Google Ads vs Meta Ads ROAS: A Fair Comparison Method

A reported 6.2× Google Ads ROAS can lose to a reported 4.7× Meta Ads ROAS once both answer the same attribution question.

Ignoring that mismatch does more than produce a messy report. It moves budget towards whichever platform claims credit most generously.

Google Ads vs Meta Ads ROAS as reported compares two different questions: normalise the window and attribution model first, or the comparison decides your budget for you. Treating the interface figures as direct rivals is a category error.

Why reported Google Ads vs Meta Ads ROAS cannot choose a winner

Google Ads commonly starts with a 30-day click-through window. Meta commonly uses seven-day click and one-day view attribution. Both settings can be changed, so the live account configuration remains decisive.

The default comparison still gives Google longer to claim a click-led conversion. Meta can claim revenue after an impression that produced no click. Window length flatters Google; view-through eligibility can flatter Meta.

The model creates another mismatch. Google Ads may distribute credit across Google interactions using data-driven attribution, while Meta applies its own attribution logic to interactions it can observe. One £20,000 sale can consequently appear in both platforms.

Neither platform report sees the complete journey. It sees evidence available inside its own identity, consent and tracking boundaries. Our separate guide to challenging reported Google Ads returns covers verification; the task here is making two platform definitions comparable.

Reported ROAS measures platform credit, not comparative economic value.

The seven adjustments required before comparing platforms

Freeze platform-level budget recommendations until one dataset applies seven identical adjustments.

1. Use one outcome. Choose net closed-won revenue for both platforms. Do not compare Google revenue with Meta lead values or predicted conversion values.
2. Use one valuation basis. Exclude VAT, refunds and cancelled contracts from both numerators. Convert revenue into the same currency using one declared exchange-rate date.
3. Use one eligible interaction. Click-only attribution is the cleanest shared basis for Google Search versus Meta because a comparable Search impression does not exist. Remove Meta view-through credit for the primary comparison.
4. Use one lookback window. Apply the same 30-day or 60-day window to every eligible paid touch. Select it from the business’s conversion-lag distribution, not the more flattering interface default.
5. Use one attribution model. Reassign revenue outside the platform interfaces using a declared rule. Last eligible paid touch is blunt but reproducible; alternative B2B revenue attribution models should be compared separately rather than blended.
6. Use one cohort basis. Group results by interaction month and wait for the cohort to mature. Mixing click-date reporting with conversion-date reporting moves delayed sales between periods.
7. Use one spend denominator. Include media spend after platform credits and refunds on both sides. Keep management fees outside both figures unless the comparison explicitly uses fully loaded cost.

Store the resulting assumptions beside the output. The implementation belongs in a ROAS dashboard with declared source rules, not in another undocumented platform export.

Comparable inputs come before comparable ratios.

Google Ads vs Meta Ads ROAS, reported and normalised

The apparent winner can reverse without changing delivery, creative or sales outcomes.

Take a fully illustrative UK cyber-security consultancy comparing one Google Search campaign with one Meta campaign. Seven labelled inputs drive the calculation.

As reported, the arithmetic is:

Google reported ROAS = £49,600 ÷ £8,000 = 6.2×

Meta reported ROAS = £37,600 ÷ £8,000 = 4.7×

Google therefore appears to win at 6.2× versus 4.7×.

After normalisation, the same campaign pair becomes:

Google normalised revenue = £12,000 + £12,800 = £24,800

Google normalised ROAS = £24,800 ÷ £8,000 = 3.1×

Meta normalised revenue = £14,000 + £16,400 = £30,400

Meta normalised ROAS = £30,400 ÷ £8,000 = 3.8×

Our falsifiable claim is that an unnormalised winner cannot reliably predict the better next pound; three consecutive mature cohorts where it does would prove that claim wrong.

The honest limit of this illustration is that another declared model could allocate the shared revenue differently. Normalisation creates a fair comparison, not an unquestionable truth.

Normalisation can reverse the winner without changing campaign delivery.

Five decision rules for moving budget

A 0.1× advantage is too small to buy confidence. Apply these five operating rules instead.

1. Mismatch rule: if the attribution windows differ by even one day, or one numerator contains views while the other contains clicks only, freeze reallocation until both are rerun.
2. Google window rule: for a 30–60-day B2B consideration cycle, Google’s common 30-day click default flatters Google against Meta’s seven-day click window. Estimate the advantage as Google-attributed value from days 8–30 divided by total Google-attributed value. On an illustrative £50,000 numerator containing £12,000 from days 8–30:

£12,000 ÷ £50,000 = 24%

Google is flattered by roughly 24% of its reported numerator. If the share exceeds 15%, compare both platforms using 30-day click-only attribution.

3. Meta view rule: if view-through revenue exceeds 10% of Meta’s reported numerator, remove all view-through value before comparing it with Google Search.
4. Materiality rule: if the normalised ROAS gap is below 15%, hold both budgets. If fewer than 90% of historically expected sales have matured, postpone the decision regardless of the gap.
5. Reallocation rule: if one platform leads by at least 25% across two mature monthly cohorts, with at least 10 closed-won sales per platform, move 10% of the next budget towards the winner.

These rules also set the measurement boundary for our specialist Google Ads management: bidding changes follow comparable economics, not interface confidence.

Budget follows mature, material and consistently normalised differences.

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

What did not work: four shortcuts that leave the comparison broken

Putting two platform exports in adjacent columns creates symmetry without comparability. Four common shortcuts fail.

1. Selecting 30 days in both interfaces aligns the window label but not eligible interactions, attribution logic, revenue definition or identity coverage.
2. Making GA4 the automatic referee replaces two platform models with GA4’s selected model. Consent loss, cross-device gaps and channel classification remain, while consistency still does not establish causation.
3. Replacing ROAS with cost per lead hides the revenue problem rather than solving it. A £90 lead versus a £140 lead says nothing decisive when their close rates and contract values differ. Our guide to B2B cost-per-lead economics treats that metric as a separate decision.
4. Giving both campaigns the same target ROAS changes bidding behaviour, not measurement. Each algorithm still optimises towards its own conversion values and attribution evidence.

Interface tidiness cannot repair incompatible attribution.

What normalisation still does not settle: incrementality

A paid touch can receive credit for a sale that would have happened without the advert. Neither the reported number nor the normalised number proves incrementality.

Three confounds remain.

1. Demand capture versus demand creation: Google may collect an existing brand or category search, while an earlier exposure shaped the search. Attribution assigns the sale; it cannot reconstruct the no-ad outcome.
2. Cross-channel interaction: a buyer can encounter both campaigns, return directly and speak to sales later. A common model prevents duplicate credit but cannot prove which exposure changed behaviour.
3. Website and offer readiness: channel comparisons remain confounded by landing-page relevance, audience routing and sales follow-up. Our selected web design case studies document Lanteria’s stakeholder routes and Lake Erie Shores’ separate stay and ownership journeys; those are architecture decisions, not performance claims.

What this cannot tell you is the revenue that would exist with either platform switched off. If the normalised gap is below 15%, or more than 20% of won deals contain touches from both channels, cap any reallocation at 10% and use a controlled holdout before moving further.

Attribution assigns credit; experiments estimate causation.

FAQ

Should agency fees be included in the comparison?

Use media-only ROAS for the platform comparison, with fees excluded from both sides. If management, creative or technology fees differ by more than 10% of media spend between platforms, add a second fully loaded contribution-return figure rather than contaminating ROAS.

How many sales are enough for a platform decision?

Fewer than 10 closed-won outcomes per platform in a mature cohort is too thin for a direct ranking. Pool up to three comparable monthly cohorts; if either platform still remains below 10 outcomes, keep budgets stable and treat the result as directional.

Should branded Google searches be removed?

Split brand and non-brand results whenever branded terms consume more than 20% of Google spend. Do not delete brand revenue automatically; make removal contingent on a brand holdout showing that the same demand arrives without paid coverage.

Can qualified pipeline replace closed-won revenue?

Use qualified pipeline only when the sales cycle prevents a mature revenue cohort. Lock one stage probability for a quarter, and abandon pipeline ROAS if forecast revenue differs from realised revenue by more than 15% across two consecutive quarters.

Low-volume decisions need stricter rules, not louder platform reports.

Summary

Use these five rules.

• Freeze budget changes whenever attribution windows or eligible interaction types differ.
• If Google’s day-8-to-30 value exceeds 15%, compare both platforms at 30-day click-only.
• Remove Meta view-through credit when it exceeds 10% of reported revenue.
• Hold budgets when normalised ROAS differs by less than 15%.
• Shift 10% only after a 25% lead persists across two mature cohorts.

Comparable inputs must precede every budget verdict.

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