Conversion tracking is how you tell which visits turned into something real. Without it you are buying clicks and hoping. With it you can see which campaign produced the enquiry, and which one produced nothing at all.
Every other number in a performance marketing report sits on top of it. So a broken tag does not only lose data. It quietly moves next month's budget to the wrong place.
What conversion tracking actually records
A conversion is an action you decided matters. A form sent, a call answered, an order placed, a booking confirmed.
Conversion tracking joins that action back to the visit that caused it. A visit usually arrives carrying a label that says where it came from. When the action happens, your site hands that label back to the platform. So the report can say which advert paid for the result.
Choosing what counts as a conversion
Most accounts count too many things. Then every number looks healthy and none of them means much.
Pick one primary action per campaign. Usually that is the enquiry or the sale. Everything else is context, and context should never sit inside the cost per result you report.
- Primary: the thing you would happily pay for, such as a sale or a real enquiry.
- Secondary: useful signals like a brochure download or a visit to the price page.
- Ignore: scrolling, time on page and anything nobody would ever act on.
Putting a value on a conversion
A conversion with no value attached is only a count. Two campaigns can produce the same number of enquiries and still be worth very different amounts.
Where you know the order total, send it. Where you do not, send a considered figure rather than nothing. An average order value from your own accounts works. So does the worth of a typical job, taken from your own records rather than an industry figure.
Then keep that figure honest. Review it when your prices change. Also write down where it came from, because a value nobody can explain gets ignored the moment somebody questions it. Values are what let you compare two very different campaigns. That comparison is the basis of using data to decide where the budget goes.
How conversion tracking works behind the scenes
Three pieces have to line up. A tag on the site, an event with a name, and a way to join that event to the original click.
The tag loads with the page. Your event fires once somebody finishes the action. That click label, stored when the visitor arrived, travels along with it. If any one of the three goes missing, the platform records nothing, and a working campaign looks like a failure.
Conversion tracking through a tag manager
A tag manager gives you one place to add and change tracking without touching the site code. That suits most teams, because a marketing change no longer waits for a developer.
It also creates a single point of failure. When somebody edits the container badly, several conversions stop at once. So write down every change, and check the account the day after you make one.
Why two reports never quite agree
Your ad platform and your analytics will show different totals. That is normal, and chasing an exact match wastes weeks.
Each tool answers a different question. An ad platform asks what a click eventually produced, so it credits the day of the click. An analytics tool asks what happened during a visit, so it credits the visit that finished the job. Windows and credit rules differ too, and these are defaults rather than fixed laws. Read your own settings instead of assuming. Then pick one report as the source for decisions and treat the other as a sanity check.
Attribution windows, in plain English
A window is how long after a click the platform still takes credit for a sale. Short windows undercount slow decisions. Long ones flatter whatever ran most recently.
Match the window to how people really buy from you. Somebody with a burst pipe decides in an hour. A school choosing a supplier takes a term. Then leave the setting alone, because moving it halfway through a quarter makes every comparison meaningless.
Consent, and what it does to conversion tracking
A consent banner is not decoration. If your site honours a refusal, part of the measurement stops, so reported conversions fall even though sales did not.
Expect a gap between what you sell and what the platform sees. Some platforms fill part of that gap with an estimate rather than a counted event. Ask which of your figures are counted and which are modelled, then keep the two apart in every report. Consent rules also differ by country, and they change. So check what applies to your own business rather than copying another site's banner.
The conversions that happen away from the website
Phone calls, walk ins and quotes agreed over email never touch a form. Any setup that ignores them describes a fraction of the truth.
Tracked phone numbers help. A short source question on the quote form helps. An import from your sales list helps most of all, since it carries the outcome rather than the enquiry. None of these is perfect. Still, a rough record of where a customer came from beats an exact record of nothing.
Junk conversions, and keeping them out
Robots fill in forms. Staff test the checkout. Suppliers ring the tracked number.
Every one of those becomes a conversion, and your cost per result improves for no reason at all. Add a hidden honeypot field to catch the robots. Filter traffic from your own office. Mark a bad lead as junk in the report rather than deleting it, so you can still see how much of it arrives.
A worked example, with invented round numbers chosen to make the arithmetic easy rather than to describe a typical result. Say a campaign spends 200 and the platform reports 20 conversions, so each one appears to cost 10. Now suppose 5 of the 20 were robots or wrong numbers. The real count is 15, and each real conversion cost about 13. Nothing about the tracking changed. Only the definition did.
Testing conversion tracking before you trust it
Never assume a tag works because somebody says it does. Test it yourself, on the live site.
- Submit the form from a phone, on mobile data rather than the office network.
- Watch the event arrive in whatever preview or debug view your platform offers.
- Check the value and the currency field carry what you expect.
- Repeat the whole test after any website change, however small it looked.
- Ask a colleague to try it in a browser you never use.
Then put a monthly reminder in the calendar. Tracking breaks quietly, usually during a site update, and nobody notices until a report looks strange weeks later.
Conversion tracking mistakes that cost the most
A handful of faults turn up again and again, and each one bends a decision.
- Counting a thank you page a visitor can reload, which inflates the total.
- Two tags firing for one purchase, so every sale appears twice.
- Sending no value at all, which makes the return on your spend look terrible.
- Treating every click as a conversion, until nothing stands out.
- Changing a definition halfway through a campaign, which ruins the history.
Reading the numbers without fooling yourself
A tracked conversion is a signal, not a sale. Somebody who fills in a form has bought nothing yet.
So look at what happens next. If half your enquiries never answer the phone, your true cost per customer is double what the dashboard claims. That is why cost per acquisition only means something once you know how many enquiries turn into customers.
The same applies when you report upward. Keep a small set of numbers, write the definitions down, and say which ones came from a real event. Our note on choosing marketing numbers covers the shortlist worth keeping.
When conversion tracking breaks, what to check first
Start with the date. Find the day conversions stopped, then look at what changed that day.
Usually it is a website release, a new cookie banner, a renamed thank you page, or a tag somebody removed while tidying. Confirm the tag fires on the live page before rebuilding anything. Most of the time the repair takes minutes once you know the cause.
Where conversion tracking ends and judgement begins
Tracking tells you what happened. It will never tell you what to do about it.
A campaign with a poor cost per enquiry can still deserve its budget, because it brings the customers who stay longest. A page can convert well and pull in the wrong people entirely, which is why landing page work and tracking belong together.
So treat the numbers as evidence in an argument you are still having. Good conversion tracking makes that argument short and honest, which is all any reporting can really do.
Frequently asked questions
What is conversion tracking in simple terms?
It is the link between an advert and something valuable that happened afterwards. A visitor clicks, arrives with a label attached, and later sends a form or buys something. Your site reports that action back to the platform along with the label. So the platform can name the advert or campaign that produced the result, instead of only counting clicks. It cannot tell you whether the enquiry was any good.
How do I know whether my tracking is working?
Test it yourself rather than trusting a screenshot. Send a real form from a phone, off the office network, and watch the event appear in whatever preview view your platform offers. Then wait a day and check that the same action shows up in the report itself. A tag can fire correctly and still land under the wrong campaign, so the report is the only proof that settles it. Note the date you tested, because the next person will want it.
Why do the ad platform and analytics show different numbers?
Because each tool answers a different question, and each applies its own window and credit rules. A gap is normal and it will never close completely. What matters is whether the gap stays steady. A stable difference between two reports is a quirk you can live with, while a gap that suddenly widens usually means something broke on the site. Pick one report as the source for decisions, then use the other to spot that change.
Should a phone call count as a conversion?
Yes, whenever calls are how people actually buy from you. Call tracking usually works by showing a different number to different visitors, so the number dialled points back to the source of the visit. If your setup can ignore very short calls, use that, since wrong numbers otherwise inflate the count. Suppliers and staff will ring the tracked number too. Review the call log now and then rather than trusting the total.
How many conversions should I track at once?
One primary action per campaign, and a small number of secondary ones for context. Beyond that the reports fill up and nothing stands out. Keep secondary actions out of any cost per result you quote, since mixing a brochure download into the same figure as a sale makes the campaign look cheaper than it is.
What does a cookie banner do to my numbers?
It usually reduces them. When somebody refuses consent, part of the measurement stops, so the platform sees fewer conversions than the business actually earned. Nothing has gone wrong with the campaign itself. The trap is comparing a period before the banner with a period after it, then reading the drop as a decline in performance. Compare like with like instead, and treat your own sales records as the real total.
Can I track sales that close offline?
Yes, by sending the outcome back rather than only the enquiry. If your platform accepts an import, export the leads that became customers and match them on the click label captured when they first arrived. Where that is too much work, a source question on the quote form gets you most of the way. Conversion tracking that stops at the website form describes only part of the business.
How often should I test conversion tracking?
Once a month as a habit, and immediately after any change to the website, the forms or the consent banner. Put the monthly test in the calendar so it survives holidays and staff changes. Most broken conversion tracking goes unnoticed for weeks, and the damage is not the missing data. It is the budget decisions made from a report that quietly lost a campaign.
Do I need a tag manager to set this up?
Not always. A tag manager helps when several people need to change tracking and nobody wants to wait for a developer. A small site with one form may be simpler with the code placed straight into the page. The trade with a container is that one careless edit can stop several conversions at once. Whichever you choose, keep a written log of changes, then check the account the day after anybody edits anything.