Hello, hello! 👋🏻
The guy who literally wrote the book on web analytics just published a manifesto declaring that human-powered analytics is dead.
Avinash Kaushik (the guy who built Google Analytics' measurement practice and literally wrote the book on web analytics) just published a 3-part manifesto declaring that human-powered analytics is dead (I highly recommend his newsletter).
If you've an old timer in digital marketing, you probably know his famous 10/90 rule: invest $10 in tools, $90 in human analysts.
His NEW 10/90 rule:
"If you have $100 to invest in smart decisions, invest $10 in brilliant human analytical strategists, invest $90 in AI activation."
The Core Problem: Insights Latency
The issue isn't that we don't have enough data. We are all drowning in it.
The problem is the delay between data collection and actionable insights.
Here's what traditional analytics looks like in most organizations I work with:
Days generating reports and dashboards
Manual segmentation of "known knowns"
Presentations to executives for months
Complex insights competing with other team priorities, getting missed and misinterpreted
This workflow is fundamentally reactive. Linear. And it's struggling to keep up with the speed of AI marketing.
AI changes these human limitations:
Cognitive constraints. We can only track a finite number of correlations simultaneously. AI processes hundreds/thousands of variables per user engagement.
Insights latency. AI identifies emerging patterns in real-time, predicting what's coming instead of just reporting what happened.
Pattern recognition limits. Us humans can't identify subtle, non-linear, multidimensional relationships. AI finds the "unknown unknowns" hidden in our messy data.
Where to Start
Kaushik outlines 9 specific applications (please read them). I'll focus on the 4 to start with.
1. Propensity Modeling > Impact: Transformational
This is the shift from "who converted" to "who WILL convert."
Instead of spreading your budget across all users, you focus on high-propensity humans. ML algorithms identify subtle combinations of behavior that signal conversion readiness across hundreds of variables.
The results Kaushik reports:
35-60% improvement in conversion rates for targeted segments
20-35% reduction in acquisition costs
Shift from reactive to proactive marketing
This isn't theoretical. Companies are already using tools like XGBoost and LightGBM to predict who will convert, when they'll convert, and what will push them over the edge.
📺 Watch: Propensity Modeling to Identify High-Value Customers (Retail Summit, 15 min)
📺 Watch: Value Based Bidding with Propensity Modeling (Technology for Marketing, 2024)
2. Behavior & Intelligence (BTI) > Impact: Transformational
This goes beyond A/B testing into what Kaushik calls "continuous intelligence."
Traditional A/B tests pick a winner. BTI creates a flywheel: the AI learns from one user and improves predictions for ALL subsequent users. It assembles personalized experiences in real-time from a library of headlines, images, layouts, and offers.
One media company Kaushik worked with:
- Assembled experiences from 50+ headlines, 15+ formats, 10+ layouts in under 100 milliseconds
- Users showing "scanning behavior" got key fact summaries, resulting in +45% article completion
- Users showing "lingering behavior" got deep dive content, resulting in +70% engagement time
The kicker? One company ran the equivalent of 5,000+ A/B tests per week, autonomously.
Results:
45% increase in NPS from intent-responsive experiences
60% reduction in bounce rates for high-intent segments
25% uplift in LTV for users who converted via optimized BTI
3. NLP for Unstructured Data > Impact: High
I worked for an early Natural Language Processing company so we are finally connecting the WHY (surveys, chat transcripts, support tickets) with the WHAT (behavioral data).
Example from an online retailer:
"Size uncertainty" mentions in chat = 47% higher cart abandonment
Positive "easy returns" sentiment = 30% higher LTV
The system predicted negative reviews BEFORE they happened
Results:
15-20% improvement in conversion rate from language insights
+15 point improvement in Employee NPS (from automating qual analysis)
Tools like BERTopic and transformer models can process thousands of documents in minutes, extracting themes, sentiment, and intent that would take human analysts weeks.
4. Real-Time Pricing Optimization > Impact: High
Most pricing is rules-based: seasonal adjustments, competitor matching, promotional calendars. This leaves money on the table.
AI-powered pricing processes dozens of variables simultaneously: user behavior, competitor prices, inventory levels, demand forecasts, and individual price elasticity.
Example from a large company (45 pricing dimensions):
14% revenue lift in first six months (with no increase in acquisition cost)
22% savings in inventory costs
Significant market share shift from responsive pricing
Industry benchmarks:
Fashion/ecom: Revenue +30%, Gross Margin +6%
Airlines: Revenue +15% per flight
Retailers: Pricing accuracy jumped from ~70% to ~95%
What This Means for You
The good news: You don't need to be a data scientist. You don't need to code neural networks from scratch. The tools are increasingly accessible in your AI model on your desktop.
The reality check: Your competitors are already doing this. The gap will widen exponentially, not linearly. Every quarter you get 2-3 quarters behind.
Kaushik puts it bluntly:
"We've turned analytics into a reporting factory that creates zero competitive advantages. We've confused activity with impact. We've mistaken data collection for business transformation. AI isn't coming for your job. The status quo already did."
Where to Start
Tier 1: Start Here (Existing data, high impact)
Propensity modeling for conversion prediction
Advanced segmentation (move beyond demographics)
Anomaly detection with automated root-cause analysis
Tier 2: Build Toward (May need data integration)
Voice of Customer + behavioral data integration
NLP for support tickets and surveys
Dynamic pricing/offers
Tier 3: Advanced (Significant investment)
Full BTI implementation
Liquid merchandising
Whole-company LTV modeling
NEW! Digital Marketing Boot Camp 2.0
The AI Execution Engine
Friday, February 13 | UNH Manchester
Your Marketing Plan Deserves Execution—Let AI Do the Heavy Lifting
In this one-day boot camp, participants will build real AI-powered marketing systems, including content engines, persona-driven email sequences, GEO (Generative Engine Optimization) tactics, and automation workflows, during the session. Instead of leaving with ideas, you’ll leave with working templates, prompts, and automations tailored to your audience and ready to deploy.
Led by Alec Newcomb and Marlana Trombley, this program helps marketing teams, agencies, and nonprofits execute their 2026 marketing plans faster and more efficiently.
Questions to Ask Your Team This Week
Are we doing predictive propensity modeling, or just reporting on conversions?
How many variables does our segmentation actually consider?
Are we connecting behavioral data with voice-of-customer data?
What's our average time from anomaly to diagnosis to action?
If the answers are uncomfortable, that's the point.
Gif by theoffice on Giphy
The Uncomfortable Question
Kaushik asks: "Does the human Analyst role exist by Dec 2027? If it does, what's the job?"
My take: The future analyst is a strategist + AI orchestrator. They define the right questions, validate AI outputs against business outcomes, and translate insights into action.
What they're NOT doing: building reports manually, hunting for patterns in spreadsheets, or setting threshold colored alerts.
Go Deeper
📺 Avinash Kaushik on AI & Digital Analytics (Super Data Science Podcast): Watch on YouTube
📺 Answer Engine Optimization & AI-Resistant Content (Avinash Kaushik): Watch on YouTube
📖 Avinash's Blog (decades of analytics wisdom): kaushik.net/avinash
📖 XGBoost Documentation (propensity modeling): xgboost.readthedocs.io
Your Turn
Pick ONE high-impact use case from this list. Just one. Start a conversation with your analytics team or agency this week.
Ask: What would it take to implement this in Q1?
Hit reply and tell me: What's your biggest barrier to AI-powered analytics? Data, tools, boss barrier, or mindset?
Alec
P.S. if you're thinking about adding a second AI into your life. Anthropic just dropped Opus 4.6. More on multi-ai next week…
