Marketing attribution is a model for assigning credit. It is not a perfect reconstruction of why a person bought. That difference matters because the same sale can look different in Meta, Google Ads, GA4, a CRM and the finance system — without any of those systems necessarily being “broken.”
ROAS remains useful. The mistake is treating one platform's attributed revenue as the complete truth about the customer journey.
Attribution assigns credit; it does not recreate causality
GA4 defines attribution as assigning credit to touchpoints on the path to a key event. Its reporting attribution options currently include data-driven attribution and last-click variants.
Those models answer a credit-allocation question. They do not prove which touchpoint would have caused the sale if the others had not existed.
A customer may see a social ad, search the brand, visit directly, read reviews, return through email and then purchase. A model has to decide how to distribute credit across that sequence. Different models can produce different answers while the underlying sale stays exactly the same.
Platform ROAS answers a platform question
Meta and Google Ads report conversions using their own attribution systems and available signals. That data is important for optimizing activity inside each platform.
But platform reporting can differ because of attribution windows, view-through credit, identity resolution, consent, cross-device behavior, modeled events and the platform's own role in the path. Treat platform ROAS as a powerful optimization lens — not audited company revenue.
GA4 is also a model, not an oracle
GA4's data-driven attribution uses property data to distribute credit across click interactions for each key event. Google also uses modeled key events when some conversions cannot be directly observed because of privacy, technical limitations or cross-device journeys.
Google notes that attributed conversion data in Analytics can continue updating for up to 12 days after a conversion is recorded because of processing and model training. That is a reminder to avoid treating a same-day dashboard as final truth.
Blended performance answers a different question
Blended ROAS or MER-style reporting looks at the business more broadly by comparing total relevant revenue with total media or marketing spend.
Its strength is economic: two platforms cannot both create two separate units of actual company revenue from the same order. Its weakness is diagnostic: a blended number does not tell the media buyer which campaign, keyword or creative to change tomorrow.
That is why platform and blended views should coexist.
Use a measurement hierarchy
- Business layer — revenue, gross margin, contribution, pipeline value, new customers and cash efficiency.
- Acquisition layer — CAC, cost per qualified opportunity, cost per booked outcome, blended ROAS or MER.
- Channel layer — platform ROAS, cost per conversion, conversion value, impression share, frequency and campaign-level performance.
- Diagnostic layer — CTR, CPC, CPM, landing-page rate, form completion, add-to-cart and other operational signals.
The hierarchy prevents a diagnostic metric from becoming a business outcome.
Lead generation needs deeper conversion feedback
Attribution is harder when the valuable event happens after the form. A dental inquiry may convert over the phone. A B2B lead may become a qualified opportunity weeks later. A high-ticket project may pass through consultation, proposal and contract.
Google's current enhanced-conversions-for-leads system is specifically designed to connect online lead activity with later offline outcomes using hashed first-party data and CRM imports. In 2026, Google also consolidated enhanced conversion settings and moved offline uploads toward Data Manager.
The technical implementation matters, but the strategic point is simpler: the ad platform should learn from deeper commercial outcomes when the business can provide them responsibly.
A real pipeline shows why this matters
Nestra's acquisition system did not stop reporting at the lead. Across 18 weeks, the account generated 730+ qualified leads, 186 design consultations, 74 proposals and 19 signed projects. That created four very different acquisition costs: approximately PKR 4.2K per qualified lead, PKR 16.7K per booked consultation, PKR 41.9K per proposal and PKR 163K per signed project.
If management only looked at cost per lead, it would miss the entire commercial shape of the funnel. The question is not which number is “the real one.” The question is which number is appropriate for the decision being made.
Vela shows why blended economics matter
Vela Skin was evaluated using revenue, blended ROAS, CAC, merchandising behavior and conversion rather than ad-platform performance in isolation.
Over the 16-week engagement, blended ROAS moved from 2.4× to 3.4×, customer acquisition cost fell 24% versus baseline and final-month ecommerce revenue increased 41% versus the pre-engagement baseline.
The case study explicitly notes that ROAS is not net profitability. That is an important measurement discipline: revenue efficiency and profit are related, but they are not the same metric.
Attribution cannot answer incrementality by itself
A channel can receive attribution credit for customers who may have converted anyway. That is an incrementality question, not an attribution-model question.
Where the budget decision is material, use holdouts, geo tests, lift studies or other experimental designs where feasible. You do not need an incrementality experiment for every small campaign. You do need to understand when attribution alone is insufficient for the decision.
Fix data hygiene before building a sophisticated model
- UTMs are inconsistent.
- Forms lose source information.
- Phone calls are not categorized.
- CRM stages mean different things to different people.
- Refunds or cancellations are ignored.
- Campaign naming is chaotic.
- Revenue is not reconciled with finance.
In that environment, a more complicated model creates precise-looking uncertainty.
Use attribution to answer specific management questions
- Which channels are producing the highest-quality demand?
- Are we becoming more or less efficient as total spend increases?
- Does one platform look exceptional only under its own attribution view?
- Are branded conversions being influenced by another channel?
- Are returning customers inflating acquisition efficiency?
- Are offline outcomes being fed back into the bidding system?
- Did total company revenue or qualified pipeline rise when marketing investment rose?
That is what turns attribution from reporting theatre into a decision layer.
Our operating view
Use platform data to optimize the platform. Use GA4 to understand and compare paths. Use CRM data to judge lead quality and downstream conversion. Use blended economics to decide whether the acquisition system deserves more capital. That is the measurement half of a performance marketing strategy.
ROAS is useful. The error is asking one attribution model to explain an entire business.
