Scaling E-commerce with LTV, CAC, Attribution and AI

Recorded in Polish. English subtitles are available on the recording via YouTube's auto-translation.

Meet Marek, the manager of a mid-sized e-commerce business. Marek is a composite character based on real people, but every figure shown in this session comes from his actual store.

Marek has Q1 targets: more revenue, more profit. The trouble is that click costs are rising, competitors keep multiplying, and he cannot raise budgets. Performance Max, search, brand, social, and a newsletter to everyone who signed up worked for years — now they are running out of road.

Marek noticed that market leaders hold two advantages: in-house analytics teams, and AI implementations fed by that data. Hiring an analyst takes months and costs more than he can justify. This webinar is about winning that advantage without an analytics department.

Part 1: The customer analysis framework

Who counts as a customer. A customer is someone who has transacted with you and from whom you expect further purchases. And, crucially, a customer is not a user. Analysing how someone behaves on your site and where they came from is user analysis in GA4. Customer analysis happens on hard CRM data — unaffected by consent, disabled cookies, or a payment method that was never tracked.

Step 1 — collect the data. From your store database (PrestaShop, IdoSell, WooCommerce, Magento, custom): email address, every order with its date, the date and value of the first and last order, and the total count and value of all orders.

Step 2 — segments: RFM 2.0. Classic RFM (recency, frequency, monetary) is methodologically sound, but splitting into deciles and dozens of segments is so complex that it tells no story — over the years I have seen very few companies actually act on such an analysis. So we keep the same dimensions and rearrange them into segments that narrate the customer lifecycle:

• New — first order
• Returning — second order
• Loyal — third and subsequent orders
• Gold — loyal customers past a value threshold (e.g. 3–5× average order value)
• At risk — approaching the inactivity boundary
• Lost — no order in a year
• Winback — recovered from the lost segment

Step 3 — three analyses. The snapshot shows today's state: customer count, revenue, share, average LTV and AOV per segment. In Marek's store, Loyal Gold is 21,500 customers yet drives 40% of revenue — against 95,000 new customers alone. It also exposes 76,000 lost customers and a handful of gold customers at risk. Change over time (30/60 days) shows the dynamics — for instance a 60% jump in the lost segment exactly a year after a big promotional push. Flow shows where customers actually moved: how many new ones came back for a second order, and how many went straight to the graveyard.

Step 4 — an action plan per segment. For new customers the goal is the second order: a dedicated email with a discount, matched to what they bought. With 95,000 new customers and 5% conversion, that is nearly 5,000 additional orders — roughly half a million in revenue, while also saving around 70,000 PLN you would otherwise pay in media CPA. For lost customers: a custom audience built from their email addresses in Meta Ads.

Step 5 — monitoring. A day-by-day report showing which segment each ordering customer came from. You react daily, not after a month.

AI instead of manual work. This whole process normally takes hours, if not days. AI fed with the data produces a monthly summary of key findings, a base-health analysis, a flow analysis, a financial impact assessment, and concrete day-by-day recommendations.

Attribution per segment. Joining CRM data with GA4 in the warehouse reveals which segment each source actually delivers. Suddenly you see that Google Ads brings mostly new customers, while loyal and gold customers return through direct — people you paid for once and no longer pay for. That inverts how you judge campaigns. Includes a comparison of attribution models, such as last click versus linear.

Part 2: Customer acquisition analysis

Buying transactions versus buying customers. 99% of e-commerce buys transactions: a conversion costs 15–25 PLN, returns 150 PLN, and you calculate ROAS. Knowing LTV tells you that customer will bring not 150 but 450 PLN — which means you can outbid competitors on customers you know will come back, without raising budgets.

The metrics. New customer count, average CAC, average LTV, and the value of first versus subsequent orders. The LTV-to-CAC ratio: 3:1 is healthy and stable; 1:1 is alarming; above 5:1 means you are investing too cautiously and should push harder. Plus ERS (effective revenue share, the inverse of ROAS) in short and long variants — is the customer growing in value over time.

The cohort report. How many customers were acquired in a given month, how much they spent in that month, and how much in the months that followed. Essential for retention-driven businesses — meal-prep subscriptions, apparel, anything seasonal. GA4 cannot genuinely deliver this, because it does not run on CRM data.

The finishing touch: are welcome coupons worth it? Did the customer who used a first-order discount actually come back? In Marek's store it turned out almost nobody redeems the coupons, and those who do are price hunters who never become real customers — meaning the margin sacrifice may simply not pay off for that group. In another business the opposite held: coupon customers returned and placed substantially larger orders.

Who it is for

E-commerce managers who need to hit targets without raising budgets, and agencies who want to do better work on their clients' data. This session builds on the previous episode about constructing a data warehouse — worth watching first.