How this track works
Eight analyst sessions, one store. You meet Mango Lane and its event trail today (b1), compute the core four metrics by hand (b2), build the GMV driver tree in a live simulator (b3), measure the full funnel including PDP penetration (b4), diagnose a real drop (b5), cut every metric by segment (b6), turn the numbers into a weekly review pack (b7), and finish by writing the metric spec sheet that keeps the whole company honest (b8). The leader track (a1-a6) teaches the same numbers in reading mode - what to ask, not what to compute.
The store: Mango Lane 6 min live
Mango Lane is a mobile-first fashion and lifestyle store: about 3,000 products across womenswear, menswear, accessories, and home. It gets traffic from four places - organic search, paid ads, CRM (email and push), and social. In a typical month, 240,000 people open the app or site, generating 360,000 sessions, and 9,720 of those sessions end in an order averaging $68. Every number in this course is one of these, or derived from them.
LiveWhy one made-up store beats ten real screenshots3 min▶
Real dashboards change weekly and hide their definitions. Mango Lane holds still: the same 360,000 sessions and 9,720 orders appear in every session of this course, so when b4 tells you PDP penetration is 40%, you can check it against numbers you already know. By b8 you will have decomposed one month of one store completely - which is exactly how real analysts learn a new company's numbers.
- One month, one store, every metric - the numbers cross-check across all 8 sessions.
- Fashion, mobile-first - typical mid-size ecommerce: enough traffic for stable rates, small enough to reason about.
- Four channels - organic, paid, CRM, social - because b6 (segment cuts) needs something to cut by.
The event trail: five things a shopper can do 9 min live
A store's analytics is just a diary of shopper actions. The industry-standard vocabulary (this is what Google Analytics 4 calls them, and every tool has an equivalent) is five events: session_start, view_item (a product detail page - a PDP - gets viewed), add_to_cart, begin_checkout, and purchase. Every ecommerce metric in existence is a count of one of these events, or a ratio between two of them.
LiveEach step loses people - that is normal3 min▶
Of Mango Lane's 360,000 monthly sessions, only some view a product, fewer add to cart, fewer still start checkout, and 9,720 pay. That narrowing is the funnel, and it is not a failure - no store converts everyone. What matters is how much narrows at each step compared to last month and compared to peers. Session b4 measures each step precisely; today you just need the shape.
The vocabulary is genuinely universal. GA4 uses these exact event names; Amplitude, Mixpanel, Adobe, and Shopify each use near-identical ones (Shopify's funnel report literally shows "added to cart → reached checkout → converted"). Learn the five steps once and every analytics tool you ever open becomes readable.
Self-studyWhat each event carries with it3 min read▶
Events are not bare ticks - each carries properties. A view_item knows which product; an add_to_cart knows the quantity and price; a purchase carries the order id, every line item, and the total. Two consequences worth remembering:
- Metrics can be sliced by any property - CR by channel, AOV by category - which is all "segmentation" (b6) means.
- Money lives only on purchase. Everything before it measures attention and intent; only the purchase event measures revenue. When a report mixes the two, ask which event each number came from.
Visitor, session, pageview: the counting rules ★ 15 min · everyone works
The most common wrong number in ecommerce is not a wrong formula - it is a right formula fed the wrong count. One person (a visitor) can visit three times (three sessions) and look at ten pages (ten pageviews). Divide orders by the wrong one and your "conversion rate" is off by half. Work these three rules with real Mango Lane numbers.
LiveRule 1 - a session is a visit, and it expires5 min▶
A session is one continuous visit. In GA4 a session ends after 30 minutes of inactivity - walk away for lunch mid-browse and your return is a new session. Mango Lane's 240,000 visitors produce 360,000 sessions: 1.5 sessions per visitor. Fashion shoppers browse, leave, come back - that ratio is the store's window-shopping habit made visible.
Maya opens Mango Lane on Monday lunch, browses, leaves. Tuesday she returns and buys. Count: 1 visitor, 2 sessions, 1 order.
Her Monday session converted at 0%. Her Tuesday session converted at 100%. As a visitor, she converted at 100%.
Now scale that up: session CR and visitor CR are different numbers for the same store, and both are correct. You must say which one you mean.
LiveRule 2 - conversion rate needs a declared denominator5 min▶
Mango Lane, same month, two honest conversion rates:
| Definition | Formula | Mango Lane |
|---|---|---|
| Session CR (Shopify's "online store conversion rate") | purchase sessions ÷ sessions | 9,720 ÷ 360,000 = 2.7% |
| Visitor CR (how GA4's purchase-journey funnel counts - by user) | buying visitors ÷ visitors | ≈ 9,300 ÷ 240,000 = ≈ 3.9% |
Same store, same month, 2.7% vs 3.9% - a 44% difference with zero change in the business. This is why b8 makes you write a spec sheet: a metric without a declared denominator is a rumor.
LiveRule 3 - pageviews inflate, events discriminate3 min▶
Pageviews count every screen load - home, listing, PDP, cart, all mashed together. They are fine for "is the site up and busy" and nearly useless for diagnosis. The event trail is sharper: view_item counts only PDP views, which is why b4's PDP penetration (sessions that viewed at least one product ÷ all sessions - 40% at Mango Lane) tells you something pageviews never can: whether shoppers are actually reaching products or bouncing around navigation.
The classic trap: a redesign ships, pageviews jump 20%, the team celebrates. Then someone checks: shoppers now need three screens to reach a product instead of two. Pageviews rose because the journey got worse. PDP penetration - flat. Orders - flat. Count events, not screens.
Try it yourself - this week ◐ 15-20 min total
- Open any analytics tool you have access to (GA4, Shopify, anything). Find where it reports visitors vs sessions. Compute the sessions-per-visitor ratio - is your audience more or less of a window-shopper than Mango Lane's 1.5?
- Find the tool's conversion rate. Hunt down its denominator - sessions or visitors? (It is usually in a tooltip or the docs, rarely on the dashboard itself.)
- Say Mango Lane's baseline out loud from memory: 240K visitors, 360K sessions, 9,720 orders, $68 AOV, $660,960 GMV. You will use these numbers in every remaining session.
- Write the five event names on a sticky note. Next session (b2) you compute the core four metrics from them by hand.
Sources this session draws on
Ecommerce metrics have no single certificate - the definitions live in the analytics platforms' own documentation. This course teaches the working core of these, applied to one store end to end. This page draws on:
Three questions before you go 🎯 ◐ 90 seconds
1 · Maya visits Mango Lane on Monday, leaves, returns Tuesday and buys. What did the store count?
One person = one visitor, however often she returns. Each separate visit is its own session (they expire after ~30 min of inactivity). This is why session CR and visitor CR are different numbers.
2 · Mango Lane's session CR is 2.7%, but a user-counting funnel (like GA4's purchase journey) shows roughly 3.9%. Why?
Both are correct. 9,720 orders ÷ 360,000 sessions = 2.7%; ~9,300 buying visitors ÷ 240,000 visitors ≈ 3.9%. A conversion rate without a declared denominator is a rumor.
3 · After a redesign, pageviews jump 20% but orders are flat. What should you check first?
Pageviews count screen loads, not progress. If the journey got longer, pageviews rise while the funnel is unchanged or worse. PDP penetration (sessions reaching a product page) tells you whether attention is reaching products.