Evidence status: Agency-published case study; reported measurement and CPA effects, not a causal business-outcome experiment
This case is about seeing more, not necessarily creating more
Virgin Media O2 operates in a market where paid search performance depends on accurately connecting ad interactions with broadband and telecom conversions.
Privacy changes make that harder.
Consent choices, browser restrictions and fragmented first-party signals can reduce the share of customer journeys that appear directly in advertising reports.
The risk is operational: bidding systems learn from what they can observe.
If high-value conversions disappear from measurement, the platform can make poorer optimization decisions even when the underlying customer behavior has not changed.
Merkle’s public case study says Virgin Media O2 implemented Google Consent Mode and Enhanced Conversions as part of a modernized measurement framework.
The reported outcomes were:
- 9.5% increase in conversions reported via Consent Mode
- 42.1% increase in conversions reported via Enhanced Conversions
- 6.6% improvement in CPA across paid search campaigns
Those figures sound like performance growth.
The first analytical task is to separate measurement recovery from incremental demand creation.
What Consent Mode changes
Consent Mode changes how Google tags behave based on a user’s consent state.
Google describes the system as a way to respect consent choices while using conversion modeling to estimate some conversions that cannot be directly linked to advertising interactions.
That means an increase in conversions “reported via Consent Mode” does not automatically mean more customers converted after implementation.
Part of the increase can come from previously unobservable or modeled activity becoming measurable inside the reporting system.
That is still valuable.
A bidding system can only optimize from the data available to it.
But the metric should be interpreted correctly.
Measurement recovery is not the same claim as sales lift.
What Enhanced Conversions changes
Enhanced Conversions uses consented first-party customer data, transformed before being sent, to improve the platform’s ability to match conversions to ad interactions.
In practical terms, it can help reconnect a conversion with an advertising journey when ordinary browser identifiers are incomplete.
Merkle reports a 42.1% increase in conversions reported via Enhanced Conversions for Virgin Media O2.
Again, the wording matters.
The result describes additional conversions observed or matched in the measurement system.
It does not establish that Enhanced Conversions caused 42.1% more people to buy broadband.
That would require a different experimental design.
Why better measurement can still improve CPA
The 6.6% CPA improvement is the most commercially interesting number in the Merkle case.
Unlike a reporting-recovery percentage, CPA combines advertising cost with conversion count.
If the platform observes more valid conversion signals and uses them in automated bidding, it can potentially allocate spend more efficiently.
There are at least three mechanisms through which that can happen:
Better attribution feedback. More conversions are linked to the campaigns and queries that influenced them.
Stronger bidding signal. Smart Bidding receives a denser set of outcomes from which to learn.
Reduced false negatives. Campaigns that were creating valuable conversions but appeared weaker can receive more appropriate investment.
The case is therefore plausible operationally.
The public evidence still does not isolate exactly how much of the CPA change came from measurement, bidding adaptation, market conditions or other campaign work during the period.
That limitation should remain visible.
The case illustrates a signal-density problem
Imagine two Search campaigns.
Campaign A generates 1,000 true conversions and the measurement system observes 900.
Campaign B also generates 1,000 true conversions but the measurement system observes only 650 because more users decline tracking or convert through paths that are harder to match.
A bidding algorithm trained on reported outcomes may treat Campaign B as weaker even when true business performance is equivalent.
The problem is not merely reporting aesthetics.
It changes capital allocation.
Privacy-aware measurement tools are valuable when they reduce that distortion without ignoring consent.
That is the core operational lesson from Virgin Media O2.
Why telecom makes the problem harder
Telecom purchases can involve:
- multiple devices;
- research over several days;
- store and online touchpoints;
- address eligibility checks;
- call-center interactions;
- contract decisions;
- household rather than individual behavior.
A single browser session is an incomplete representation of that journey.
This makes first-party data especially important.
The closer the measurement stack can connect paid search with validated customer outcomes, the less dependent it becomes on a single cookie or session.
The same logic applies to banking, insurance, travel, B2B and other categories where the final economic outcome occurs away from the initial click.
Failure mode: celebrating recovered conversions as growth
This is the most important failure mode in the case.
Suppose an implementation increases reported conversions by 30%.
If the business reports the project internally as “30% more conversions,” marketing leadership may assume customer demand increased.
That is a category error.
The implementation may have:
- recovered previously missing conversions;
- improved matching;
- increased modeled conversions;
- reduced duplicate loss;
- changed attribution.
Those are measurement improvements.
They become business improvements only when they change decisions in a way that produces more valuable outcomes.
A strong post-implementation review should therefore reconcile platform changes with:
- total orders or contracts;
- qualified customer volume;
- revenue;
- finance data;
- channel spend;
- downstream retention if relevant.
Failure mode: assuming more measured data is always better
Measurement quality is not measured by the maximum possible event count.
A system can increase reported conversions by introducing duplicates or low-quality proxy events.
The validation question should be:
Does the platform now represent business truth more faithfully?
For Virgin Media O2, the public case frames Consent Mode and Enhanced Conversions as privacy-conscious measurement improvements.
An external operator should still build QA around:
- transaction or lead identifiers;
- event duplication;
- consent state;
- destination acceptance;
- reporting latency;
- CRM reconciliation;
- unexplained jumps after deployment.
Signal density is useful only when signal integrity remains high.
The optimization loop should be monitored after implementation
The most interesting period may begin after the measurement project launches.
Automated bidding can change behavior as more conversion data becomes available.
Queries that previously looked weak may receive more traffic.
Campaigns may expand.
CPA targets may become easier to hit.
That makes the measurement change a model-input change, not merely an analytics change.
The team should therefore monitor:
- spend distribution before and after;
- query mix;
- conversion lag;
- bidding strategy behavior;
- campaign-level CPA;
- total business outcomes;
- modeled versus directly observed share.
This helps distinguish a healthier optimization system from a reporting-only change.
What the case does not prove
The public case does not prove:
- Consent Mode created 9.5% more underlying conversions;
- Enhanced Conversions created 42.1% more sales;
- every advertiser will improve CPA by 6.6%;
- the implementation alone caused the CPA improvement;
- modeled conversions are equivalent to independently verified incremental conversions.
The correct takeaway is narrower and more useful.
Virgin Media O2 appears to have improved the observability of paid-search outcomes and reported a subsequent efficiency improvement.
That is exactly the kind of measurement project that deserves operational attention in a privacy-constrained environment.
How to apply the lesson
A media team considering similar work should define three baselines before implementation:
Business truth baseline: How many validated conversions or customers exist in the source system?
Measurement baseline: How many appear in Google Ads and analytics, and through which mechanism?
Optimization baseline: How is Smart Bidding currently allocating spend and what performance does it achieve?
After implementation, compare all three.
That prevents the team from treating a measurement change as a growth event while still allowing it to recognize genuine improvements in optimization.
Evidence note
The headline metrics in this article are reported by Merkle, the agency that worked with Virgin Media O2. The public case does not provide a controlled experiment isolating the causal effect of the measurement implementation. Radar therefore treats the 9.5%, 42.1% and 6.6% figures as agency-reported case outcomes, with the first two primarily describing measurement visibility rather than direct incremental customer growth.
