E-commerce Metrics: The KPIs That Matter and How They Fit Together

Most stores track too many metrics and act on too few. The point of a metric is to tell you where to look next: a falling revenue per visitor sends you to conversion or order value, a rising CAC sends you to channel mix and creative, a rising refund rate sends you to product pages and fulfilment. Read on their own, metrics mislead; read in pairs, they explain each other.

This guide walks through the core KPIs in roughly the order money flows: what you pay for attention, what a visitor is worth, what an order leaves you after costs, and how long a customer stays. Each section links to the deeper guides on that metric.

CPA: cost per acquisition

Cost per acquisition is ad spend divided by the number of conversions you count as an acquisition: an order, a sign-up, a lead. It's the metric ad platforms optimise for, which makes the definition of 'acquisition' important. If the platform counts every purchase, including those from returning customers, CPA will look better than the real cost of a new customer. Always say which conversion a CPA refers to.

CAC: customer acquisition cost

Customer acquisition cost is everything you spend to win new customers in a period, divided by the number of new customers in that period. Unlike CPA it counts only first-time buyers and, in its fully loaded form, includes agency fees, tools, creative and discounts, not just media. Blended CAC (all spend over all new customers) and paid CAC (paid spend over paid-attributed new customers) answer different questions; track both.

CAC only means something next to what a customer is worth. A rising CAC is fine if lifetime value rises faster; a falling CAC is a problem if it comes from buying customers who never return.

LTV: customer lifetime value

Lifetime value is the total gross profit a customer brings over their relationship with the store. The simple version is average order value × purchase frequency × expected lifetime × gross margin; the more reliable version follows real cohorts and adds up what they actually spent in their first 3, 6 and 12 months. Use margin, not revenue, or LTV will justify acquisition costs you can't afford.

Segment it. LTV by first product, by acquisition channel and by first-order discount usually varies more than any average suggests, and those differences tell you which customers to pay more for.

ROAS: return on ad spend

Return on ad spend is attributed revenue divided by ad spend. It's quick to read and easy to misuse. ROAS is a revenue ratio, so the same ROAS can be profitable for a high-margin store and loss-making for a low-margin one. Work out your break-even ROAS (1 ÷ contribution margin percentage) before judging any campaign against it.

Platform-reported ROAS also tends to over-credit the platform, because every channel claims the conversions it touched. Compare it with your own analytics and, for large budgets, with holdout or incrementality tests.

ROI: return on investment

Return on investment is (gain − cost) ÷ cost, using profit rather than revenue. It's the metric for comparing unlike decisions: a new channel, a site redesign, an A/B testing program, a warehouse move. For marketing, ROI differs from ROAS in two ways: it uses margin, and it subtracts the investment. A campaign with a ROAS of 3 and a 30% margin has a negative ROI.

AOV: average order value

Average order value is revenue divided by number of orders. It moves with bundles, free-shipping thresholds, cross-sells, pricing and product mix. Read it next to conversion rate: a free-shipping threshold set too high can lift AOV and lower conversion by more, leaving revenue flat or down. Check the median as well as the mean, because a few very large orders can drag the average around.

Revenue per visitor: the metric that combines both

Revenue per visitor (RPV) is revenue divided by visitors, which is the same as conversion rate × average order value. That makes it the best single number for judging site changes and A/B tests: it can't be gamed by trading conversion for order value or the other way round. Use it as the primary metric for tests that might affect basket size, and break it down by device and channel to find where visits are worth least.

RPV is noisier than conversion rate, because order values vary a lot. Tests on RPV need more traffic to reach a conclusion; plan the sample size before you start.

Cart abandonment rate

Cart abandonment rate is the share of carts created that don't become orders: 1 − (orders ÷ carts created). Most carts in most stores are abandoned, because shoppers use the cart to save items, compare prices and check delivery costs. The rate becomes useful when you split it: by device, by cart value, by the step where people leave, and by whether they saw shipping costs before checkout.

Gross margin

Gross margin is (revenue − cost of goods sold) ÷ revenue. It sets how much every other lever is worth: a store with a 70% gross margin can afford a CAC that would bankrupt a store at 25%. Track it by product and category, not just overall, because product mix shifts it without any price change. Discounts come straight out of gross margin, which is why a promotion that lifts revenue can still lose money.

Contribution margin

Contribution margin goes further: revenue minus all variable costs of an order, including shipping, payment fees, packaging, returns and the marketing spent to acquire it. It answers the question gross margin can't: does this order, channel or customer actually make money? Use it to set break-even ROAS, to price free shipping, and to decide which products to promote.

Retention rate

Retention rate is the share of customers from a period who are still active (buying, or still subscribed) in a later period. For non-subscription stores, define 'active' by your natural purchase cycle: a coffee store might use 60 days, a furniture store two years. Read retention by cohort, grouping customers by the month of their first order, so that growth in new customers doesn't hide a decline in loyalty.

Repeat purchase rate

Repeat purchase rate is the share of customers who have placed more than one order. It's the simplest loyalty metric for non-subscription stores and the biggest driver of LTV. The second order is the hardest; once a customer has bought twice, the chance of a third rises sharply in most stores. Measure time to second order and which first products lead to it.

Churn rate

Churn rate is the share of customers (or subscriptions) lost in a period. It's the mirror of retention and the core metric for subscription stores. Separate voluntary churn (cancellations) from involuntary churn (failed payments), because they have different fixes: one needs a better product or offer, the other a better dunning process.

Refund rate

Refund rate is refunded orders (or refunded revenue) divided by total orders (or revenue). It's often left out of conversion analysis, which is a mistake: a product page change that raises conversion by over-promising can raise refunds by more. Track refund reasons by product, and include refunds when you judge the result of any test that changes how products are described.

Net revenue retention

Net revenue retention (NRR) compares the revenue a group of existing customers brings this period with what the same group brought in an earlier period, including expansion and subtracting churn and downgrades. It's mostly a subscription metric. Above 100%, existing customers grow revenue on their own; below 100%, new customers have to make up the loss first.

Average session duration

Average session duration is total session time divided by sessions. It's a diagnostic metric, not a goal: a longer session can mean engaged browsing or a shopper who can't find what they want. Compare duration between converting and non-converting sessions, and between devices, rather than trying to push the average up.

Bounce rate

Bounce rate is the share of sessions that view one page and leave without interacting. A high bounce rate on a landing page bought with ads is expensive, and it's often a mismatch between the ad's promise and the page. Analyse it by landing page and traffic source; a sitewide bounce rate mixes too many different visits to act on.

Engagement rate

Engagement rate is the share of sessions that count as engaged, typically by lasting longer than a set time, viewing more than one page or triggering a conversion event. It's the positive version of bounce rate and the default in GA4. Because the definition is configurable, check it before comparing stores or tools.

CTR: click-through rate

Click-through rate is clicks divided by impressions, for an ad, an email link, a search result or a button on your own site. A high CTR on an ad only helps if the landing page keeps the promise; otherwise it just buys more bounces. On-site, CTR on category tiles, recommendations and banners shows which elements draw attention and which are ignored.

CPC: cost per click

Cost per click is ad spend divided by clicks. It's set by auction competition and ad quality, and it connects directly to on-site performance: if CPC is €1 and revenue per visitor is €0.80, every click loses money before any other cost. Raising RPV through conversion work is often the cheapest way to make expensive clicks affordable.

CPM: cost per thousand impressions

CPM is the cost of a thousand ad impressions. It's the base cost of attention on social and display channels and moves with season and competition, rising sharply in the run-up to big retail events. CPM × (1 ÷ CTR) × (1 ÷ conversion rate) gives you the media cost of an order, which shows which part of the chain is most expensive.

Checkout conversion rate

Checkout conversion rate is orders divided by checkouts started. Shoppers who reach checkout have shown strong intent, so losses here are usually caused by surprises (shipping costs, required accounts, missing payment methods) or form friction. It's one of the most actionable metrics in the store; compare it by device and payment method, and watch session recordings of abandoned checkouts.

How the metrics fit together

Reading metrics in chains keeps you honest. Acquisition: CPM → CTR → CPC → conversion rate → CPA and CAC. Order economics: AOV × conversion rate = RPV, and RPV × contribution margin = profit per visit. Customer economics: first-order contribution + repeat purchases × margin = LTV, compared with CAC. When one number moves, follow the chain to see whether profit moved with it.

Pick one primary metric per decision. For site changes and A/B tests that's usually revenue per visitor or conversion rate; for acquisition it's contribution after marketing cost; for retention it's repeat purchase rate or cohort revenue. Keep the others as guardrails, so a win on the primary metric can't hide a loss somewhere else.