Enhanced Conversions: The ROAS Lift Nobody Configures Right

Enhanced conversions can lift reported ROAS without improving a single lead, opportunity or sale. The lift comes from matching conversions Google Ads previously failed to connect to an ad interaction, not from creating demand. In B2B, enhanced conversions ROAS recovers match, not truth—and recovery is concentrated in exactly the segment already known to the business: people who submitted stable first-party identifiers.

Why enhanced conversions ROAS rises without better marketing

A conversion can be real even when Google Ads cannot associate it with the ad interaction that preceded it.

Standard conversion tagging may record the event but lose the join because the buyer changed devices, switched browsers or was no longer carrying the expected advertising identifier. Enhanced conversions provide Google with another matching route: customer data supplied during the conversion.

That restriction matters. The feature cannot recover anonymous researchers, unseen buying-committee members, word-of-mouth influence or people who withheld consent. It works on identifiable converters whose details have already reached the business.

The resulting ROAS lift is therefore concentrated rather than comprehensive. Google may attribute more of the known conversion set, increasing its reported numerator while spend and commercial outcomes remain unchanged.

Better observations may eventually give Smart Bidding a more complete signal. That is a possible downstream benefit, not evidence that implementation itself created demand. The distinction between platform reporting and commercial performance is covered in our guide to separating Google Ads ROAS from real business return.

Enhanced conversions repair attribution gaps without creating commercial truth.

The five-stage enhanced conversions mechanic

The useful unit is not a better cookie. It is a normalised customer identifier attached to an existing conversion event.

The end-to-end process has five stages:

1. Collect the permitted fields. At the conversion event, the implementation reads eligible first-party data supplied by the customer, usually an email address, telephone number or permitted name-and-address fields. Spend, campaign names and conversion values are not customer-data hashes.
2. Normalise each value. Email addresses are trimmed and lowercased, telephone numbers use the required international format, and names or addresses follow Google Ads formatting rules. Normalisation must happen before hashing because "[email protected] " and "[email protected]" otherwise produce different outputs.
3. Hash the customer data once. SHA-256 converts each normalised value into a deterministic one-way digest. With a Google tag or Google Tag Manager implementation, the tag can hash raw values in the browser before transmission. With a server-side or API implementation, the business hashes the values before sending the request.
4. Transmit the hash with the conversion. Google receives the hashed identifiers alongside the conversion event and its normal attribution metadata. Raw customer data should not be placed in page URLs, analytics parameters or application logs. Pre-hashed data must not be hashed again.
5. Match against Google-held identifiers. Google compares the submitted hashes with equivalently hashed data associated with authenticated Google accounts. A valid match can reconnect the conversion to an eligible Google Ads interaction without revealing the person’s identity to the advertiser.

No match means no recovered attribution. A hash cannot compensate for an incorrect email, malformed telephone number, missing consent signal or conversion fired twice.

Correct normalisation and single hashing determine whether identifiers can match.

Measuring enhanced conversions ROAS without inventing growth

The useful denominator is the existing match gap, not total website traffic.

Take an explicitly illustrative 30-day audit set with 100 unique, confirmed eligible conversions. Standard tagging attributes 60 conversions, while the correctly configured enhanced setup attributes 68. The comparison is 60 before versus 68 after.

Previously unmatched conversions = 100 − 60 = 40
Newly recovered conversions = 68 − 60 = 8
Recovery uplift as a share of previously unmatched conversions = 8 ÷ 40 × 100 = 20%
Reported conversion-count uplift = 8 ÷ 60 × 100 = 13.3%

The platform has recovered 20% of the previously unmatched set. Calling that 20% more conversions would be false: the business received 100 conversions in both views.

Compare 30 complete days before implementation versus 30 complete days after it. Keep the conversion action, attribution window, attribution model, deduplication method and consent behaviour fixed. Apply the same conversion-lag allowance to both periods.

The honest limit here is that before-versus-after measurement is confounded by traffic mix, seasonality, consent rates and conversion lag. Google Ads diagnostics can support the interpretation, but they cannot remove those confounds.

Our claim that a measured uplift represents match recovery is falsifiable: event-level reconciliation showing duplicate fires rather than newly matched identifiers would prove it wrong.

Now take an explicitly illustrative UK consultancy spending £8,000 per month. The five labelled inputs are:

• Monthly Google Ads spend: £8,000
• Commercially attributable revenue: £40,000, held constant
• Google Ads matched conversion value before enhanced conversions: £24,000
• Newly matched conversion value after implementation: £4,000
• Signed contracts: 2 before versus 2 after

Google-reported ROAS before = £24,000 ÷ £8,000 = 3.00x
Google-reported ROAS after = (£24,000 + £4,000) ÷ £8,000 = 3.50x
Reported ROAS lift = (3.50 − 3.00) ÷ 3.00 × 100 = 16.7%

Business ROAS before versus after
= (£40,000 ÷ £8,000) versus (£40,000 ÷ £8,000)
= 5.00x versus 5.00x

Google’s reported ROAS improves from 3.00x to 3.50x. Commercial return remains 5.00x versus 5.00x because neither revenue nor contract count changed.

Reported ROAS can rise while commercial return stays perfectly flat.

Four decision rules for whether implementation is worth it

Engineering time has an opportunity cost. These four gates prevent a technically interesting project from outranking a commercially useful one.

1. Volume floor: Fewer than 20 eligible, consented primary conversions containing usable customer data in a rolling 30-day period → defer enhanced conversions. This is our prioritisation floor, not a Google Ads eligibility rule.
2. Baseline integrity: More than 2 mismatches in a 50-event sample between unique confirmed conversions and primary tag fires → repair the underlying B2B conversion tracking first.
3. Identifier availability: Fewer than 70% of eligible conversions expose at least one correctly formatted, permissioned identifier → correct the approved data mapping before adding hashing.
4. Materiality: After at least 60 eligible post-launch conversions, an attributed-count increase below 5% → leave the feature running only when maintenance is trivial and do not use it to justify higher spend. At 5% or more, report attribution recovery while holding commercial targets unchanged.

Businesses clearing all four gates can place the work within ongoing paid-search management or commission a scoped Google Ads implementation when the requirement is self-contained.

Low volume makes enhanced conversions a distraction, not a priority.

If the path from ad click to signed revenue still has gaps, Actualyse will trace each hand-off with you — book a call

UK consent and data handling come before hashing

A SHA-256 digest can still be personal data under UK GDPR. Hashing reduces direct readability; it does not make a work email address anonymous or erase the purpose for which it was collected.

Implementation should not ship until four legal conditions are evidenced:

1. A lawful purpose and basis are documented. The business must establish why customer data is being processed for advertising measurement and record the applicable UK GDPR lawful basis. Hashing after unlawful collection does not cure the original processing.
2. PECR requirements are satisfied. Where the implementation stores or accesses information on a user’s device, valid consent must be obtained before the relevant activity begins unless a genuine exemption applies. Consent Mode communicates consent state to Google tags; it does not obtain consent or create a lawful basis.
3. The processing is transparent. The privacy notice must explain that customer-provided data may be sent securely to Google for measurement and matching. The organisation must also accept the applicable Google Ads customer-data terms and document the parties’ data roles.
4. Data is minimised and protected. Send only supported fields needed for matching, restrict access, define retention, prevent raw identifiers entering logs and assess international transfers. Complete a data protection impact assessment when the risk assessment requires one.

Legal review must reflect the actual configuration rather than a generic tag-management policy. Server-side transmission, browser hashing and automatic collection expose different operational risks even when their matching objective is identical.

Lawful collection must precede technical transmission.

What does not improve enhanced conversions ROAS

The broken versions arriving in audits are often busy rather than sophisticated. Six common interventions fail for specific reasons.

1. Hashing an existing hash: A second SHA-256 pass produces a different digest, so Google cannot match it against the once-hashed identifier it holds.
2. Firing hard-coded and Google Tag Manager events together: Two implementations can produce duplicate conversions. A larger count caused by duplicate events is inflation, not recovered matching.
3. Sending every available customer field: Stale telephone numbers and malformed addresses do not become useful through volume. Extra fields increase compliance exposure without repairing bad inputs.
4. Changing the attribution window or model during rollout: The before-versus-after comparison then mixes matching recovery with a different crediting system. The enhanced-conversion effect becomes unidentifiable.
5. Changing conversion values at the same time: Adjusting how Google weights different conversions changes reported value, not identifier match quality. Run that decision separately.
6. Treating an external ROAS benchmark as validation: Commercial efficiency context for different industries can inform targets, but it cannot show whether a hash matched correctly in one account.

More data cannot rescue a broken identifier.

FAQ

Four implementation questions expose the remaining ambiguity.

Do enhanced conversions affect Smart Bidding?

Recovered primary conversions can enter bidding when they are included in the account’s conversion goals. Hold budgets and targets fixed for at least one full conversion-lag cycle, with a 14-day minimum. If attributed primary volume changes by 10% or more, re-baseline the target rather than declaring an algorithmic improvement.

Should we use automatic or manual collection?

Choose manual data-layer or server mapping when the site has more than one conversion template, dynamic fields or a single-page application. Reserve automatic collection for a stable, predictable implementation and require 10 out of 10 test conversions to pass identifier and event checks.

Which diagnostic finding should block release?

One raw email address or telephone number in an outbound URL, analytics parameter or application log is enough to stop deployment. One duplicate conversion event within 20 controlled tests also blocks release until its second source is removed.

Should enhanced conversions be enabled for micro-conversions?

If an action can occur more than once per visit without representing a commercial outcome, mark it as Secondary and exclude it from bidding. Reserve enhanced matching for genuine primary actions containing permissioned customer data.

Configuration decisions should follow evidence, not the feature’s availability.

Summary

Use these five operating rules:

• Fewer than 20 eligible conversions in 30 days → defer implementation.
• More than 2 mismatches in 50 events → repair standard tracking first.
• Identifier availability below 70% → correct approved data mapping before hashing.
• Compare 30 complete days before versus 30 complete days after under fixed settings.
• Flat revenue with higher reported ROAS → report match recovery, not commercial growth.

Match recovery deserves disciplined measurement, not a victory lap.

Actualyse builds measurement and attribution setups that tie B2B ad spend to real revenue. Book a call to talk through where yours stands.