ROAS
Return on Ad Spend
ROAS depends on the attribution model. Last-click ROAS = 4 and first-click ROAS = 2 from the same data. Always ask which attribution the ROAS is calculated under.
What is ROAS →Marketing concepts
Marketing attribution: who gets credit for the sale
Marketing attribution is the rule for deciding which ad (or channel) deserves credit for a conversion. A customer typically sees 5–20 touchpoints before buying — attribution answers which of those touchpoints "made" the sale.
The core question
Which channel gets credit when a customer saw a banner, searched the brand on Google, clicked an email, and only then bought?
Example
Customer journey: Instagram ad → a week later Google brand search → bought for $100. Last-click credits Google search. First-click credits Instagram. Linear model splits it: $50 each.
Heads up: no attribution model shows truth. Each is just a rule for how revenue gets divided in a report.
Why attribution is modern marketing's biggest headache
In the nineties it was simple. You ran a newspaper ad, the phone rang, you sold. The ad got the credit, done. Today a customer averages 5–20 touchpoints before buying — an Instagram ad, a podcast, a blog banner, a friend's recommendation, a brand search, a retargeting ad, an email. The question "what made the sale?" suddenly has a dozen answers.
Ad platforms all want to claim credit. Google Ads shows you ROAS 4. Meta Ads shows you ROAS 3. You add them up: ROAS 7? No. Part of the revenue is being claimed by both platforms at once. That's "double counting" — and it's one of the main reasons companies overspend on ads.
Marketing attribution is an attempt to bring order to this chaos. You pick one rule ("credit the first touch," "credit the last touch," "split evenly") and apply it consistently. No rule is "correct" — they're all just simplifications of reality.
The main attribution models
You'll encounter six or seven models in practice. Here are the most common, what they do, and when they make sense:
Principle
Attribution = the rule that decides who among many touchpoints deserves credit for the conversion.
Six models you should know
01
Last-click
All credit goes to the last touchpoint before conversion. Default in Google Ads, Meta Ads, and GA4.
Pro: simple. Con: ignores upper funnel.
Inflates brand campaigns and retargeting. If you decide by this alone, you'll slowly stop building new audiences.
02
First-click
All credit goes to the very first touchpoint. Useful for B2B with long sales cycles.
Pro: rewards brand discovery. Con: ignores the closing.
Good for understanding which channel brings in new audiences. Bad for optimizing conversion campaigns.
03
Linear
Credit splits evenly across all touchpoints. Five touchpoints = 20% each.
Pro: fair. Con: doesn't reflect that some touchpoints matter more.
A solid compromise for businesses unsure what to pick. Captures the whole journey but doesn't differentiate.
04
Position-based (40-20-40)
First touchpoint gets 40%, last gets 40%, middle touchpoints share 20%.
Pro: rewards both discovery and closing. Con: arbitrary numbers.
Often the most realistic model for ecommerce. But the 40/20/40 split is a convention from attribution literature, not a result of measurement.
05
Time-decay
The closer to conversion, the more credit. A touch an hour ago gets more than one a week ago.
Pro: matches how decisions actually work. Con: still inflates conversion campaigns.
Good for impulse purchases and short decision cycles. Less suited for B2B.
06
Data-driven (Google's ML model)
Algorithm learns from your data who actually drives conversions. Requires at least 300 conversions per month.
Pro: most accurate platform-side model. Con: black box, you can't verify it.
If you have enough data, it's the best choice in Google Ads. Below 300 conversions a month, Google keeps the default last-click.
Which model fits which business
There's no universal "best" model. It depends on your business, decision length, and data quality:
| Business type | Recommended model | Why |
|---|---|---|
| Ecommerce, impulse purchases (under $50) | Last-click or Time-decay | Short decision, last touch often really did decide |
| Ecommerce, expensive items ($500+) | Position-based or Data-driven | Long decision, deserves more touchpoints in attribution |
| SaaS with annual billing | First-click + Position-based | First-click shows acquisition channels, Position shows balance |
| B2B contracts (6+ months) | Linear or Data-driven CRM | Customer journey is long, last-click almost always lies |
| Established brand, brand-heavy | MMM (Marketing Mix Modeling) | Touchpoint attribution isn't enough — you need a statistical model |
| Startup with sparse data (<100 conv/mo) | Last-click + common sense | No model has enough data — attribution is secondary, focus on MER (whole-marketing return) |
Rule of thumb: if you can't explain why you picked the model, use last-click and watch MER on top. Worry about attribution once you have 300+ conversions a month.
When attribution inevitably lies
Even the best model has blind spots. Four situations where attribution systematically misleads:
iOS 14+ and the death of cookies
Since 2021, Apple has blocked much of cross-app tracking. Meta Ads today typically sees only 50–70% of actual conversions. Attribution in Meta Ads therefore systematically undervalues performance — campaigns look worse than they really are.
Cross-device and cross-browser
A customer sees an ad on their phone, buys on desktop. Without a logged-in account, that counts as two different people. Attribution doesn't know the buyer is the same person.
Dark social (WhatsApp, Messenger, link shares)
A customer sees your Instagram post, sends it to a friend via WhatsApp, the friend clicks — where in that journey is the "channel"? Analytics records it as direct traffic. The real cause (the Instagram post) disappears from the report.
Brand halo effect
You saw a billboard on your commute. A week later you remember and search the brand on Google. Attribution credits Google, not the billboard. For TV, podcasts, and influencer campaigns this effect is huge — and classic attribution practically can't see it.
Related concepts and when to use them
Attribution isn't enough on its own. Metrics and methods that complement it:
MER
Marketing Efficiency Ratio
MER sidesteps attribution entirely — total revenue / total marketing spend. When attribution is unreliable, MER is your safety net.
What is MER →MMM
Marketing Mix Modeling — statistical attribution
Top-down approach. A statistical model estimates each channel's contribution based on spend and revenue over time. For companies with annual marketing budgets around $1M+.
Incrementality
Incrementality — measuring causal lift
Measured via A/B tests: pause a channel for 4 weeks in selected regions, watch the difference. Most accurate method, but expensive and slow.
Three mistakes most companies make
- 01
Treating last-click as truth
Last-click is the default in Google Ads and Meta Ads, so companies get used to it. But it's the worst possible model for understanding the actual customer journey. Always pick the model consciously — don't inherit the default blindly.
- 02
Comparing channels under different models
Meta Ads UI shows 7-day view-through attribution by default. Google Ads defaults to last click. If you compare ROAS straight from two platforms, you're comparing apples and oranges. Always normalize to the same model (typically last-click) or look at GA4.
- 03
Thinking attribution = truth
Attribution is a rule, not reality. The actual truth can only be told by incrementality testing — and even that has a large margin of error. If you want precision, pay for an experiment. If you want orientation, attribution is enough.
Frequently asked questions
- Which attribution model is best?
- None universally. If you have at least 300 conversions a month, Google data-driven or Meta data-driven are the best choices in platform. Below 300 conversions, use position-based for ecommerce, first-click for B2B. And always watch MER as a backstop.
- Why do Google Ads and Meta Ads report higher revenue than my accounting?
- It's called double counting — both platforms claim the same conversion if both contributed. Summing ROAS across platforms makes no sense. Either use GA4 as a single source of truth, or watch MER and don't worry about platform-specific attribution.
- What's the difference between attribution and MMM?
- Attribution counts credit per touchpoint: each conversion is split across specific channels based on a rule. MMM (Marketing Mix Modeling) is the opposite — top-down, statistically modeling how much of monthly revenue each channel caused based on spend and seasonality. MMM needs no cookies or tracking.
- How did iOS 14 affect attribution?
- Drastically. App Tracking Transparency in 2021 removed Meta Ads' access to most iOS conversions. Click-through windows shrank from 28 to 7 days; view-through to 1 day. Meta Ads today typically sees 50–70% of actual conversions. The impact on Google Ads is smaller (they have their own logged-in users).
- If attribution lies, how do you actually make decisions?
- Combine three things: 1) MER as an overall sanity check, 2) incrementality tests (A/B pause of a channel for 4 weeks) for your most expensive channels, 3) attribution as orientation, not truth. Most good marketers steer by trend feel, not absolute numbers.
Want to see attribution across all your platforms?
Lupli connects Google Ads, Meta Ads, and GA4 into one view. Instead of switching between attribution models in each platform, you ask directly: "how much does a new customer actually cost?"
- Attribution unified across Google Ads, Meta Ads, and GA4
- MER as a safety net when attribution misleads
- 30 seconds to connect your first account