalued predictor of consumer behavior on the market Visual Rank Sentiment Layout Shift Scroll Depth Pixel Depth CTR Modeling Ad Copy Suggested Keywords # of Elements Ad Location Locations Reviews Merchant IDs Business Titles Ranking URLs Title, Meta, etc. Custom Extraction SERP Screenshots Pixel Height SERP Features Search Intent Keyword Clusters Topic Modeling Search Volume Ad Presence Ad Performance Co-Occurrence Competitor Performance Competitor Discovery Share of Voice Visual Share of Voice URL Screenshots NLP Analysis Video Discussions & Forums Social Knowledge Graph SERP Feature Interiors FAQ Flight Details Hotel Details Review details PLA Text Ads People Also Ask Refine this search Organic Commerce Related Products Shops News News Details Price & Currency Google vs. Bing Buying Guide Howto Job Details Speakerdeck.com/raygrieselhuber @raygrieselhuber
8. Speakerdeck.com/raygrieselhuber @raygrieselhuber What? Why? Going beyond “what happened?”
9. Speakerdeck.com/raygrieselhuber @raygrieselhuber Traditional Google Search Console metrics
10. Old School SERPs (“Ten blue links”) Speakerdeck.com/raygrieselhuber @raygrieselhuber
11. GSC metrics show a basic view of the “search funnel” Speakerdeck.com/raygrieselhuber @raygrieselhuber Avg. Position (Max traffic potential) Impressions (depends on Avg. Pos.) Clicks CTR (clicks / impressions)
12. Half of GSC metrics are derived Speakerdeck.com/raygrieselhuber @raygrieselhuber Avg. Position (Max traffic potential) Impressions (depends on Avg. Pos.) Clicks CTR (clicks / impressions) What happened? Derived metric What happened? Derived metric
13. The missing detail is the visual impact on the user Speakerdeck.com/raygrieselhuber @raygrieselhuber Avg. Position (Max traffic potential) Impressions (depends on Avg. Pos.) Clicks CTR (clicks / impressions) What happened? Derived metric What happened? Derived metric WHY?!
14. Pixels = Attention Speakerdeck.com/raygrieselhuber @raygrieselhuber 330 pixels 540 pixels Pixel depth: 900 pixels Pixel depth and fold appearance determines attention Attention determines click-through rates (CTR) SERP shape is one of the best predictors of user behavior on the internet Visual SERP analytics enables pre-click predictions
15. Impacts: • Fold visibility • CTR • Clicks / Sessions • Conversions • Revenue Let a million mini-SERPs bloom Paid Content • Pixel depth: 120px • Pixel height: 320px • Visual position: 1 • Nested position: 2 Organic Result • Pixel depth: 440px • Pixel height: 330px • Visual position: 3 • Title changed
16. Pre- vs. Post- click analytics Speakerdeck.com/raygrieselhuber @raygrieselhuber Pre-click (SERP data) • Pixel depth • Rank • Visual rank • Market landscape • SERP Features • Competitors (Easy) Post-click (website data) • Cookie settings • Ad blockers & privacy laws • GA4… • Tag issues • ERP integration • CRM integration (Harder)
17. Pixels & CTR Impact Speakerdeck.com/raygrieselhuber @raygrieselhuber
18. Be careful with simplistic, position-weighted CTR models Speakerdeck.com/raygrieselhuber @raygrieselhuber ~20% CTR ?? ~10% CTR ?? ~8% CTR ??
19. Instead: merge and segment the best data available Speakerdeck.com/raygrieselhuber @raygrieselhuber Pixel and fold metrics will help to explain the “why” on segmented CTR models Visual SERP metrics Search Console Bulk Export GA4 Search Volume Long-tail Branded High Volume Keyword Groups Clusters CTR Segments BigQuery
20. A concrete example of visual impact: AI Overviews Speakerdeck.com/raygrieselhuber @raygrieselhuber Google generative overview Key points 394 Average Pixel Height Expanded AIO example Links: Links overview Links (show all) Links are generally pulled from top 12 rankings and have a high cosine similarity to the generated summary
21. Ray Grieselhuber CEO DemandSphere How Google's AI Overviews Are Changing Consumer Behaviors Tim Resnik Global VP of Professional Services Botify
22. AI Overview Research Study Speakerdeck.com/raygrieselhuber @raygrieselhuber 120,778 Total Keywords 36,000 Commercial Keywords 85,638 Informational Keywords 22 Websites 10 E-Commerce 10 Brands 2 Publishers
23. Framing the study Speakerdeck.com/raygrieselhuber @raygrieselhuber Measuring AI Overview impact for your own brand is critical
24. Framing the study Speakerdeck.com/raygrieselhuber @raygrieselhuber Measuring AI Overview impact for your own brand is critical Global AI Overview rollout is not complete
25. Framing the study Speakerdeck.com/raygrieselhuber @raygrieselhuber Measuring AI Overview impact for your own brand is critical Global AI Overview rollout is not complete User behaviour is diverse
26. Leveraging Commercial + PAAs to “force” AIOs Speakerdeck.com/raygrieselhuber @raygrieselhuber running shoes What are the best running shoes? Commercial Informational Can you wear running shoes for walking all day? Informational What shoes to run on a trail? Informational Are running shoes machine washable? Informational
27. Speakerdeck.com/raygrieselhuber @raygrieselhuber Total SERPs 120,778 Commercial 36,000 Informational 85,638 # of Sites 22 With AIOs 57,263 AIO % 47% Keyword Set Breakdown SERPs with AIOs AIO Intent Breakdown AI Overviews appeared for 47% of the keywords
28. Speakerdeck.com/raygrieselhuber @raygrieselhuber AIOs appear far more frequently for Informational Intent 19% of Commercial triggered AIO 59% of Informational triggered AIO
29. Speakerdeck.com/raygrieselhuber @raygrieselhuber Featured Snippet Avg. Pixel Height 237.75 AIO Avg. Pixel Height 394.44 Combined Pixel Height 632.19 Avg. Screen Resolution Height (US) 942.29 394.44 pixels 237.75 pixels https://gs.statcounter.com/screen-resolution-stats/desktop/north-america AIO and FS pixels combined take up 67% of Avg. Screen Height
30. Speakerdeck.com/raygrieselhuber @raygrieselhuber The fundamentals still matter Performing well in organic rankings is the strongest influencing factor for appearance in AI Overview links AI Overview links can be thought of as the new SERP – when they appear 75% of AI Overview links came from position 12 or higher 4 Median Organic Rank 12.03 Average Organic Rank 12 75th Percentile
31. Speakerdeck.com/raygrieselhuber @raygrieselhuber Generally appear in positions 1-12 in organic rankings Presence is currently correlated with medium to high Cosine Similarity score All best practices apply with regard to organic rankings Cosine Similarity allows us to mathematically compare the similarity between two bodies of text. How sources are selected for inclusion in AI Overviews
32. Speakerdeck.com/raygrieselhuber @raygrieselhuber Vector Embeddings 0.001 .0044 .012 …. 0.001 .0043 .011 …. 0.001 .0041 .009 …. Embedding model “marathon training” “running shoes” “route tracker app” Cluster visualization with projections
33. Speakerdeck.com/raygrieselhuber @raygrieselhuber Evaluating AIO links with Cosine Similarity 0.82 0.78 0.73 Cosine Similarity
34. Speakerdeck.com/raygrieselhuber @raygrieselhuber Relationship of Cosine Similarity to AIO Links
35. Speakerdeck.com/raygrieselhuber @raygrieselhuber Let’s look at LLM / Chat-based Search
36. Speakerdeck.com/raygrieselhuber @raygrieselhuber Live Retrieval (RAG) is the key Search Index Foundational Model
37. Speakerdeck.com/raygrieselhuber @raygrieselhuber Back to simple search results (but interactive)
38. Speakerdeck.com/raygrieselhuber @raygrieselhuber Start measuring and working on visibility Proper indexation & visibility in organic results Domain selection & rejection Semantic similarity What are searchers asking?
39. Speakerdeck.com/raygrieselhuber @raygrieselhuber AI Overviews are Google’s RAG in the SERP
40. Speakerdeck.com/raygrieselhuber @raygrieselhuber Prediction: AI Organized results demonstrate agentic behavior, with humans as the agent (for now)
41. What makes a good analytics workflow? Speakerdeck.com/raygrieselhuber @raygrieselhuber
42. Speakerdeck.com/raygrieselhuber @raygrieselhuber What? Why? Answering more questions beyond “what?” Who?
43. Speakerdeck.com/raygrieselhuber @raygrieselhuber Addressing gaps in SEO software https://speakerdeck.com/techseoconnect/michael-king-accounting-for-gaps-in-seo-software Ultimately a question of use cases & workflows
44. Two types of workflows in dynamic systems Exploratory Regression We can take a cue from UX testing
45. Adding the Team vs. Individual axis Exploratory Operational Team Individual
46. Mapping this framework to SEO use cases Exploratory Operational Team Individual Daily SERP Monitoring Dashboards & BI Tools Alerts APIs Data Science / Notebooks Data Exports from Tools Integrations (GSC / GA4) Automations Data Warehouses (BigQuery, etc.) Data Management SERP Archival
47. There are many good reasons to standardize exploratory workflows on Python & BigQuery Speakerdeck.com/raygrieselhuber @raygrieselhuber
48. The winners have repeatable processes Speakerdeck.com/raygrieselhuber @raygrieselhuber
49. Intelligence has a lifecyle that requires action Speakerdeck.com/raygrieselhuber @raygrieselhuber Review and plan Allow for some time after the campaign to ensure that data is vetted. Review successes and failures, establish regression baselines, and plan next phase. Execute campaign Create new content, optimize existing content, work on digital PR, etc. This execution phase should be tied to achieving strategic and operational goals Define monitoring Hone research data set into campaign and group-focused segments, which can be prioritized for action Define strategic goals Intelligence lifecycles are derived from strategic vision and initiatives Research opportunities Initial SERP and search volume focused analysis of competitive and market landscape 01 05 04 03 02
50. Key Takeaways Speakerdeck.com/raygrieselhuber @raygrieselhuber Measuring for yourself is critical Pixel depth, fold, and visual rank are the new metrics for SERP analytics Organic rankings still matter a lot because of live retrieval Analyze and understand similarity metrics Be clear about your workflows Use the opportunity that AI presents to break SERP analytics out of the SEO silo
51. Come visit our booth! Speakerdeck.com/raygrieselhuber @raygrieselhuber x.com/demandsphere linkedin.com/in/raygrieselhuber