Problem
Conversions dropped after Consent Mode: how to fix it
A drop in recorded conversions after a consent implementation goes live is expected up to a point, because some visitors decline and their activity should not be collected. Beyond that point the loss is avoidable, and it usually comes from tags being blocked outright rather than told to wait, or from signals that never arrive after a choice is made. The aim is measurement that degrades gracefully, not measurement that stops.
By CartKernel · Last reviewed 2026-09-07
Does this look familiar?
- Recorded conversions fell sharply on the day the consent banner or its settings changed
- The loss is concentrated in European traffic while other regions look normal
- Ad platforms report far fewer conversions while store orders are unchanged
- Smart bidding became unstable or re-entered a learning period without any campaign edit
- Sessions fell as well as conversions, not just conversions on their own
- Modelled conversions never appeared in reporting even though the implementation is described as advanced
Causes, ranked
Why it happens, most common first
Check them in this order. The first two account for most cases we open.
- most common
The consent tool blocks tags instead of signalling to them
In a blocking setup the tag never loads until consent is granted, so nothing at all is sent for visitors who decline or ignore the banner. In a signalling setup the tag loads immediately in a restricted state and sends a consent aware ping. Only the second allows any modelling to happen.
- most common
The consent update never fires
Defaults are set to denied and the tool is expected to send an update the moment a choice is made. If the update fires late, only on a page change, or not at all when a visitor closes the banner without choosing, every one of those sessions is treated as denied for its entire life.
- common
Not all four consent signals are being sent
Advertising measurement now depends on separate signals for advertising storage, analytics storage, advertising user data and advertising personalisation. Sending only the older two leaves advertising features unable to use the data even when the visitor granted permission for it.
- common
Ad click identifiers are being redacted or lost
When advertising storage is denied, click identifiers can be removed from requests and are no longer stored in the browser. Without the setting that carries the identifier through page navigation, the click cannot be joined to the conversion even for visitors who later grant consent.
- common
There is not enough data for modelling to run
Modelled conversions require a minimum volume of both consented and unconsented traffic before the platform will estimate the gap. Smaller accounts and accounts that only recently switched to a signalling setup may see no modelling at all for a period, which reads as a permanent loss.
- occasional
Banner design pushed the decline rate up
A redesign that gives equal prominence to reject, or that adds a granular preferences layer, changes the share of visitors who grant permission. The measurement is working correctly, the input changed. This shows as a step change in the consent rate rather than a technical failure.
- occasional
Region settings apply denied defaults too widely
Region specific defaults are meant to apply strict behaviour where it is required and lighter behaviour elsewhere. A configuration that applies denied defaults globally, or that lists regions incorrectly, silently reduces measurement in markets where it was not necessary.
The fix
In this order
Each step is something you can do today. Do them in sequence; skipping ahead is how a review fails twice.
Prevent it next time
- Monitor the consent grant rate as a metric, alongside sessions and conversions
- Test both a grant and a decline journey after any change to the banner or the tag setup
- Keep region defaults in a documented list that someone owns and reviews
- Note consent configuration changes in the same log as site releases and campaign edits
Measure the consent rate before diagnosing anything
Find what proportion of visitors grant each permission, broken down by region and device. Without that number you cannot tell a legitimate decline rate from a broken implementation, and every other reading is guesswork.
Confirm tags load in a waiting state
Open the site with tracking inspection enabled, decline the banner, and check whether the measurement requests are sent at all. If nothing leaves the browser, the tool is blocking. Reconfigure it so tags load with denied defaults and send consent aware requests.
Verify the update signal on both paths
Test a grant journey and a decline journey separately, and watch for the consent update immediately after the choice. Then test closing the banner without choosing, which is the case most implementations handle badly.
Send all four signals with correct defaults
Make sure advertising storage, analytics storage, advertising user data and advertising personalisation are all declared, with defaults appropriate to each region you sell in, and all four updated together when a choice is made.
Turn on identifier pass through and redaction settings
Enable the setting that carries the ad click identifier through page navigation when storage is denied, and set redaction behaviour deliberately rather than leaving it at whatever the tool installed. This preserves the link between click and conversion where it is permitted.
Add enhanced conversions to recover matched orders
Pass hashed customer details from the order record for visitors who consented to advertising use. This restores conversions that would otherwise be lost to storage limits, independently of the consent signal path.
Move purchase collection to the server where practical
Sending the purchase from the order record rather than the browser reduces dependence on browser storage while still respecting the consent state recorded for that visitor. It is the durable answer for stores where the consented population is large but browser collection is fragile.
Rebaseline reporting and tell everyone the date
Record the day the configuration changed, note the new consent rate, and compare performance forward from that date. Historical comparisons across the change will mislead every report and every bidding decision made from them.
When to get help
Get help when the store sells across several regulatory regions, when legal counsel has set requirements the current tool does not meet, or when bidding has degraded because the platforms lost too much conversion signal. That work sits between legal obligation and technical design, and it usually resolves into a signalling configuration with server side collection and enhanced conversions rather than any single setting. It is worth doing once, carefully, rather than repeatedly adjusting a banner.
Free. We reply within one business day.
How much of a conversion drop is normal after implementing consent controls?
It depends entirely on your consent rate and where your customers are, so there is no standard figure. The useful test is whether the drop matches your measured decline rate. If far more conversions were lost than visitors declined, something is blocking rather than signalling.
Does modelling put the missing conversions back?
Modelling estimates conversions from visitors who declined, using patterns from those who consented, and it only runs when there is enough of both. It narrows the gap in platform reporting rather than restoring individual records, and it will not appear at all below the volume thresholds.
Can I set defaults to granted to avoid the loss?
Defaults describe the state before a visitor chooses, and in regions with prior consent requirements they must be denied. Setting them otherwise creates a compliance exposure that is far more expensive than the measurement it recovers. Set defaults by region and take legal advice for each market.
Should the banner appear for visitors outside regulated regions?
That is a policy decision rather than a technical one, and many stores choose a single global experience for simplicity. If you do apply different behaviour by region, verify that the region list is correct, because a misconfigured list quietly applies the strictest settings everywhere.
Services that fix it for good
Related problems, answers and terms
Stop the leak.
A free Growth Analysis finds what is broken and ranks it by the revenue it costs you.