Conversion tracking setup that survives a website change

Every business that has rebuilt a website has a version of this story. The new site launches, everyone is pleased with how it looks, and three months later somebody notices the lead numbers went to zero on a Tuesday in March.

The leads did not stop. The counting stopped. And because nobody checked for twelve weeks, there is now a hole in the data that cannot be filled in, right through the period you most wanted to compare against.

Why does tracking break during a redesign?

Because most tracking is attached to things that were never meant to be permanent.

A tag fires when a button with a particular ID is clicked. The new site rebuilds that button and the ID changes. Nothing errors, nothing alerts, the tag simply never fires again.

A form plugin gets replaced by a different one, so the listener that watched the old form has nothing to watch.

URLs change during a rebuild, so a conversion defined as a visit to a specific page stops matching.

The old tag manager container never gets installed on the new site, usually because it lived in a theme file rather than in a settings field.

And the most common one of all: the tracking was set up inside an account belonging to a previous agency or developer, and nobody currently working on the site can even see it.

Tracking does not break loudly. It goes quiet, and quiet looks exactly like a slow month.

What does a silent tracking gap actually cost?

Put a number on it once and you will never skip the monthly check again. Use your own figures in place of these.

Set the baseline. Your site averaged 22 counted conversions a month across the six months before the rebuild.

Set the break. The new site launches 3 March. The submit button is rebuilt, its ID changes, the tag stops firing. Nobody notices until the June report lands on 12 June.

Count the dark days. 3 March to 12 June is 101 days, about 3.4 months. At 22 a month that is roughly 74 conversions you can never attribute to anything.

Price the spend running blind. If you were spending $1,800 a month on search ads with bidding optimized against that conversion signal, 3.4 months is about $6,100 of spend optimizing toward an event that stopped existing. The platform did not error. It just stopped learning.

Price the comparison you lost. You rebuilt the site to convert better. You now have no before and after for the three months following launch, which is the exact window the rebuild was supposed to prove itself in. That comparison cannot be reconstructed, because analytics does not backfill events that never fired.

Then price the prevention. One test submission on the first of each month, checked against the analytics report, takes about sixty seconds. Twelve of those is twelve minutes a year.

Twelve minutes a year against $6,100 and a quarter of missing history. That is the entire argument for the calendar reminder at the end of this article.

What setup actually survives a rebuild?

Four decisions, made before anyone touches design.

Own the accounts. Analytics, tag manager, ad platforms, and search console should be created under an email address your business controls, with your developer and your agency added as users. This one decision prevents most of the disasters in this article.

Use a thank you page as the primary conversion. A form redirect to a unique URL is the most durable event available, because it survives plugin changes, theme changes, and most platform migrations. Keep the same URL through the rebuild and the conversion definition still works.

Put everything in one tag manager container, installed through a supported setting rather than pasted into a template file. Templates get overwritten. Settings usually carry across.

Name events plainly and consistently. form_submit_contact, call_click_header, booking_complete. Six months later you will not remember what event_17 was, and neither will anyone else.

Then add secondary events on top, like the field-level ones that let you do form abandonment analysis without paying for a tool. Those can break without costing you the core number.

What should be in the tracking document?

One page, written before the build, handed to whoever is doing it.

Every conversion you count, listed by name, with the URL or the element it fires on. Every ad platform that needs its own tag. Every URL that must not change, especially the thank you page. Who owns which account. And what the numbers looked like the month before launch.

That last item is the one people skip and regret. Without a baseline you cannot tell whether the new site converts better, worse, or the same. You will have a launch date and no before picture, and every argument afterward becomes a matter of opinion.

Make it an acceptance criterion, not a hope. The site is not done when it looks right. It is done when a test submission appears in analytics as the correctly named event. Put that sentence in writing with your developer, because what a small business website actually costs should include the tracking work rather than treating it as an extra later.

What can I actually measure, and what is invisible?

This is the part of the conversation most agencies skip, and it changes what you should promise yourself.

Your Google Business Profile reports actions, not rank. Google's own performance report for a profile publishes Interactions, Searches, Views, Directions, Calls, Website clicks, Messages, Bookings, Booking clicks, Products, Menus and Offers. There is no rank dimension and no position dimension anywhere in it, and Search Console has none either. Anyone converting a local ranking change into a traffic forecast is using a number Google does not publish.

A growing share of search never produces a click at all. SparkToro, working with Similarweb panel data across US Google searches from January to April 2026, put the share of searches ending without any click at 68.01%, and reported that per 1,000 US searches only 276 clicks now reach the open web against 374 in 2024. Google has publicly disputed SparkToro's methodology and a separate 2026 measurement using a different definition put zero-click at 27.6%, so use the direction rather than the decimal. Either way, your analytics can only ever see the visits, never the impressions that resolved without one.

Clicks from AI features are not separable. Google states that traffic from its AI features arrives inside ordinary organic search traffic. There is no referral parameter, no separate channel in your analytics, and no way to isolate it. In June 2026 Google added generative AI performance views to Search Console, which report impressions by page, country, device and date, and explicitly do not report clicks, click-through rate, position, or the prompt the user typed. AI Overviews and AI Mode are not separated from each other inside that view either. Search Engine Land has already reported a data bug in the report, so sanity check anything unusual before you act on it.

Native app traffic lands in Direct. Clicks from a ChatGPT or Perplexity mobile app often arrive with the referrer stripped, which means your analytics files them under Direct alongside people who typed your URL.

None of this is an argument against tracking. It is an argument for tracking the things you own, which are your own events, your own phone lines and your own booked jobs, and for refusing to pay anyone who promises visibility into channels the platforms do not expose.

How do I verify it after launch?

Manually, on launch day, and then again a week later.

Submit the form yourself with a real name and a note saying it is a test. Confirm three things: the email arrives in the inbox someone reads, the thank you page loads, and the event shows in analytics as the name you chose.

Tap the phone number on a phone and confirm the click event fires.

Then check the ad platforms separately. A conversion visible in analytics is not necessarily being imported into the ad account, and campaigns optimizing on a dead conversion signal will quietly waste money. That is a fast way to end up paying for clicks that land somewhere that cannot convert without any warning in the reports.

Do it again seven days later, because caching, plugin updates, and last minute developer changes all land in that first week.

Then put a calendar reminder for the first of every month to confirm the numbers are still moving. Sixty seconds a month is the whole maintenance cost of never losing a quarter of data again.

What to do this week

Log in to your analytics account and check whose email address owns it. If the answer is a former agency or a developer you no longer work with, fix that first, this week.

Submit a test through your own form and confirm it appears as a counted conversion. Most owners have never done this, and a meaningful share discover the count has been broken for months.

Write the one page tracking document even if no rebuild is planned. It takes twenty minutes and it is what you hand to the next person who touches the site.

Run the gap calculation above with your own conversion count and ad spend. It converts a maintenance chore into a dollar figure, which is the version that survives a busy month.

Note your current monthly conversion count somewhere outside the analytics tool, in a spreadsheet or a notebook. That is your baseline, and it is what makes the conversion math that decides whether ads make sense possible to run at all.

Questions about your tracking?

Email me at eric@seod.com with a short description of your setup: what platform the site is on, what form tool you use, and whether you have a thank you page. If a rebuild is coming, say so. I will send back the list of what is most likely to break and the one change that would prevent it.

It takes me a few minutes and there is nothing attached. If your setup is already durable, I will tell you that instead of finding work to do.

There is more on conversion in the library.