Perfect Attribution Doesn't Exist: What to Track Instead of ROAS

There's a scene that repeats itself in nearly every marketing meeting we sit in on. The ad manager pulls up the Meta dashboard, points to a ROAS of 4.2, and says "we're crushing it." Then the owner opens the bank statement, looks at the month's revenue, and none of those numbers show up anywhere. Two truths that don't match, in the same company, in the same month.

It's nobody's fault. It's a signal that the marketing attribution you've been leaning on for years has stopped working, and maybe never worked as well as you thought. In this article I'll explain why multi-touch attribution became an illusion after 2021, why ROAS pushes you toward bad decisions, and above all what to look at instead: MER, margin, LTV/CAC, and real incrementality measurement. No textbook theory, just the reasoning we go through when we help a company figure out whether it's actually making money from advertising.

What attribution promises (and why we believed it)

Attribution, in theory, answers a simple question: which channel, which campaign, which ad caused this sale? If you knew that for certain, you'd shift the whole budget to the winners and shut down the losers. Problem solved.

For a decade we pretended it was possible. The models got more refined: from last-click (all the credit to the final touchpoint before purchase) to multi-touch models that spread credit across the whole journey, all the way to data-driven models that promised to calculate "objective" weights for every step. If you want the full picture of the logic, pros and cons of each approach, we cover it in detail in our guide to marketing attribution models.

The problem is that all these models rest on a single assumption: knowing who that person is and following them from one touchpoint to the next. And that assumption has collapsed.

Why multi-touch attribution is a post-privacy illusion

Between 2021 and today, everything happened at once, and it dismantled the foundations of deterministic tracking. It's worth listing it out, because plenty of business owners still look at their reports as if 2019 never ended.

• Apple's App Tracking Transparency (2021). On iOS, users must give explicit consent to be tracked. Most say no. On its own, this move blinded pixels across a huge slice of premium traffic.
• Third-party cookies on their way out. The timeline has been shaky, but the direction is clear: browsers are making cross-site tracking harder, and Safari and Firefox have been blocking it for a while already.
• Consent and GDPR. In Europe, a properly built cookie banner means a portion of your visitors aren't tracked at all. That's not a bug, it's the law.
• Statistical modeling instead of real data. To fill the gaps, platforms started estimating conversions. Meta and Google today show you numbers that are largely modeled, not counted. We dig into what that really means in the piece on what conversion modeling is.

Put it all together and you get an uncomfortable sentence: the attribution report you check every morning is, in large part, a probabilistic reconstruction, not a snapshot. And every platform reconstructs it its own way, with a fairly obvious incentive: claim as many conversions as possible for itself.

The double-counting paradox

Here's practical proof anyone can run in five minutes. Take the conversions Meta says it generated this month. Take the ones Google says it generated. Add email and affiliates. Then compare the total to the real orders in your order management system.

Almost always, the sum of the channels is higher than the real total, sometimes by 30-50%. Why? Because the same order gets claimed by multiple platforms. The customer saw a post on Instagram, then searched the brand on Google, then got an email: all three take credit. No one is lying, but no one is telling the whole truth either. And you're deciding where to put your money based on these inflated numbers.

Why ROAS pushes you toward bad decisions

ROAS (return on ad spend) inherits every problem attribution has, and adds a few of its own. It's the most quoted metric and, in our view, the most dangerous one when you look at it in isolation.

Three concrete reasons.

1. It measures revenue, not profit. A ROAS of 4 on a product with a 20% margin is a loss once you've paid for the product, shipping, and returns. A ROAS of 2 on a high-margin product can be excellent. ROAS doesn't know that, because it ignores your P&L.
2. It rewards whoever steals credit. Campaigns that intercept people who've already decided to buy (tight retargeting, brand search) show sky-high ROAS. But those people would have bought anyway. You're paying to take credit for sales that were already yours.
3. It doesn't see past the first purchase. A campaign with a ROAS of 1.5 that brings in customers who repurchase three times a year is worth far more than one with a ROAS of 5 that brings in people who vanish after one order. First-purchase ROAS is short-sighted by definition.

We see the extreme case often: a company cuts "unprofitable" campaigns with low ROAS (the ones reaching new audiences at the top of the funnel), keeps only high-ROAS retargeting, and within six to eight weeks total sales collapse. It had switched off the engine feeding the retargeting. If you've ever wondered whether the return you're seeing is real or just the brand claiming credit for itself, comparing MER against ROAS as your reference metric is the right place to start.

The metrics that actually matter

The good news: you don't need perfect attribution to make good decisions. You need a handful of solid metrics, hard to fake, that look at the business from above instead of channel by channel. There are three of them.

1. MER (Marketing Efficiency Ratio)

MER is the most honest metric there is, because it doesn't depend on any attribution at all. The formula is this:

MER = Total revenue / Total marketing spend

Take everything that comes in (from your order management system, not from the platforms) and divide it by everything you spend on advertising, across every channel combined. A MER of 4 means every euro spent on marketing is associated with four euros of real revenue. No double counting possible, because there's only one numerator: real revenue.

MER doesn't tell you which channel is working, and that's its limit. But it tells you the truth about the whole. And when you raise the budget on a channel and overall MER improves, you have your answer without needing to trust any pixel.

2. Contribution margin, not revenue

Stop thinking in revenue and switch to margin. The question isn't "how much did I bill" but "how much is left in my pocket after paying everything that varies with the sale": product cost, shipping, fees, returns, and the ad spend itself.

Many companies discover, once they run this math, that the products they push hardest are the ones with the thinnest margins. A flattering ROAS on a loss-making product is a trap: the more you sell, the more you lose. Thinking in margin realigns you with the only thing that actually matters, which is cash.

3. LTV/CAC: the ratio that decides whether you have a business or a hole

This is the heart of it all. Two numbers:

• CAC (customer acquisition cost): how much you spend on average to acquire a new customer. If you want to work on the denominator, we've gathered the practical levers in how to reduce customer acquisition cost.
• LTV (customer lifetime value): how much margin that customer leaves you over the entire relationship, not just the first order. The step-by-step calculation method is in how to calculate customer lifetime value.

The LTV/CAC ratio tells you whether your acquisition model is healthy. As a widely used benchmark, a ratio below 1 means you're losing money on every customer; around 3 is considered solid; much higher can even signal that you're investing too little and leaving growth on the table. These aren't sacred thresholds, but they give you the scale. And above all: a healthy LTV/CAC ratio lets you tolerate a higher CAC (and therefore a lower ROAS) on acquisition campaigns, because you know that customer pays you back over time.

This trio, MER, margin, and LTV/CAC, is the real directional dashboard. If you want the broader picture of every unit-economics indicator and how they fit together, we laid it out in the analysis on CAC, CPL, and LTV as acquisition KPIs.