Conversion Tracking

Running ads made visitor counts visible immediately, but what those visits led to stayed invisible. Three thousand clicks sat on one screen and 180 new sign-ups for the month sat on another, with nothing anywhere saying where those 180 people came from.

Early Pagedeck did the reconciliation by hand every month. Ravi exported the sign-up list, lined up each sign-up date against the campaign calendar, and credited the ads with anyone whose dates overlapped. The method collapsed as soon as several campaigns ran on the same day. When Grace asked how many people from last week's newsletter had reached a payment, the answer she got back was an estimate.

The missing piece was a device that recorded the moment of success as it happened. Instead of matching lists after the fact, the payment confirmation screen can write one line at the instant it loads: this person arrived through the newsletter and paid. Conversion tracking is the setup that plants that line in advance.

Start by deciding what counts as a conversion. That's an agreement, not an engineering task, and the definition settled here determines what every number downstream means.

1. Define the conversion. Whether a free sign-up counts, whether the bar is inviting teammates and actually using the product, or whether only a first payment qualifies will swing the same campaign's reported performance several times over. When there are multiple goals, rank them, name one primary conversion, and mark the rest as secondary.
2. Plant the moment as an event. This means adding code that fires a signal when the payment confirmation screen loads, usually the few lines known as a tracking pixel. Some teams have their own server send the signal instead, which is server-side tracking.
3. Attach the source. Knowing a conversion happened says nothing about which path earned it, so the tag carried in on the link — typically a UTM parameter — gets recorded alongside it. Traffic that arrives untagged has no recoverable origin later.
4. Send the result back to the ad platform. Platforms need conversions to report performance and to run automated bidding. That signal is a postback, and the form that has settled in is our server telling the platform's server directly.
5. Fire one real conversion. Put a test payment through, confirm it appears on the platform's dashboard, and confirm it isn't counted twice. Skipping this step is how teams lose a month of data, and the check costs about ten minutes.
6. Write the definition down with the date it changed. Touching the definition breaks comparability between last month and this one, so record when it moved and don't read across the seam.

It fits best when the goal action happens at one clear moment and that moment lives on a screen you own. Sign-up complete or payment complete gives you a single page to point at, a clean record, and a conversion rate you can calculate directly.

Four situations fit poorly.

The first is when the recording itself gets blocked. Browsers and phones keep restricting tracking storage, so conversions go missing. Catching only 80 of every 100 is common, and because the loss isn't even across channels it distorts comparisons too.

The second is a journey that crosses devices. Someone sees the ad on a phone during the commute and pays on a desktop at the office, and the two records stay two different people. Logged-in users can be stitched together by account; visits before login mostly can't, which means the channel that started the journey often loses its record entirely.

The third is a conversion that happens off-screen. When a form fill is followed by a salesperson closing the deal over the phone, the moment of success isn't on any page you control. Track the inquiry and reconcile the outcome against sales records separately.

The fourth is a long consideration cycle. If four months pass between first visit and contract, putting this month's ad spend next to this month's conversions stops meaning anything.

One more limit is interpretive rather than technical. Conversion tracking records what happened; it doesn't divide credit. When one person arrives through search, then a newsletter, then an ad before paying, the attribution rule decides whose column the conversion lands in. Accurate records still flip the channel rankings the moment that rule changes.

"There's no conversion tracking on this campaign. We can't report on this week."

The ads ran without the recording device, and the missed period can't be reconstructed.

"I'm moving the conversion from sign-up to first payment, so it won't line up with last quarter."

Changing the definition changes what the number means and breaks the time series.

"The platform says 40 and our own data says 31."

Platform-side and first-party counts disagree, and someone needs to find out whether it's duplication or loss.

At Pagedeck, installing this is usually a thirty-minute session with Ravi and Theo sitting together. Ravi decides what gets counted and which platforms receive the signal; Theo adds the code at the payment-complete step. Then Ravi runs a real card payment through and waits until that single conversion appears on the platform dashboard before any ads go live.

A month later Ravi moves budget around based on conversions by channel. A channel with few conversions stays if its return on ad spend (ROAS) is high, and a channel with plenty of conversions that never reach payment gets cut.

The most common assumption is that once the setup is on, the numbers are simply true. It all tallies automatically, so there seems to be no room for human error. In practice blocking, duplication, and loss are always mixed in, and gaps above 10 percent between platform dashboards and first-party data are ordinary. Report the platform's figure without knowing that, and budget keeps flowing to whichever channel overstates itself most.

Tracking as many conversions as possible is the other one. Tag every button click and you end up with more than twenty conversions, at which point nobody can say which one is the actual goal. Automated bidding optimizes toward whichever conversion it is pointed at, so spreading the target across many events spends ad budget buying cheap actions. Keep one primary conversion and demote the rest to reference. The secondary events still get recorded and still answer questions later; they just stop competing for the budget.