Conversion Tracking Without the Guesswork

A conversion is whatever you would be pleased to see a visitor do. That definition is doing more work than it looks, because most conversion tracking goes wrong at the point of deciding what to count, not at the point of implementing it.

Pick the thing that changes a decision

The test is simple: if this number moved, would you do anything differently? If not, do not track it.

"Visited the pricing page" fails that test for most sites. It moves with traffic, so it tells you what your traffic did, not what your site did. "Started a trial" passes, because if it drops you go looking for a cause.

Start with one. Add the second when you have run out of things to learn from the first.

The three shapes a goal comes in

In Galytics a goal is one of three things, and choosing the right shape matters more than the name you give it.

A page goal fires when someone reaches a URL. Wildcards are supported, so /checkout/success* catches every variant with one goal instead of forty. Best for anything with a confirmation page.

An event goal fires when your own code triggers a custom event. You call galytics.triggerEvent('trial-started') at the moment it happens, and the goal counts it. Best for anything without a page change, which today is most interactions.

A scroll goal fires when a visitor gets past a depth you choose, on pages you choose. Best for sites where the valuable thing is being read. If you publish documentation or long-form writing, this is often your only honest conversion.

One detail worth knowing about events: the event name has to be registered in settings before it will be accepted. Anything unrecognised is discarded. That feels fussy for about a minute, until you realise it means a typo in a template cannot invent a phantom event and quietly consume your quota.

Count the same thing your ad platform counts

If you run Google Ads, the conversions you optimise against should be the ones you actually care about, not landing page views. Otherwise bidding gets very good at buying visits from people who leave.

Galytics can export chosen goals to Google Ads as conversions. You pick which goals count, they go back to the ad platform, and bidding optimises against something real.

The reason to do it this way rather than installing Google's own conversion tag: you get the feedback loop without putting an advertising company's tracking script on your pages. If you moved off Google Analytics for privacy reasons and then added Google's conversion tag, you have mostly moved the problem.

Read conversion rate carefully

Two traps, both common.

The denominator matters. A conversion rate of "conversions divided by all visitors" answers a different question from "conversions divided by people who reached the form". The first tells you about your traffic mix, the second about your form. Neither is wrong. Confusing them is.

Small numbers lie confidently. Four conversions from 80 visits is 5%. Five from 80 is 6.25%. That is a 25% relative improvement and it is one person, who may simply have been in a good mood. Wait for volume before you believe a change, and be especially suspicious of a redesign that "improved conversion by 30%" in its first three days.

Then go and watch what actually happens

The number tells you how many people finished. It cannot tell you why the others stopped, and it will never volunteer that your submit button sits under a sticky header on a 375px screen.

That is what session replay is for, and it is the reason the two belong in the same tool. Notice the drop in the goal, filter to the sessions that failed, watch four of them. The cause is usually embarrassing and takes ten minutes to fix.

A workable order

1. Decide the one action that matters
2. Pick the goal shape that matches how it actually happens
3. Let it run for two weeks without touching anything
4. Export it to your ad platform if you buy traffic
5. Watch the sessions that did not convert
6. Change one thing, then measure again

Most conversion tracking setups fail at step one and get very sophisticated about steps two onwards. Getting step one right is worth more than everything after it.