Microsoft Advertising Advanced Consent Mode 2026: Measuring CPA Conversions with UET When Users Decline Tracking
Consent refusal no longer has to create a complete blind spot in Microsoft Advertising measurement. In 2026, Advanced Consent Mode allows Universal Event Tracking, or UET, to operate with a denied consent state before a visitor makes a choice and to remain in that restricted state if advertising consent is refused. The important distinction is that Microsoft is not restoring conventional user-level tracking after a refusal. Instead, limited anonymous signals can support conversion modelling, while advertising cookies and identifiers remain unavailable for normal attribution. For advertisers managing acquisition campaigns, this can produce a more realistic view of conversion volume and CPA than a setup in which UET is completely blocked for every non-consenting visitor. The quality of that measurement depends heavily on correct consent configuration, complete UET coverage and enough campaign data for Microsoft’s modelling systems to work with confidence.
How Advanced Consent Mode Changes UET Measurement in 2026
Microsoft Advertising Consent Mode controls how UET behaves according to a visitor’s advertising consent status. The central setting is ad_storage, which can be either granted or denied. With granted consent, UET can use the normal advertising storage needed for standard conversion attribution and remarketing. With denied consent, that storage is restricted. This distinction matters particularly for traffic from the European Economic Area, the United Kingdom and Switzerland, where Microsoft has required advertisers to send appropriate consent signals since enforcement took full effect on 5 May 2025. A cookie banner by itself is therefore not enough. The visitor’s decision must reach UET correctly so that Microsoft receives the appropriate consent state alongside measurement activity.
Basic and Advanced Consent Mode handle the period before consent differently. Under Basic Consent Mode, UET is held back until the visitor gives permission. If permission never arrives, Microsoft receives no UET activity from that visit. Advanced Consent Mode takes another approach: UET is allowed to load immediately, but its default consent state is denied. If the visitor subsequently accepts advertising cookies, the state changes to granted. If the visitor rejects them, UET remains in the restricted denied state. Microsoft recommends the advanced configuration as the stronger technical option for advertisers that want to improve consent-based modelling because it preserves limited measurement signals without treating a refusal as permission for conventional advertising tracking.
This difference is particularly important for CPA analysis. Imagine that two campaigns spend the same amount and generate a similar number of real purchases, but one has a much higher share of visitors declining advertising consent. A Basic setup can make that campaign appear weaker simply because a larger part of its customer journey becomes unobservable. Advanced Consent Mode cannot recreate every lost journey or identify each non-consenting buyer. What it can do is give Microsoft’s modelling systems more information about aggregate behaviour. For an eligible advertiser, the resulting modelled conversions may reduce the gap between directly observed results and the broader number of conversions that the campaign is statistically likely to have produced.
What Happens When a User Rejects Advertising Tracking
A denied consent signal should not be interpreted as a different form of full tracking. Microsoft states that under Advanced Consent Mode, UET can continue operating with the consent state set to denied, while advertising identifiers are not attached in the same way they would be after permission is granted. Anonymous information can be collected and aggregated before it is used for modelling. The advertiser therefore does not receive a deterministic record saying that a particular non-consenting visitor clicked an ad and later became a customer. That individual path remains unavailable. The useful output comes later at an aggregated reporting level, where modelling can estimate some conversions that could not be directly observed because of the consent restriction.
This is why the phrase “measuring conversions after tracking refusal” needs to be understood carefully. An advertiser is not bypassing the visitor’s decision. Instead, Microsoft uses observable campaign data, historical patterns and privacy-restricted signals to estimate missing conversion activity. Directly measured conversions continue to come from journeys where the necessary signals are available, while modelled conversions help fill some of the remaining gap. Microsoft reports that these modelled conversions do not use the personal data of non-consenting end users. This approach is valuable for CPA analysis because campaign performance is less dependent on the proportion of visitors who happen to accept advertising storage, although statistical estimates should never be treated as individually verified sales or leads.
The reporting effect is visible because modelled conversions can be added directly to the Conversions column for eligible advertisers. Microsoft combines observed and modelled conversions in that total, and modelled values may also contribute to the Revenue column when conversion values are being tracked. Modelled conversions can appear as fractional numbers rather than whole conversions, which is normal for statistical estimation. If campaign spend is £10,000 and the report contains 100 observed conversions plus 10 modelled conversions, the reported CPA based on 110 conversions is about £90.91 rather than £100. That does not mean ten named customers have been identified. It means the reporting model estimates additional conversion activity that was not directly observable.
How to Set Up UET So CPA Measurement Remains Useful
The most important requirement for Advanced Consent Mode is the order in which UET and the consent system operate. Microsoft’s 2026 guidance says that UET needs to load before the consent banner or consent management script prevents it from reading the initial state. The default must be denied before a visitor has granted advertising consent. When Google Tag Manager is used, Microsoft recommends the official UET template, configured across the relevant pages, with the tag-management container available early enough for UET to receive the default consent state. A consent tool should not simply block UET completely if the intention is to use Advanced Consent Mode, because a tag that never executes cannot contribute the restricted signals required for the advanced setup.
The second requirement is equally important: when a visitor accepts advertising consent, UET must receive a granted update. That state needs to remain accurate on subsequent pages where the consent applies. A common measurement failure occurs when a cookie banner correctly records acceptance but the choice is never passed to UET. In that situation, Microsoft continues to see denied consent and UET remains in restricted operation even though the visitor has agreed to advertising storage. The opposite problem is more serious from a privacy perspective: sending granted before permission has actually been obtained. Consent logic therefore needs to reflect the visitor’s real choice rather than merely whether a banner was displayed or closed.
Coverage also matters. Microsoft’s current guidance for modelled conversions recommends 100% UET coverage across landing pages and no breaks in conversion tracking. If paid traffic can enter through landing pages where UET is missing, both observed measurement and modelling quality can deteriorate. Conversion goals should also represent genuine business outcomes. A lead-generation advertiser may count a qualified form submission, while an ecommerce advertiser may count completed purchases and their values. Mixing minor interactions such as page views with primary acquisitions can make CPA look artificially low and can send weak signals to automated bidding. The measurement setup should therefore define clearly which actions represent an acquisition before consent modelling is considered.
Checking Consent Signals and Interpreting the CPA Report
Testing should cover at least three states: a new visitor who has not yet made a choice, a visitor who rejects advertising consent and a visitor who accepts it. Microsoft’s UET Tag Helper can be used to confirm that UET is present, that the initial consent state is denied and that the state changes to granted after a positive choice. The same test should confirm that advertising identifiers are not being attached as though consent had been granted when the visitor refuses. Browser network tools can provide an additional check. Testing only the “accept all” path is not enough because many faulty configurations appear to work normally for consenting users while mishandling rejection or the period before a decision.
Advertisers should then distinguish observed business activity from reported conversions that include modelling. Microsoft says modelled conversions are integrated directly into campaign reporting, which means a change in the Conversions figure does not always represent a sudden change in the number of directly recorded transactions. This becomes particularly important when Advanced Consent Mode is introduced on an established account. A rise in reported conversions and a lower CPA may partly reflect improved modelling rather than an immediate improvement in sales performance. Internal order data, CRM records or another lawful first-party source can be used as a business-level reference for checking whether overall acquisition trends remain sensible without attempting to rebuild identities for visitors who refused advertising tracking.
CPA itself should be read over a useful period rather than judged from isolated days. The basic relationship remains cost divided by conversions, but the conversion denominator can now contain both observed and modelled results. If the share of modelled activity changes significantly, short-term CPA comparisons with older periods can become misleading. A practical approach is to annotate the date when consent configuration changed, keep campaign goals stable for the comparison period and monitor spend, conversions and actual customer outcomes together. This gives the advertiser enough context to distinguish a genuine change in acquisition efficiency from a reporting change caused by better consent signals or the activation of modelled conversions.
Using Modelled Conversions for CPA Decisions Without Overstating Accuracy
Modelled conversions are not automatically available to every advertiser. Microsoft’s June 2026 documentation states that the feature is generally available to eligible advertisers in the EEA, Switzerland and Great Britain. Eligibility includes the use of UET conversion tracking, correct implementation of UET Consent Mode or a comparable consent framework, and at least 700 ad clicks per week in the targeted market. Microsoft also applies a confidence threshold, so meeting the stated requirements does not guarantee that an account will immediately receive modelled conversions. Eligible accounts begin receiving them automatically, and the effect can appear gradually in reporting rather than as a single one-off adjustment.
Once modelling contributes to reporting, automated bidding can also benefit from the additional conversion information. Microsoft states that modelled conversions improve the conversion data used by its bidding algorithms. This is relevant to advertisers working towards a CPA target because bidding systems perform poorly when a large and uneven share of real acquisitions disappears from measurement. There is also a 2026 account-management detail worth noting: since 4 August 2025, newly configured Target CPA is generally set as an optional target within Maximise conversions rather than as a new standalone Target CPA strategy. Existing campaigns already using the older standalone strategy and eligible portfolio strategies are treated separately.
Campaign managers should still avoid reacting too quickly to a newly modelled CPA. Microsoft recommends allowing conversion-based bidding enough data before judging performance and commonly advises having at least 30 conversions before evaluating Target CPA behaviour. After a consent implementation change, the safer approach is to keep budgets, conversion definitions and bidding targets reasonably stable while reporting settles. If an advertiser previously raised or lowered Target CPA to compensate for missing consent data, Microsoft recommends monitoring spend and adjusting targets gradually when modelled conversions become active. Large immediate changes can make it difficult to separate the effect of consent modelling from the effect of the bidding change itself.
Limits, Compliance Checks and Mistakes That Can Distort CPA
The most important limitation is that modelled conversions remain estimates. They can improve the completeness of reporting, but they are not a replacement for accurate consent handling, reliable conversion goals or sound business records. A fractional modelled conversion should not be presented internally as a confirmed individual sale, and modelled revenue should not automatically be treated as reconciled accounting revenue. For commercial decisions, Microsoft Advertising data should be considered alongside confirmed orders, approved leads, cancellations, refunds and other business outcomes. This is especially important for businesses with long sales cycles, delayed lead qualification or substantial order cancellation, where the advertising conversion event and the final economic result may be separated by days or weeks.
Server-side measurement should not be used as a way to ignore a refusal either. Microsoft’s Conversions API supports an explicit advertising-storage consent value. When the value indicates denied consent, Microsoft states that the event is not used for advertising purposes, including conversion attribution and retargeting. Moving an event from the browser to a server therefore does not automatically create permission to use it for advertising measurement. The correct approach is to preserve the visitor’s consent state throughout the measurement process and apply the appropriate legal basis to any CRM, offline or server-side data that is later sent to Microsoft. Consent requirements and lawful processing obligations can vary by jurisdiction, so technical configuration does not replace legal review where one is required.
For most advertisers, the practical 2026 objective is therefore straightforward: implement Advanced Consent Mode correctly, keep UET coverage complete, pass the visitor’s real choice consistently, maintain meaningful conversion goals and allow enough data for modelling to become reliable. CPA after a tracking refusal should then be understood as an aggregated campaign metric supported partly by statistical modelling, not as deterministic tracking of people who said no. This distinction protects the integrity of both privacy controls and performance analysis. When the implementation is tested properly and reporting changes are interpreted in context, Microsoft Advertising can retain substantially more useful acquisition measurement without pretending that consent restrictions have disappeared.