Every other list of tracking mistakes is ranked by how often they happen, which tells you what to expect rather than what to fix first. Ranked by recoverability instead, the order changes completely: the faults that quietly corrupt months of historical data outrank the ones that break loudly, because a broken tag gets noticed and fixed while a wrong number gets reported to the board.
There is a reason tracking faults are more expensive than they look. A campaign that fails is visible, gets investigated, and gets fixed. A campaign that reports success it did not earn gets more budget. The mistakes below are ordered by how much of the damage survives the fix, because that is the property that decides what you should check this afternoon.
Which conversion tracking mistakes are hardest to undo?
The ones that corrupt historical data. A misfiring tag can be fixed today, but the months it already recorded cannot be recovered.
Ranked by
How much of the damage survives after the fault is found and corrected. A fault you can repair completely ranks below one that leaves you with a permanently unreliable record of the past.
Counting the same conversion twice
First because the damage is total and silent. Every historical number is inflated by an unknown multiple, nothing in the interface flags it, and no correction recovers the real figures for the months already recorded.
The usual causes are mundane. A Google Ads tag and a GA4 import both counting the same action. A tag hardcoded on the page by a developer while the same tag also fires from the tag manager. A conversion firing on a redirect and again on the destination. In every case the account looks like it is performing about twice as well as it is.
What makes this the worst entry on the list is what happens downstream. Automated bidding optimises toward the inflated number, so it will pay roughly twice what it should for the traffic that produces it, and it will do so confidently. The cost per acquisition on your report is half the real one, budgets get set from it, and the whole thing is self-reinforcing until somebody compares the platform against the CRM.
You can stop the double count today. You cannot go back and work out which of last quarter's conversions were real, so every year-on-year comparison that spans the fault is permanently compromised.
A conversion action that measures the wrong event
Second because the fix is easy and the historical damage is just as permanent. The data is not double counted, it is measuring something else entirely, and no amount of correction retrofits the right event onto past months.
The most common version is a conversion action firing on a page view rather than a form submission, usually because the thank-you page is reachable directly or because a single-page app never navigates. Any visitor who wanders onto that URL is recorded as a lead.
Ranked below double counting only because the direction of the error is often obvious once anyone looks: the conversion count is implausibly high and the sales team has never heard of most of them. Double counting produces numbers that look merely good, which is far harder to notice.
Optimising toward an action nobody qualified
Third because the tracking is working correctly and the choice of what to track is the fault. The record stays accurate, so the past remains readable, but every bidding decision made in the meantime was aimed at the wrong target.
This is not a bug, it is a decision that was never really made. Somebody marked a newsletter signup as a conversion because it was easy to measure, and the bid strategy has been dutifully buying more newsletter signups ever since, at the expense of the traffic that produced revenue.
It ranks below the first two because the data itself is intact. Once you change what you optimise toward, you can still read the history and see what happened. The waste was real but it is bounded and it is legible.
Consent and tag blocking quietly removing conversions
Fourth because the loss is partial, ongoing and at least estimable. You lose a share of your conversions rather than the truth about all of them, and the share can usually be bounded.
Ad blockers, tracking prevention in browsers and consent choices all remove some proportion of client-side conversion data. The result is under-reporting rather than corruption, which is a materially better failure: your cost per conversion looks worse than it is, so you make conservative decisions rather than expensive ones.
It is still worth fixing, and it is the case where server-side measurement genuinely earns its complexity rather than being adopted because it sounds advanced.
Tracking events that never drive a decision
Last because nothing is corrupted and nothing is lost. This is clutter, and clutter costs attention rather than money.
Scroll depth, video starts, and every button on the site marked as a key event. None of it is wrong, and none of it is load-bearing. The cost is that the two numbers that matter are harder to find among the thirty that do not, and that reports get longer while getting less useful.
It appears on this list at all because it is frequently ranked near the top of other lists, which is a good illustration of why frequency is the wrong axis. It is the most common item here and comfortably the least damaging.
Why rank by recoverability instead of by frequency?
Because frequency tells you what to expect and recoverability tells you what to check first, and only one of those is a priority order.
A list ordered by how often something happens is a description of a population. It is genuinely useful if you are writing a checklist for a hundred accounts. It is close to useless if you have one account and an afternoon, because the most common fault on the list may also be the most trivial, and it usually is.
Recoverability answers the question you actually have, which is what it costs you to find this late. A fault you can fully repair costs you the waste it caused until you found it. A fault that corrupts the record costs you that, plus every decision made from the record afterwards, plus the ability to ever prove what happened. Those are different kinds of expensive and they deserve different urgency.
How do you find these before they cost you a quarter?
Compare what the platform reports against what your own records show, monthly. Any gap that cannot be explained is a tracking fault until proven otherwise.
The single highest-value habit in paid media is also the least technical one. Once a month, take the conversion count the ad platform reports and put it next to the number of leads your CRM or your inbox actually received. They will not match exactly and they are not supposed to. What matters is whether the gap is stable and explainable.
A platform reporting 180 where the CRM shows 90 is not a rounding difference, it is the first entry on this list. A platform reporting 40 where the CRM shows 90 is the fourth. The direction of the gap tells you which fault you are looking at before you open a single tag.
This is why every account review we run checks measurement before it checks anything else, and why a bidding recommendation made on unverified conversion data is worth nothing. The reasoning is set out on marketing analytics, and what to count in the first place is ranked in conversion actions ranked by how well they predict revenue.
Marketing analytics services
Analytics work has two halves. The first is technical: events firing correctly, deduplicated, with the right parameters, server-side where the browser is unreliable. The second is interpretive: knowing which numbers can carry a decision and which cannot. Most reporting problems are actually the second half, dressed up as the first.
Read about Marketing analytics