AI for Analytics
Over half your web traffic is no longer human. Learn to measure all three storefronts — the website, the chatbot, and the AI agents — and turn the data into revenue you can defend.
A hands-on course in analytics for the two-visitor web, where your visitors are humans and machines. You run the analytics of LaceYard — a fictional online shoe retailer with a live simulated store, a 13-file data corpus, a chatbot, and an agent-commerce endpoint — using your AI assistant as the analyst's workhorse while you own every number. You learn to read websites (funnels, customer value, honest regression), chatbots (containment, chat-assisted revenue), and AI agents (tasks, mandates, protocol orders), and you finish by writing the strategy memo that finds +30% revenue, every number defended from the files.
Who it's for — Marketing, product, operations, and business professionals who work with websites, apps, chatbots, or AI-facing storefronts and want to make data-backed decisions — no coding background required.
In class you work on a synthetic company corpus — a fictional business modeled on real-world challenges, so you practice on realistic, messy data, not toy examples.
- Measure the two-visitor web: know who's really at your door — humans, chatbot users, verified bots, declared agents, and stealth scrapers
- Apply the Reach–Engage–Convert framework as controllable inputs — and never present a rate without its base
- Direct an AI assistant to sessionize raw clickstream data and fit an honest conversion model — catching the leakage, confounders, and self-selection traps a flattering model hides
- Read GA4 like an analyst — and know exactly which questions (margin, identity, machines) need the warehouse instead
- Compute RFM tiers, CLV, LTV:CAC, and cohort retention — and join behavioral tribes to wallets to find the highest-ROI fix on the site
- Grade a chatbot by its job: containment with a satisfaction signal next to it, false-containment-adjusted rates, and chat-assisted revenue without the self-selection lie
- Review AI agents like employees — a shared agent handbook, a per-agent ledger, and Know-Your-Agent (identity, authority, compliance) — and audit agent commerce by reconciling ledgers, not funnels
- Read a raw server log as a behavioral record: take the census of humans and machines, find the crawler sweeping your catalog, and decide serve, price, or block
You leave with Your AI assistant as the analytics workhorse — loading the corpus, modeling, segmenting, and generating verified reports — plus four agents you hire, brief, review from their ledgers, and retire..
Every Enorasi student leaves with their own AI agent — built to their own custom spec — and uses it in their daily work. Whatever your role, you decide what it knows, what it does, and how it works — then put it to work.
How you build them, class by class
Foundations, Data & Honest Analysis
- The course thesis: majority-machine traffic — and the three analytics (website, chatbot, agent) that share one discipline
- Where the data comes from: tags, cookies, pixels — plus the server logs where bots and agents are honest
- Reach–Engage–Convert as controllable inputs, computed on LaceYard's real numbers
- From 61,812 raw events to an honest conversion model: leakage, confounders, A/B tests, and the duration paradox
GA4, Your Best Customers & the Chat Storefront
- GA4 by example — reports, key events, funnels, UTMs — and what it quietly hides
- People, not sessions: identity stitching, RFM tiers, CLV vs CAC, cohorts, and the replenishment heartbeat
- Find the tribes with clustering, join them to wallets, and design a segment A/B with a guardrail
- Chatbot analytics: every web metric's chat twin, containment traps, false containment, and when NOT to build a bot
When Customers and Employees Are Agents
- How a person briefs an agent and turns a task into a routine — the three hires, the shared handbook, the ledger
- The server log as a behavioral record: the census, the hourly profile, and the crawler that swept the catalog
- A purchase with zero pageviews under a mandate — and how agents find each other (card + registry) and talk (MCP)
- The two-visitor scorecard: one measurement language across website, chatbot, agent-employees and agent-customers
The names the whole industry runs on.
Your AI assistant
Claude, ChatGPT, or Gemini as the analyst's workhorse — it loads the files, computes, and drafts the reports; you verify and own every number.
Google Analytics 4
The industry-standard web analytics cockpit — you learn what it answers well and exactly where the data warehouse must take over.
MCP & agent protocols
The fast-emerging rails of agent commerce — mandates, protocol ledgers, llms.txt, the agent card and registries — instrumented before they're big numbers.
Plus the rest of the working stack — Google Analytics 4 (GA4) · Warehouse-style event exports · Claude / ChatGPT / Gemini as the analysis workhorse · AI-generated, human-verified reports · Web-server logs · Agent protocol ledgers · MCP · llms.txt, product feeds & the agent card at /.well-known/agent-card.json · Agent registries, mandates & ledgers.
Your hands-on capstone: the +30% memo, plus a deck of no more than 10 slides. As LaceYard's head of analytics you find the revenue levers hiding in the corpus — across the website, the chatbot, and the agent surfaces — size each one in dollars with a method a classmate could re-run, and attach a validation plan to every recommendation. Graded on five lines: correctness · lever quality · validation plans · two-visitor coverage (at least one lever from human data and one from machine data) · governance (one handbook amendment). This is the work your certificate is based on — and a portfolio piece that proves you can measure a business whose visitors are no longer only human.
Your certificate is based on what you build — not attendance.
Ready to build your AI agent?
A small, high-quality cohort for professionals — fully online, hands-on from minute one.