Marketing metrics that matter: funnels, conversion, CAC, LTV and attribution

Most marketing reports show plenty of numbers and few answers. Visits went up, followers went up, cost per click went down, and nobody can say whether marketing brought in more customers. The fix is a small set of metrics that follow a customer from first visit to purchase, plus an honest view of how credit for a sale is shared between channels.

This guide covers the marketing funnel, conversion rate, customer acquisition cost (CAC) and lifetime value (LTV) as marketing uses them, and the attribution models in Google Analytics 4 (GA4). Our guide to unit economics covers the finance side of CAC, LTV and payback in more depth.

Build a funnel that matches how people buy

A funnel is the series of steps people take towards a purchase or enquiry. Each step loses some people. Measuring the step-by-step loss shows you where to work.

Keep the funnel close to real behaviour on your site. For an online shop that might be: visit, view a product, add to basket, start checkout, purchase. For a service business: visit, view a service page, start the enquiry form, submit it, book a call.

GA4’s funnel exploration lets you build this from events. You can define up to 10 steps. In a closed funnel, users must enter at the first step. In an open funnel, they can enter at any step. GA4 also notes that “if the user misses a step, they fall out of the funnel and aren’t counted in any subsequent steps”, so step order matters when you set it up.

Conversion rate, step by step

Conversion rate is the share of people who complete an action out of those who could have. Always say what the top and bottom of the fraction are, because “conversion rate” alone can mean many things.

A funnel example (illustration only)

An online shop records a month of figures. They are invented to show the method.

The overall conversion rate is 360 ÷ 20,000 = 1.8%. On its own that says little. The step rates say more. Only 15% of product viewers add to basket, which points at product pages: price, photos, delivery information or stock. Half of those who start checkout do not finish, which points at checkout: surprise delivery costs, forced account creation or payment problems. You now have two specific places to test, instead of a vague aim to “improve conversion”.

Mark the right events as key events

GA4 calls the actions that matter most to your business key events. Any event can be marked as a key event. Google Ads conversions are then created from those events if you want to use them for ad measurement and bidding.

Choose few. A purchase, a submitted enquiry and a booked call are key events for most small businesses. A newsletter sign-up might be. Scrolling halfway down a page usually is not. If everything is a key event, conversion rates rise and mean nothing.

CAC and LTV from a marketing point of view

CAC is the cost of winning a new customer. LTV is the gross profit you expect from that customer over the relationship. Marketing needs both for a simple reason: a channel that brings cheap customers who never return can be worse than an expensive one that brings loyal ones.

Marketing adds two points to the finance view.

• Blended versus channel CAC. Blended CAC divides all marketing and sales cost by all new customers. Channel CAC divides one channel’s cost by the customers attributed to it. Channel CAC depends entirely on your attribution model, which is why the next section matters.
• New versus returning customers. Ad platforms and analytics often count every purchase, including repeat orders from existing customers. If you divide ad spend by all purchases, CAC looks lower than it is.

Attribution: who gets the credit

Customers rarely buy after one touch. Someone might see an Instagram ad, search for the brand a week later, then buy after clicking an email. An attribution model decides how credit for that sale is shared between the three.

GA4’s attribution models page lists three options for reports:

• Data-driven. Credit is shared using a machine learning model that compares paths that converted with paths that did not. Each model is specific to the advertiser and the key event.
• Paid and organic last click. All credit goes to the last channel clicked before conversion, ignoring direct visits.
• Google paid channels last click. All credit goes to the last Google Ads click. If there was none, GA4 falls back to paid and organic last click.

You may still read about first click, linear, time decay and position-based models. GA4 states they “are no longer available as of November 2023”. The ideas remain useful for thinking, but you cannot select them in GA4.

GA4’s attribution settings also set a lookback window: how far back a touch can earn credit. The default is 30 days for acquisition key events such as a first visit, and 90 days for other key events, with 30 or 60 days as options. Changing the reporting model applies to historical data as well as future data. Changing the lookback window applies only going forward.

An attribution example (illustration only)

Take the path above: Instagram ad click, organic search visit, email click, purchase worth £90. Under paid and organic last click, email gets all £90. Instagram gets nothing, even though it introduced the customer. Under data-driven attribution, GA4 might give some credit to each, depending on what it learns from your data. If you cut Instagram spend because last click shows no sales, you may find brand searches and email sales fall a few weeks later.

A short metrics set for a small business

1. Visitors by channel, to see where traffic comes from.
2. Step conversion rates through your funnel, to find the leak.
3. Key events by channel, using one attribution model consistently.
4. Blended CAC, counting only new customers.
5. Repeat purchase rate or a simple LTV estimate, to judge customer quality.

Review them monthly. Note the attribution model on the report so nobody compares numbers built on different rules.

Common mistakes and limits

• Quoting a conversion rate without saying what was divided by what.
• Marking minor actions as key events, which inflates every rate.
• Treating ad platform conversion numbers and GA4 numbers as the same. They use different rules and rarely match.
• Cutting a channel because last click shows few sales, without checking what it starts.
• Switching attribution models mid-year and comparing the results with last year.

Attribution models describe patterns in tracked data. They miss offline word of mouth, people who decline cookies and journeys across devices that cannot be linked. For bigger decisions, run a simple test: pause or reduce a channel in one period or region and watch what happens to total sales, not just attributed sales.