Why MTA Is Broken – And Why Unified Measurement Is the Only Way Forward

The Promise vs. the Reality of Multi-Touch Attribution

When multi-touch attribution (MTA) arrived, it felt like a long-overdue upgrade. Marketers had spent years giving all the credit to the last click, a tactic that made email and retargeting campaigns look brilliant and awareness campaigns look useless. MTA promised something smarter: a data-driven view of how every touchpoint in the customer journey contributed to a conversion.

That promise was compelling. The reality has been far less so.

After billions of dollars in ad spend optimized by MTA models, the uncomfortable truth is this: 

MTA doesn’t measure marketing effectiveness. It measures the appearance of marketing effectiveness. 

In a world where privacy regulations are tightening, cookies are disappearing, and channels are more fragmented than ever, the gap between what MTA claims to know and what it actually knows has never been wider.

This isn’t a fixable product problem. It’s a foundational methodology problem.

What Multi-Touch Attribution Actually Measures

Before diagnosing what’s broken, it’s worth being precise about what MTA does.

Multi-touch attribution works by tracking individual user journeys, collecting data on every ad impression, click, email open, and website visit that preceded a conversion, and then distributing credit across those touchpoints according to a set of rules (linear, time-decay, data-driven, etc.).

The core assumption baked into every MTA model is that the touchpoints you observed caused the conversion. 

It also assumes that you can track every touchpoint. 

Those assumptions are wrong.

Five Reasons MTA Is Broken

1. MTA Confuses Correlation with Causation

This is the original sin of attribution. Just because a user saw a retargeting ad before converting doesn’t mean the ad caused the conversion. They may have already decided to buy. The ad showed up in the right place at the right time. A classic case of appearing valuable while doing nothing.

MTA has no mechanism to distinguish between:

• A touchpoint that drove a purchase
• A touchpoint that reached someone already predisposed to purchase

When you optimize toward touchpoints identified by MTA as high value, you’re often not increasing incremental revenue. You’re increasing spend on channels that disproportionately reach people who would have converted anyway. This is the retargeting trap: high attributed ROI, low true incremental ROI.

A study by Uber found that its retargeting campaigns had nearly zero incremental lift, but looked outstanding in attribution models. The same pattern repeats across industries.

2. MTA Requires Data It No Longer Has

MTA depends on stitching together individual user journeys across devices, channels, and sessions. This requires persistent identity tracking, historically enabled by third-party cookies.

Third-party cookies are now effectively dead. 

Safari and Firefox blocked them years ago. Chrome, which still accounts for over 60% of browser traffic globally, has been deprecating them progressively though refuses to officially get rid of them. iOS 14.5’s App Tracking Transparency (ATT) framework decimated mobile tracking.

The result: MTA models are now operating on a fraction of the data they need. 

Most enterprise advertisers have tracking visibility on 30–60% of touchpoints in a user journey, at best. The model fills in the gaps with assumptions, and those assumptions are not disclosed, not audited, and almost never validated.

You’re making multi-million dollar budget decisions based on a model that can’t see most of what it’s supposed to be measuring.

3. MTA Is Structurally Blind to Offline and Upper-Funnel Channels

MTA is built on clickstream data. That means it lives and dies by what’s trackable in a browser or app.

What falls outside its view entirely?

• TV and connected TV (CTV) — reaching audiences who then convert through organic search days later
• Out-of-home (OOH) advertising — no click, no data
• Podcast and audio — brand awareness with no traceable user journey
• In-store purchases triggered by digital campaigns
• Word-of-mouth and earned media — attribution’s perpetual blind spot

If you run a significant upper-funnel TV campaign and see a 20% lift in organic branded search conversions the following week, MTA will credit Google Search. The TV campaign will look worthless. Your TV budget gets cut. Your branded search volume slowly declines. By the time you notice, the damage is done.

This bias toward measurable, lower-funnel, click-based channels systematically distorts budget allocation and it systematically undervalues brand investment. But, no one can search for you if they’ve never heard of you.

4. MTA Breaks Entirely in B2B and Long Sales Cycles

In B2B, a buyer’s journey might span six to eighteen months, involve seven to ten stakeholders, and touch dozens of channels, many of which occur entirely offline (events, sales calls, executive briefings).

MTA can’t model committee-based buying. It can’t resolve the identity problem of multiple people at the same company contributing to one purchase decision. It can’t attribute a conversion event that occurred in a sales call to the webinar a VP attended three quarters ago.

Even in B2C contexts with longer consideration periods, high-ticket retail, insurance, financial services, even gift-giving, the cookie window (typically 30–90 days) means MTA misses the early touchpoints that actually built the relationship.

5. MTA Is Gameable by Vendors and by Walled Gardens

The platforms you’re measuring are also providing the measurement data. 

Meta reports its own attribution. Google reports its own attribution. Amazon reports its own attribution. Each uses proprietary models with undisclosed logic, different attribution windows, and different rules for what counts as an impression, a click, or influenced conversion.

The result: 

Total attributed conversions consistently exceed total actual conversions, sometimes by 2–5x. 

Add up the conversions claimed by Google, Meta, TikTok, and your email platform, and you’ll routinely exceed your actual sales numbers. Every channel is getting credit for conversions it shares with other channels, and no one is accountable for the double-counting.

This isn’t a conspiracy. It’s an inevitable consequence of letting publishers grade their own homework.

MTA vs. Causal Attribution: What’s the Difference?

“Causal attribution” sounds like marketing jargon, but it refers to something technically specific and meaningfully different from MTA.

Dimension	Multi-Touch Attribution (MTA)	Causal Attribution
Core question	Which touchpoints appeared before conversion?	Which touchpoints caused incremental conversions?
Methodology	Rule-based or algorithmic credit distribution	Counterfactual reasoning (what would have happened without this touchpoint?)
Data dependency	Requires full user-level identity tracking	Works with aggregate, panel, or experimental data
Handles unmeasured channels	No	Yes, via modeling
Distinguishes converter behavior	No	Yes, separates incremental lift from baseline
Reliable in a cookieless world	No	Yes
Answers “should I spend more here?”	Poorly. Correlation ≠ causation	Yes, with statistical confidence

The key insight is this: 

Causal attribution asks a fundamentally different question.

MTA asks: “Who was present?” Causal attribution asks: “Who made the difference?”

To answer the causal question, you need one of three approaches:

1. Randomized controlled experiments (geo-lift tests, holdout groups). This is the gold standard, but expensive and slow to run at scale
2. Marketing Mix Modeling (MMM). Statistical regression across aggregate spend and outcome data; doesn’t require user-level tracking; captures offline channels
3. Incrementality measurement. Quasi-experimental methods that estimate the counterfactual impact of specific channels or campaigns

The best-in-class measurement frameworks combine all three, calibrated against each other. That’s the foundation of unified measurement.

What Is Unified Measurement and Why Does It Solve What MTA Can’t?

Unified measurement isn’t a single technique. It’s an architecture that brings together MMM, incrementality testing, and causal attribution and reconciles them into a single source of truth.

Here’s how the three components work together:

Marketing Mix Modeling (MMM) operates at the aggregate level, using historical spend and outcome data to estimate the causal contribution of each channel. It handles TV, OOH, seasonality, price effects, and competitive dynamics. It doesn’t require cookies or user tracking. Its weakness: slow refresh rates and limited granularity at the campaign level.

Incrementality testing runs controlled experiments, geo holdouts, synthetic control groups, platform-level tests, to measure the true lift generated by specific channels or campaigns. It answers “does this channel actually drive incremental revenue?” with statistical rigor. Its weakness: can’t test everything simultaneously.

Causal Attribution (used correctly) serves as a fast, granular signal for within-channel optimization, but only when downstream-calibrated by MMM and incrementality. Used alone, it misleads. Used as one input in a unified system, it contributes signal without distorting budget strategy.

When these three are unified:

• You get MMM’s big-picture causal view for strategic budget allocation
• You get incrementality testing to validate and course-correct
• You get causal attribution’s granularity for tactical, channel-level optimization
• You get a reconciled measurement that doesn’t let any single channel self-report its own success

The Business Cost of Staying on MTA

The consequences of MTA-driven decision-making aren’t theoretical. They show up in your P&L.

Mis-allocated budgets. 

When channels that look good in MTA (retargeting, branded search, last-click email) receive outsized investment, and channels that look poor (upper-funnel display, TV, awareness social) get cut, you hollow out the top of the funnel. Revenue holds in the short term, then gradually degrades as brand equity erodes.

Undervalued awareness investment. 

Awareness campaigns that drive future demand are systematically undercredited in MTA. CMOs who rely on MTA end up justifying cuts to brand budget and then wonder why performance marketing efficiency declines six to twelve months later.

False confidence in incrementality. 

High attributed ROAS from retargeting and paid search may be predominantly capturing demand you already own. MTA gives you no way to know. A single well-run holdout test will often reveal that 50–70% of attributed retargeting conversions would have happened anyway.

Measurement fragmentation. 

Five different platforms. Five different attribution models. Five different conversion windows. No reconciliation. Your data team spends its time explaining why the numbers don’t add up rather than generating insights that improve performance.

What Good Marketing Measurement Looks Like in 2026

If you’re ready to move beyond MTA, here’s what a modern measurement foundation looks like:

1. Establish a causal MMM as your baseline. It doesn’t need to be perfectly granular. It needs to be directionally correct at the channel level for budget allocation decisions. A well-specified MMM with quality data, run quarterly or continuously, will outperform MTA for strategic decisions every time.
2. Run incrementality tests on your highest-spend channels. Start with your biggest paid channels, paid social, branded search, retargeting. A geo-based holdout will give you a ground-truth read on whether those channels are actually driving incremental revenue or harvesting existing demand.
3. Use attribution for within-channel optimization only. Attribution is still useful for deciding which creative, audience, or bidding strategy to run within a channel. It should never be used to compare channels against each other or to justify overall channel spend.
4. Require reconciliation across data sources. Your measurement system should produce a single, coherent view of marketing effectiveness, not three different answers from three different vendors.
5. Calibrate continuously. The media landscape changes. Consumer behavior changes. Your measurement model needs to be a living system, not a one-time audit.

The Bottom Line

Multi-touch attribution was the right idea for a world that no longer exists. It was built for when users could be tracked everywhere, walled gardens played fair, and correlation was a reasonable proxy for causation.

That world is gone.

MTA’s structural problems, its reliance on cookies, its blindness to causation, its inability to handle offline channels, its vulnerability to self-reporting bias, aren’t being patched in the next product update. They’re fundamental to how MTA works.

Unified measurement, built on causal MMM, incrementality testing, and calibrated attribution, gives marketing leaders the honest answer to the question that actually matters: 

Is this marketing investment generating incremental business outcomes?

The marketers and organizations that make this transition will make better budget decisions, prove ROI to finance, and build measurement infrastructure that gets more valuable over time. not less. as privacy regulations continue to tighten.

The ones who stay on MTA will keep chasing attribution windows, reconciling conflicting dashboards, and discovering, too late, that high attributed ROAS doesn’t show up in the P&L.