Your conversions are double counted: how to find it

Find and fix duplicate conversion recording using a short diagnostic sequence, and understand what the correction will do to your reported numbers.

The symptom

Your reports show roughly twice the orders your back office recorded. Or a tidier version of the same problem where the ratio is not quite double because only some journeys duplicate. Conversion rate looks excellent. Cost per acquisition looks excellent. Nothing in the bank agrees.

Double counting is a configuration fault and almost always an accident of history. Someone added tracking. Someone else added tracking again because they could not tell whether the first attempt worked. Both are now live and both are firing.

What fixing it costs and what it cannot undo

The fix does not recover the budget already misallocated and it does not repair your history. Past months stay inflated. The honest response is to annotate the date of the correction and tell anyone comparing periods that the series breaks there.

Expect a few hours of auditing and possibly developer time where a duplicate snippet is hardcoded into a theme or a template. Expect an awkward conversation too because reported performance will get worse on the day you fix it. If automated bidding was being fed the duplicated signal it needs time to recalibrate and performance can wobble while it does. That cost is real and still smaller than bidding on numbers that are wrong.

Who should fix it now and who can wait

If you spend nothing on paid media and nobody is bidding on these conversions then this is an annoyance rather than an emergency. Note the fault. Correct it when you next touch the site. Do not buy an audit this month because the only thing being distorted is a report you can mentally adjust.

The same applies if you are replatforming this quarter. Auditing tracking on a site you are about to retire is wasted work.

Fix it immediately if automated bidding or a value-based strategy is running on the affected conversion. Every duplicate teaches the algorithm that an audience is twice as valuable as it is and it will keep buying more of whatever produced the duplicate. The same urgency applies if a retainer or an incentive is paid against reported conversions.

The diagnostic sequence

Start by listing what is actually installed. Open the page source or use the vendor tag assistant and search for every measurement identifier and pixel identifier present. Finding the same identifier twice is the single most common cause. A theme setting with a field for the ID plus a plugin plus a tag manager container will happily install three copies of the same thing.

Then complete one real conversion yourself and watch the debug stream while you do it. Count the events that arrive from your single action. Refresh the thank you page and check whether the count rises again because a purchase event with no order identifier attached fires on every reload and every back navigation. Next check your ad platform for two paths to the same outcome. A native conversion tag and an imported analytics conversion both marked as primary will each claim the sale. Check the counting setting on each conversion action since one counts every conversion and the other counts one per click. For Meta confirm that browser and server events share an event identifier because without one the deduplication cannot work.

Finally reconcile a full week. Count recorded conversions against orders or enquiries in your own records and the ratio tells you how much of the problem you have found. Repeat the sequence on mobile since some duplicates appear on only one device class.

Knowing the fix held

The metric that matters is the ratio of recorded conversions to real orders. It should settle slightly below one because consent refusals and blockers always cost you some events. Watch it weekly for a month. A ratio that drifts upward again means something was re-added.

The metric that flatters is any platform diagnostic saying no issues detected. Those checks confirm that a tag fired correctly and not that it fired once. A falling conversion count after your fix is the signal of success which makes this the rare repair where the numbers are supposed to get worse.

The common mistake

The expensive error is deleting a duplicate without checking which one the bidding depends on. Remove the tag the campaign optimises against and the signal goes to zero rather than halving. Bidding then throttles delivery and a measurement problem becomes a revenue problem within days. Identify the primary conversion action first. Remove the other. Verify delivery for a week afterwards.

The quieter mistake is leaving both installed and dividing by two in the monthly report. It works until a journey appears that only fires once and then your correction becomes the error. If the implementation is tangled enough that this feels tempting the cleaner answer is a rebuild of the measurement layer and the published scope of an analytics and measurement setup covers that work.

What to expect for a fortnight afterwards

Tell the people who read the reports before you change anything. Record the date in your analytics property as an annotation. Hold campaign budgets steady for a couple of weeks so the only variable is the correction. Do not judge any channel on the days immediately after the change because bidding is recalibrating and the comparison is meaningless until it settles.

This guide explains how something works. It is not legal, tax or financial advice, and rules change. Confirm anything regulatory with a qualified professional before you act on it.