By Malin Andersson, Commercial Director Retail & DTC, Criteo
If you turned the hands of time back a decade or so, marketing measurement was a simpler creature.
The questions at the heart of measurement were straightforward: which touchpoint got the conversion? Which campaign performed best?
The data was trackable, the paths to purchase were significantly shorter and more linear. The questions keeping CMOs up at night today are far more complex – did our marketing drive incremental business outcomes? Where should we invest the next $2M? And can we prove to the CFO that ad spend drove results we wouldn’t have achieved without it?
74% of CMOs now report being under more scrutiny to prove marketing ROI than at any point in their careers, and 63% face increased pressure from their CFO to justify spend, which is up from 52% just two years ago. For DTC brands, where every acquisition dollar is already under scrutiny, this pressure is even more acute.
At the same time, the consumer journey is becoming harder to observe. 84% of European consumers now use AI tools in everyday life. LLM recommendations, creator influence, connected TV, cross-device browsing – purchase decisions are shaped by touchpoints that leave no clicks to track.
In this piece, I want to lay out a more sophisticated way of thinking about measurement. I’ll cover:
- The triangulation of attribution, incrementality, and marketing mix modelling (MMM)
- The organizational challenge that trips most teams up
- Where I think measurement is heading next
Let’s dive in.
The question attribution was never designed to answer
Multi-touch attribution (MTA) is a genuine step forward from last-click.
It distributes credit across touchpoints based on relative impact and gives a more sophisticated view of cross-channel influence and upper-funnel contribution.
For DTC brands running campaigns across social, search, and the open web, MTA is increasingly important and worth investing in.
However, attribution (whether last-click or multi-touch) still doesn’t answer the questions your CFO is asking – did our marketing cause the sale, or would it have happened without us?
Questions like this are the reason why CMOs are under growing pressure to prove the incremental value of their marketing – not just what correlated with a sale, but what caused it.
Why one measurement method will never be enough
No single method of measurement is sufficient on its own.
The industry is converging on a triangulated approach, and for good reason: each method answers a different question.
Nearly half of US marketers are already investing more in marketing mix modelling, and over a third more in incrementality testing – the shift is well underway.
I think of these as three lenses that work together to give you a complete picture.
Attribution: The daily pulse
This tells you what happened, which touchpoints fired, and which campaigns performed. It feeds your daily optimization and performance reporting.
Last-click has survived this long not because anyone still defends it as the truth, but because it’s easy. It’s an operational compromise, one that’s common in organizations that haven’t yet solved cross-functional measurement alignment.
I think about it the same way I’d think about measuring the success of a film based entirely on the final scene. The writing, the casting, the character development, the tension built through the first two acts – these all matter just as much. Last-click has the same blind spot: it captures the final touchpoint and ignores everything that built toward it.
There is no universally correct attribution model. But whatever model you choose should reflect how consumers actually behave today, not how they shopped five years ago.
Incrementality: The causal proof
Attribution optimizes, incrementality validates.
This is where you move from tracking what happened to proving what your marketing achieved – and it’s the discipline that directly answers the incremental value question keeping CMOs up at night.
The methods range from geo holdouts (withholding ads in specific regions to compare sales against regions where ads ran) to ghost bidding (entering auctions for a control group but deliberately not winning, so you can isolate the ad’s effect), each with trade-offs in cost, accuracy, and setup complexity.
Something to bear in mind, though – incrementality testing carries a real cost because you’re deliberately withholding marketing from a control group to measure the difference.
Most brands run these tests in key markets or with specific customer groups, and that periodic, targeted cadence is the right model for the long term.
For DTC brands, incrementality testing answers the question that no attribution dashboard can: did this campaign bring in customers we wouldn’t have reached otherwise, or were we paying for conversions that would have happened anyway?
Something else worth knowing not all incrementality tests are equal.
Different partners use different methodologies, and the way a test is designed directly shapes the result. This is why it is critical to understand how a partner defines and constructs incrementality, because it directly determines whether reported performance reflects real business growth or simply the assumptions embedded in the measurement system.
It’s worth pushing your partners on methodology and asking:
- How do you define the eligible population and denominator? In other words, who exactly is being counted in the test?
- How is the control group constructed, and what evidence shows it is truly neutral relative to the exposed group?
- How do you ensure the test is adequately powered to detect real lift with statistical confidence?
- How do you manage the trade-off between tighter scope and cleaner signal, without introducing bias into the result?
- How have you designed the timing of the test to isolate marketing impact from seasonality, promotions, and other market noise?
The point is to work with partners who use sophisticated, defensible methods and who can explain clearly why their approach produces the result it does.
The more rigorous the methodology, the sharper the picture you get of what’s actually driving results – and the more confidence you have in the budget decisions that follow.
Here’s a useful primer on how incrementality testing works if you want to go deeper.
MMM: The strategic compass
This is the long view – where should the next dollar go? Marketing mix modeling uses historical data to forecast future ROI across channels.
A concept I find useful when explaining this to leadership is the idea that the first marketing dollar in any channel almost always pays off. But what’s the return on dollar 50,001?
Simply put, spending more in one channel doesn’t produce linear results – there are diminishing returns.
MMM helps you see where you sit on that curve, so you can shift budget before returns start declining. For DTC brands diversifying beyond walled gardens, this is where you start to see which channels are genuinely earning their share of spend.
MMM is also privacy-resilient by design – it relies on statistical modeling, not user-level tracking. And because it provides a cross-channel, cross-partner view, no single vendor is grading their own homework.
When an independent model measures across all your channels and partners, you get a picture that isn’t shaped by any one platform’s attribution logic.
Your partners can only optimize toward what you share
The three lenses give you a way to see performance more clearly. The next step is building a closed loop optimization with a conversion feedback loop – where attribution signals are fed back to the bidding system to optimize your spend.
Optimization engines maximize toward the signals they receive. When you share your attribution data with your ad tech partners – how you attribute sales, which KPIs matter most, what a valuable customer looks like for your business – their engines optimize toward your goals, not their own defaults.
This connects to something I explored in the previous piece in this series. The better the signal going in, the better the outcome coming out. Attribution data is that signal.
For DTC brands, this is where first-party data becomes a competitive advantage.
Purchase history, average order value, full-price versus discount buyers, return rates – sharing these signals closes the loop between what you measure and how your spend gets optimized.
Some brands hesitate to share attribution data, and the concerns are fair.
There can be real challenges around data quality, scale, and technical setup. But there are multiple ways to solve them, from standard real-time integrations to custom configurations.
I do want to emphasize this point, though: a light integration beats no integration, every time. You don’t need a perfect data pipeline to start closing the loop. Even a basic signal (your attribution model or your core KPIs) gives your partners something real to optimize toward.
The questions that change the conversation
You may not have all the answers, but asking all the right questions is certainly the right place to start.
For your team:
- Does our attribution model reflect how consumers actually behave today?
- How do we combine attribution, incrementality, and MM into a coherent decisioning framework?
- Do we have a shared definition of success across the organization, with a clear hierarchy of KPI’s that we actually steer by?
For your ad tech partners:
- Can your engine optimize toward our source of truth?
- How robust is your incrementality methodology – and how does it meaningfully capture true business impact with statistical rigor?
Why good measurement is undone by bad alignment
Even the best measurement model in the world won’t move the business if the organization around it isn’t aligned.
This is something I see often. MMM and incrementality outputs cut across marketing, finance, analytics, and brand – yet each function is typically optimizing toward different KPIs.
Media teams care about ROAS. Finance cares about margin, brand cares about awareness and engagement, ecommerce cares about revenue, and for analytics it’s data quality.
When these teams draw different conclusions from the same data, budget decisions either stall or get made on gut feel instead.
For most DTC organizations, the real work is embedding measurement into shared KPIs, clear budget ownership, and recurring cross-functional decision cycles. That’s what turns measurement into a competitive advantage.
You don’t need to overhaul everything tomorrow. But you do need everyone optimizing toward the same definition of success.
Where measurement is heading
The triangulated approach I’ve described is today’s best practice. But things are moving fast, and I want to share what excites me about where measurement is heading.
MMM, attribution, and incrementality are converging into a connected system, with incrementality emerging as the calibration layer. The measurement platforms built around them are evolving in step – increasingly designed to recommend where your next dollar should go, not just report on where the last one went.
How teams access insights is changing too.
Conversational AI is replacing dashboards as the primary interface to measurement data. Teams can ask, “What’s the incremental ROAS of Meta versus TikTok?” or, “What happens if we increase retargeting budget by 10%?” and get answers directly.
This means marketers, finance leads, and brand teams can interrogate measurement data directly – without having to rely on anyone else to build a report.
The next step is agentic – AI that doesn’t just surface insights but executes optimization actions within advertising platforms, from budget reallocations to bidding adjustments, with varying degrees of autonomy.
And incrementality testing is evolving from a periodic exercise into an embedded layer within media execution. For DTC brands, this means constant validation that spend is driving incremental results – not as a quarterly project, but as a persistent operating discipline.
The tools are moving fast. The starting point, as always, is being willing to ask harder questions of your data – and of the partners spending your budget.
