The Marketing Measurement Stack: How to Actually Know What's Working (And What's Burning Cash)

Published: 1 June 2026

If you're spending £50k+ a month on marketing and you're still relying on platform-reported ROAS to make budget decisions... you're flying blind.

I don't say that to be dramatic. I say it because I've sat in rooms with founders spending £200k a month across Meta, Google, TikTok, and influencer partnerships, and when I ask "what's actually driving incremental revenue?" the room goes quiet.

They have dashboards. They have numbers. They have weekly reports from their media buyer showing 4x ROAS on Meta.

But they don't actually know.

Here's the reality as of 2026. Over half of marketers are now using incrementality testing, according to eMarketer's latest research. And 75% of buy-side leaders say their core measurement approaches underperform. The industry knows the old way is broken. The question is what to do about it.

This article is the practical answer. I'm going to walk you through the three measurement methods that actually matter, when to use each one, and how to build a measurement stack that tells you the truth about your marketing. Not what Meta wants you to believe.

Jump ahead

• The three measurement methods (and what each one actually does)
• When to use what. A decision framework you can steal
• The triangulation playbook. How to combine all three
• Real examples of brands getting this right (and wrong)
• Building your measurement stack. Where to start today

The Three Measurement Methods That Actually Matter

Let's cut through the jargon. There are three approaches to marketing measurement that serious brands are using in 2026. Each one answers a different question. And none of them is sufficient on its own.

1. Multi-Touch Attribution (MTA)

What it answers: Which touchpoints in the customer journey get credit for conversions?

MTA is what most brands have been doing for years. It tracks individual users across touchpoints and assigns credit to the channels, campaigns, and ads they interacted with before converting.

It's useful. But it has serious limitations that get worse every year.

The good:

• Granular, campaign-level insights. You can see which specific ad set is performing
• Real-time data for day-to-day optimisation
• Easy to understand and act on quickly

The bad:

• iOS privacy changes, cookie deprecation, and consent frameworks have blown massive holes in tracking. You're seeing maybe 60-70% of the actual picture
• It correlates but doesn't prove causation. Just because someone saw an ad and then bought doesn't mean the ad caused the purchase
• It can't measure offline, word of mouth, or brand awareness impact
• Every platform over-claims credit. Meta says it drove the sale. Google says it drove the sale. The truth? Often neither did exclusively

Best for: Daily and weekly campaign optimisation. Think of it as your tactical cockpit. Good for spotting which creative is resonating, which audiences are converting, and where to shift budget within a channel.

But don't use it to answer "should we move £30k from Meta to TikTok?" It'll lie to you.

2. Marketing Mix Modelling (MMM)

What it answers: How should we allocate budget across channels for maximum return?

MMM takes a top-down, statistical approach. Instead of tracking individual users, it analyses the relationship between your marketing inputs (spend, impressions, campaigns) and business outputs (revenue, leads, LTV) over time. It accounts for external factors like seasonality, economic conditions, competitor activity, and even weather.

This isn't new. MMM has been around since the 1960s when CPG companies used it to measure TV ad effectiveness. But the modern version is radically different.

The good:

• Privacy-proof. It doesn't rely on tracking individual users at all
• Measures everything, including channels you can't track digitally like OOH, podcasts, PR, and organic word of mouth
• Shows diminishing returns and saturation points. This is massive. It tells you where you're wasting money
• Great for strategic budget allocation decisions

The bad:

• Historically needed 2-3 years of data. Modern Bayesian approaches (like Google's Meridian or Meta's Robyn) can work with 12-18 months, but more data is always better
• Slow. It's not giving you real-time optimisation signals
• Can be a black box if you don't understand the methodology
• Needs calibration from incrementality tests to stay accurate

Best for: Quarterly and annual budget allocation. "How much should we spend on Meta vs Google vs influencer marketing next quarter?" That's an MMM question.

According to Measured's 2026 research, 61% of retail decision-makers now use MMM for incrementality measurement. It's no longer a Fortune 500 thing. DTC brands doing £1M+ annually are adopting it.

3. Incrementality Testing

What it answers: Did this specific marketing activity actually cause additional revenue? Or would those sales have happened anyway?

This is the gold standard.

Incrementality testing uses controlled experiments to isolate the true causal impact of your marketing. The most common approach is geo holdout testing. You turn off a channel or campaign in specific geographic regions, keep it running in matched control regions, and measure the difference in outcomes.

If you turn off Meta ads in Birmingham and Manchester but keep them running in Leeds and Sheffield... and sales barely change in Birmingham and Manchester... guess what? Those Meta ads weren't driving as much incremental revenue as the platform told you.

Painful truth. But better to know.

The good:

• Proves causation, not just correlation. This is the only method that definitively answers "did this work?"
• Privacy-proof. No individual tracking needed
• Can test specific hypotheses: "Is our branded search spend incremental or are those people going to find us anyway?"

The bad:

• Requires you to turn things off, which means short-term revenue risk
• Takes time to run properly (usually 2-4 weeks minimum per test)
• You can only test one or two things at a time
• 44% of marketers cite concerns about accuracy and reliability of results as a barrier to adoption

Best for: Validating big bets before scaling. "We're about to 3x our TikTok spend. Is it actually incremental?" Run the test first. If you want a step-by-step on running these, we wrote a detailed geo holdout testing guide that walks you through the whole process.

When to Use What. The Decision Framework

Here's the framework I give every brand we work with. Screenshot this. Stick it on your wall.

MTA is your daily dashboard. Use it for campaign-level decisions within channels.
MMM is your quarterly strategy session. Use it for budget allocation across channels.
Incrementality testing is your reality check. Use it to validate assumptions before making big moves.

Or think of it this way...

MTA tells you what happened.
MMM tells you what to do.
Incrementality testing tells you the truth.

The mistake most brands make? They pick one and treat it as gospel. That's like navigating with only a speedometer. You need the speedometer (MTA), the GPS (MMM), and the fuel gauge (incrementality) to actually get where you're going.

The Triangulation Playbook

The best brands in 2026 aren't arguing about which measurement method is "right." They're using all three together in a system called triangulation.

Here's exactly how to set this up.

Step 1: Start with MTA as your daily operating layer

Use your attribution tool (GA4, Northbeam, Triple Whale, whatever) for daily campaign management. This is where your media buyers live. They need real-time signals to optimise bids, shift budgets between ad sets, and kill underperformers.

But treat these numbers as directional, not absolute. When Meta tells you an ad set has 6x ROAS, it probably doesn't. The direction might be right (that ad set is performing better than others), but the magnitude is almost certainly inflated.

Step 2: Layer in MMM for strategic allocation

Every quarter, run your MMM to answer the big allocation questions:

• Are we spending enough on Meta? Or have we hit saturation?
• Is our Google spend actually driving incremental revenue, or are we just buying branded clicks that would have happened organically?
• Should we be investing more in email, influencer, or content?
• Where are the diminishing returns?

This is where I've seen the biggest breakthroughs with brands we work with. One beauty brand we consulted with was tripling influencer spend assuming they were tripling impact. Their MMM revealed they'd actually hit saturation. By reallocating 25% of that influencer budget to paid social where marginal returns were still strong, they increased LTV-adjusted revenue by 8%.

Let's do the maths on that for a second. If that brand was doing £5M in annual revenue, an 8% lift is £400k. From a reallocation, not additional spend. That's the power of actually knowing your numbers.

Step 3: Validate with incrementality tests

Before you act on any major MMM recommendation, validate it with an incrementality test.

Your MMM says TikTok is delivering strong incremental returns? Great. Run a 3-week geo holdout test to confirm before you move £50k of budget there.

Your MMM says branded search is barely incremental? Run a test. Turn off branded search in a few regions. See what happens to direct and organic conversions in those regions.

The incrementality test is your circuit breaker. It stops you from making a £200k mistake based on a model that might have a flaw you haven't spotted.

Step 4: Feed incrementality results back into your MMM

This is the step most people miss.

Your incrementality test results should be used to calibrate your MMM. If the model said TikTok was delivering £4 return per £1 spent, but your incrementality test shows it's closer to £2.50... update the model. This feedback loop is what makes the whole system get smarter over time.

Without this step, your models drift from reality and you're back to guessing.

Real Examples of This Going Right (And Very Wrong)

The brand that discovered 40% of their Google spend was wasted

I was working with a DTC wellness brand spending about £80k a month on Google Ads. Their MTA showed solid ROAS across branded and non-branded campaigns. Everything looked healthy.

We ran an incrementality test on their branded search campaigns. Turned them off in specific regions for three weeks.

The result? Sales in those regions barely moved. The vast majority of people searching the brand name were going to find the website anyway through organic results. They were essentially paying Google to capture traffic they already owned.

We reallocated £30k of that monthly branded search budget to prospecting campaigns and top-of-funnel Meta activity. Within two months, new customer acquisition increased by 22%.

That's £360k a year they were setting on fire. And their MTA was telling them everything was fine.

The brand that scaled too fast because they trusted platform data

Different brand. Fashion DTC. Meta was showing 5x ROAS on their Advantage+ campaigns. So they tripled spend from £40k to £120k a month.

Revenue went up. But when we looked at the actual contribution margin, it had cratered. Why? The increased spend was mostly retargeting people who were already going to buy. CPA on genuinely new customers had ballooned. And their blended CPA masked the problem because the "cheap" retargeting conversions dragged the average down.

If they'd run an incrementality test before scaling, they'd have seen that the incremental ROAS on that additional £80k of spend was closer to 1.5x, not 5x. Below their breakeven threshold.

This is exactly the ROAS vs contribution margin trap I've written about before. The platform metrics look great while the actual business economics deteriorate.

Building Your Measurement Stack. Where to Start Today

I know this can feel overwhelming. Three different measurement systems? Models? Testing frameworks? Where do you even begin?

Here's my honest recommendation based on what stage you're at.

If you're doing £20k-£50k/month in ad spend

You probably don't need a full MMM yet. Start here:

1. Get your MTA house in order. Make sure you have proper server-side tracking, Conversions API set up on Meta, and enhanced conversions on Google. The data going into your attribution needs to be as complete as possible
2. Run one incrementality test per quarter. Pick your biggest channel and test it. That's it. One test. Learn from it
3. Build a simple weekly dashboard that tracks blended metrics alongside platform metrics. Compare platform-reported revenue to actual Shopify/bank revenue. The gap tells you how much the platforms are over-reporting

If you're doing £50k-£200k/month in ad spend

You're at the stage where a proper measurement stack starts paying for itself many times over.

1. Implement a lightweight MMM. Tools like Google's Meridian (open-source) or paid platforms like Northbeam and Measured can get you started without a data science team
2. Run incrementality tests monthly. Rotate through channels. Test your biggest assumptions first
3. Use MTA for daily optimisation but cross-reference major decisions against your MMM outputs
4. Start tracking incrementally-adjusted ROAS alongside platform ROAS. This is the number that actually matters

If you're doing £200k+/month in ad spend

At this level, every percentage point of wasted spend is tens of thousands of pounds a month. You need the full triangulated stack.

1. Full MMM running quarterly with scenario planning for budget allocation
2. Continuous incrementality testing programme with a testing calendar and dedicated resources
3. Calibrated MTA where your attribution weights are adjusted based on incrementality findings
4. A measurement team or partner who owns the full picture and can translate insights into action

This is where we spend a lot of our time at Elevate. Because measurement isn't just a data problem. It's a decision-making problem. The numbers are only useful if they change how you allocate budget, build creative, and structure campaigns. And that requires someone who understands the full customer journey and how all the pieces connect.

The Uncomfortable Truth About Measurement

Here's what I want to leave you with.

Perfect measurement doesn't exist. It never will. Every model has assumptions. Every test has limitations. Every platform has incentives to make their numbers look good.

But that's not an excuse to do nothing.

The brands that win in 2026 and beyond aren't the ones with perfect data. They're the ones with less wrong data than their competitors. They're the ones who triangulate, test their assumptions, and make decisions based on evidence rather than platform vanity metrics.

According to MarTech, nearly half of marketers are increasing investment in MMM this year, and over a third are investing more in incrementality testing. The shift is happening. The only question is whether you're ahead of it or behind it.

If you're spending serious money on marketing and you're not triangulating your measurement... you're not being data-driven. You're being data-deceived.

And that's a very expensive place to be.