Custom conversion tracking

A conversion is any action you would spend money to cause. That definition does more work than it looks, because it rules out most of what gets tracked as a conversion: newsletter signups nobody reads, PDF downloads from people who were never going to buy, and time-on-page thresholds invented to make a chart go up.

This page is about setting up conversion tracking that answers a real question, and about the two places it reliably goes wrong.

Defining conversions worth tracking

Start with one. Whatever the site is for, there is a single action that matters most, and every other conversion is a proxy for it or a step towards it. Define that one first and get it working before adding anything.

Then add the steps leading to it, because a single conversion number tells you whether things are working and nothing about why. A typical set for a software marketing site: reached pricing, started signup, completed signup, still active after seven days. For a shop: viewed product, added to basket, started checkout, completed order.

Give each a plain name, agree it with whoever reads the reports, and resist adding a sixth. Five well-understood conversions used weekly beat thirty nobody can define.

Building a funnel that survives scrutiny

A funnel is a fixed sequence of steps with drop-off measured between them, and its value is entirely in the drop-off. The absolute numbers tell you little; the step where half your people leave tells you where to spend next week.

Two design rules keep funnels honest. Steps must be in a genuine order, not just a list of things you care about, or the percentages mean nothing. And the funnel should span a period long enough for people to complete it: a seven-day funnel on a product with a thirty-day evaluation cycle will show you a catastrophe that is not happening.

One thing worth checking before you buy: funnels are usually a paid-tier feature. Plausible puts them on its £19 Business tier, Absolutely Analytics on its $19 Business tier, Pirsch on Plus at $12, Matomo includes them at every level including free self-hosted, and Fathom does not offer them at all.

Attribution, and why to keep it simple

Attribution is the argument about which touchpoint gets credit when somebody saw a social post, searched your name, clicked an ad and then converted. Whole careers are built on this and most of it is not worth your time.

For most sites, last non-direct click is a perfectly good default, and the honest position is that you cannot untangle the rest without cross-day identity, which a cookieless tool deliberately does not have and a cookie-based tool only has for the subset of visitors who accepted a banner.

If you genuinely need multi-touch attribution modelling across a long consideration cycle, that is Adobe Analytics or Matomo territory, and it is a project with an owner rather than a setting. If you do not, resist. The time goes further into finding the step where half your visitors leave.

Where conversion tracking goes wrong

Counting the same conversion twice. A thank-you page that people refresh, or a fire-on-load event on a page reachable by the back button, inflates everything downstream. Test by converting yourself and refreshing.

Consent loss making the rate look wrong. If your analytics needs a banner, you are measuring conversions among people who accepted tracking and dividing by traffic from the same subset. The rate may be roughly right; the volumes are not, and any comparison with a period before the banner changed is meaningless.

Bot traffic in the denominator. Around a third of raw arrivals at a typical site are automated. If those are counted as visitors but never convert, your conversion rate is understated by roughly the same third, and it will move whenever a crawler discovers you. Tools that filter bots before counting avoid this entirely.

The rest of the field

Common questions

Related guides