Ad platforms miss conversions because the signal that links a sale to an ad travels through the visitor's phone or browser, and several things now interrupt it on the way. Apple's tracking prompt, Safari's cookie limits, ad blockers and consent refusals each remove a share of events, and automatic bidding gets worse as a result.
How a conversion normally reaches the platform
When someone clicks an ad, the platform adds a click ID to the address of your landing page. Meta's is called fbclid, Google's is gclid and TikTok's is ttclid. The platform's tag on your site reads that ID and stores it in a cookie, which is a small note the browser keeps for your site.
Days later the visitor buys. The tag reads the cookie and tells the platform: this purchase, this value, this click. The platform credits the ad.
That chain has four links: the click ID, the cookie, the tag, and the request the tag sends. Each cause below breaks one of them.
iOS App Tracking Transparency
Since iOS 14.5, every iPhone app has to ask permission before it tracks a person across apps and websites owned by other companies. This is the prompt that offers "Ask App Not to Track". Apple calls the system App Tracking Transparency.
When someone declines, the app cannot use the device's advertising identifier for that person, and the ad platform limits how it connects what the person does on your site to the ad they saw in the app. Some of those conversions are reported late, in aggregate, or as statistical estimates. Some are not reported at all.
For an advertiser, the effect is a gap between sales that happened and sales the platform can tie to a specific ad, and the gap is largest among iPhone users.
Safari's cookie limits
Safari includes a feature called Intelligent Tracking Prevention. It blocks third-party cookies, which are cookies set by a site other than the one you are visiting. It also shortens the life of first-party cookies when a script writes them. Those cookies expire after seven days. In some cases, when the visitor arrives from a known tracking domain with a click ID in the address, the limit is 24 hours.
Consider a visitor who clicks your ad on a Monday, thinks it over, and comes back to buy ten days later. The cookie holding the click ID is gone. The tag still fires, and the platform receives a purchase it cannot connect to the click. Products with a long decision time lose the most this way.
Ad blockers and private browsers
Ad blockers work from lists of addresses known to belong to advertising and analytics services. When a page asks for a script from one of those addresses, the blocker cancels the request. The tag never loads, so it never fires. Some browsers ship with this kind of protection built in.
Nothing looks broken when this happens. The page works, the order goes through, and your shop records it. The platform simply never hears about it.
Consent refusals
In the EEA, the UK and Switzerland, sites generally ask visitors for permission before using advertising cookies. A visitor who refuses should not be tracked for advertising, so the tag stays silent for that visit.
This is the system working as designed. It is also a real share of your conversions that the platform will never see as individual events. Our article on consent mode and conversion tracking explains what may be sent before and after a choice is made.
Smaller leaks
- The thank-you page never loads. The buyer pays on an external payment page and closes the tab. A tag that fires on the thank-you page never runs.
- The tag loads too late. On a slow connection the visitor leaves before the script has finished loading.
- The site's own security policy blocks the tag. A strict Content Security Policy can stop a script without any visible error. We cover this in when a Content Security Policy blocks your tag.
- The click ID is stripped. A redirect between the ad and the landing page drops everything after the question mark in the address.
The causes at a glance
| Cause | What breaks | What can help |
|---|---|---|
| App Tracking Transparency | The link between the app user and the site visitor | Complete, well-matched events; your own attribution alongside the platform's |
| Safari cookie limits | The cookie that stores the click ID | Server-side delivery; hashed e-mail matching, with consent |
| Ad blockers | The tag and its requests | Server-side delivery; serving the tag from your own domain |
| Consent refusal | Nothing is broken; the visitor said no | A clear banner; modeling by the platform |
| Closed tab after payment | The thank-you page event | Reporting the order from the server |
Why bidding suffers
Most campaigns now use automatic bidding. You tell the platform what you want, such as purchases, and its algorithm decides who sees the ad and how much to bid. The algorithm learns from the conversions it receives. Missing conversions hurt it in three ways.
It learns more slowly. A campaign needs a steady flow of conversions before delivery settles. Fewer reported conversions means a longer and more expensive learning period.
It learns from a skewed sample. The conversions that go missing are not spread evenly. They come more often from iPhone and Safari users and from people who block ads. The algorithm sees fewer buyers in those groups and can drift away from them, even when they are good customers.
Your reports mislead you. Here is a hypothetical example. Suppose a shop spends 2,100 USD and gets 100 orders from its ads, and the platform only sees 70 of them.
- Reported cost per purchase: 2,100 ÷ 70 = 30 USD
- Real cost per purchase: 2,100 ÷ 100 = 21 USD
If the shop can afford 25 USD per order, the report says the campaign is losing money and the truth says it is profitable. Decisions made on the reported number would cut a campaign that works.
What you can recover, and what you cannot
Some of the gap can be closed. Server-side tracking delivers events that browsers block or drop. Sending a hashed e-mail address with the purchase, where the visitor has consented, helps the platform match buyers whose cookies have expired. Consistent UTM parameters let you run your own attribution from your own data, so you have a second view that does not depend on any platform.
Some of the gap should stay. A visitor who refuses tracking has made a choice, and a setup that works around it puts you at legal risk and treats your customers badly. The goal is to capture every conversion you are allowed to capture, and to know roughly how large the rest is by comparing platform numbers with your shop's order count.
Where Tagbreeze fits
Tagbreeze sends conversions from the browser and, for Meta and TikTok, from its servers too, with deduplication. It waits for consent in the EEA, the UK and Switzerland. Its attribution reports show each purchase with its first and last touch beside what the platform itself reports, so the gap is visible.
In short
- A conversion reaches the platform through a click ID, a cookie, a tag and a request. Each one can fail.
- Apple's tracking prompt, Safari's cookie limits, ad blockers and consent refusals are the main causes.
- Missing conversions slow down bidding, skew it, and make profitable campaigns look expensive.
- Server-side delivery, better matching and your own attribution close part of the gap. Refusals are meant to stay untracked.