We sold ourselves a lie for twenty years.
The promise was simple. Digital would make everything measurable.
Every click tracked. Every conversion attributed. Every pound spent accountable to the penny.
Finance would finally trust marketing because the numbers wouldn’t lie.
The numbers lied. They always did. We just didn’t want to see it. We’re living in a world of digital without trust.
Digital Without Trust World
The Great Marketing Measurement Circle
We’ve come full circle. Before the internet, we couldn’t measure anything meaningful.
John Wanamaker famously said half his advertising was wasted. He just didn’t know which half. 🤷♂️
Then digital arrived. Suddenly, we could track everything. The third-party cookie built an invisible surveillance architecture across the web. Marketers could follow users from an ad impression to a purchase. Multi-touch attribution promised to assign fractional credit to every touchpoint. Return on Ad Spend became the holy metric. The CFO smiled. Finally, marketing could prove its worth.
That world collapsed. And most businesses are still pretending it didn’t.
The convergence of privacy regulations like GDPR and CCPA, browser restrictions like Safari’s Intelligent Tracking Prevention, and Apple’s iOS 14.5 App Tracking Transparency has fundamentally broken the measurement system. Between 30% and 50% of users reject tracking when given a clear choice. Those users still browse and buy. But in your analytics, they don’t exist.
We’ve returned to Wanamaker’s problem. Except now we have dashboards full of impressive-looking numbers that give us false confidence. That’s worse than having no numbers at all.
Why Your Dashboard is Lying to You
Platform-reported metrics have become increasingly modelled and opaque.
When you log into Meta Ads Manager or Google Ads, you’re looking at statistical probabilities presented as facts. The platforms use machine learning to estimate conversions from users who rejected tracking, based on those who accepted it.
The problem? Consenting users are biased samples. They tend to be older, less tech-savvy, or located in regions with looser privacy laws. The models assume these users represent everyone. They don’t.
Here’s where it gets worse. Each platform operates as a walled garden with self-attribution logic.
When a user sees your Instagram ad, then searches your brand on Google and buys, both platforms claim 100% credit. Meta claims the view-through conversion. Google claims the click. Add up the attributed revenue across all your platforms, and you’ll often find it exceeds your actual total revenue.
This inflation gives a false sense of security. You think your Facebook ROAS is 4.0.
The actual business impact might be 2.0. Or 1.5. Or worse.
The platforms have a financial incentive to show higher conversions. They’re grading their own homework with a grading key they refuse to show you. When you challenge the numbers, they point to proprietary algorithms. Trust us, they say.
I don’t.
The Last-Click Trap
In the absence of reliable cross-platform tracking, many marketers retreat to the simplest metric: last-click attribution. This assigns 100% credit to the final touchpoint before conversion.
In a complex, multi-device, multi-channel world, last-click is strategically blinding.
Discovery channels get destroyed. Top-of-funnel activities like YouTube videos or TikTok content often initiate interest but rarely drive the final click. A user sees your brand video on TikTok, becomes interested, and three days later searches for your brand name on Google to make a purchase. Last-click gives Google Search 100% credit. TikTok gets nothing.
This leads marketers to cut budget from the very discovery engines fuelling their growth. It’s like measuring a football team’s success by counting only the goals, then firing everyone who doesn’t score. You won’t have any midfielders to create chances. The goals dry up. And you wonder what happened.
Offline channels become invisible. Billboards, print, word-of-mouth, and shares in WhatsApp or Slack leave no digital trace. Single-platform models implicitly value these contributions at zero. In reality, they often provide the baseline brand lift that makes performance channels work efficiently.
The Triangle That Might Save Your Budget
Forward-thinking organisations have abandoned the pursuit of perfect tracking. They’ve embraced what the industry calls Triangulated Measurement. This approach synthesises three distinct methodologies, each with its own strengths and flaws.
Marketing Mix Modelling for strategic planning. MMM uses historical time-series data to quantify the impact of marketing and non-marketing drivers on sales. It operates on aggregate data, such as weekly spend and weekly revenue. No cookies needed. No user-level tracking required. It naturally accounts for offline channels, seasonality, economic factors, and organic baseline sales. The limitation? It’s correlational. Without calibration, it can confuse correlation with causation.
Incrementality Testing for causal validation. This measures the true lift generated by a marketing activity by comparing an exposed group against a control group. Geo-lift tests are the gold standard. You select two matched geographic regions with similar sales histories. You run ads in one, go dark in the other. The sales difference reveals the incremental impact of your advertising. The limitation? It requires significant time, budget, and statistical discipline. You can’t test everything.
Attribution for tactical execution. Despite its flaws, platform attribution still tells you which creative performed better than others, which audiences converted more, and which placements drove more clicks. It’s useful for day-to-day optimisation decisions. Just don’t mistake it for business truth.
The magic happens when you layer these together:
- Run an MMM to establish a baseline understanding.
- Identify channels where MMM results conflict with platform data.
- Deploy incrementality experiments on contested channels.
- Feed the experimental results back into the MMM as calibration data.
- This mathematically forces the model to respect experimental truth.
The result? A “translation factor” between what the platform reports and what actually happened.
When Facebook shows a 4.0 ROAS but your calibrated model says the true impact is 2.0, you know to discount platform reports by 50%. You use platform data as a relative proxy, not as the absolute truth.
The Metrics Shift That CFOs Actually Respect
The obsession with platform-specific ROAS has led to efficiency traps. Marketers optimise for ratios that look good on dashboards but stifle actual business growth.
Two metrics matter now.
Marketing Efficiency Ratio captures the halo effect. Calculate it as total revenue divided by total marketing spend. MER ignores the attribution war between platforms. It asks the fundamental question: for every pound we spend on marketing, how many pounds come back? If your billboard campaign lifts Google Search conversion rates, attribution won’t credit the billboard. MER captures the total lift.
Contribution Margin protects against unprofitable scale. Revenue is vanity. Profit is sanity.
A heavy discount strategy might double conversion rates and boost ROAS. But lower average order values combined with the discount might mean you’re losing money on every sale. Mature organisations set contribution margin targets rather than ROAS targets.
The hierarchy becomes clear. Contribution margin as your north star for profitability. MER as your holistic efficiency check. Incremental ROAS validated by experiments for channel decisions. Platform ROAS only for day-to-day bidding and creative optimisation.
The Cultural Problem I’m Seeing
The hardest part of this transition is far from being technical. It’s telling a team that their “high-performing” channel is actually failing. 😅
When you switch from last-click to incrementality-based measurement, the reported ROAS for retargeting campaigns often collapses. I’ve seen branded search ROAS drop from 15.0 to 1.5 when measured properly. That’s not because performance dropped. The previous numbers were illusions.
Media buyers whose bonuses are tied to ROAS targets will panic. Agency partners whose margins depend on inflated platform metrics will push back. Finance teams who finally thought they understood marketing will demand explanations.
Leaders must educate stakeholders that the new numbers reveal the truth, not decline. The focus shifts to Incremental CPA: how much does it cost to acquire a new customer who wouldn’t have bought otherwise?
What Actually Works in Practice
I’ve implemented these approaches across dozens of client engagements. Here’s what I’ve learned.
Start with your data infrastructure: If you’re relying solely on GA4 or Meta Ads Manager, you’re flying blind. Centralise your data in a warehouse you control. Own your own truth before trying to triangulate it.
Run your first experiment: Pick your biggest channel or most contested one. Execute a geo-lift test. The results will either validate your current understanding or shatter it. Either way, you’ll learn something real.
Build cross-functional alignment: Marketing, finance, and data science must agree on definitions before they can agree on results. What counts as a conversion? What’s your attribution window? How do you handle returns? These conversations are tedious. They’re also essential.
Accept uncertainty as normal: The era of pretending we know everything is over. Probabilistic thinking replaces deterministic confidence. Every number comes with error bars, even if you don’t show them. Make decisions based on ranges, instead of points.
The Road Ahead
The constraints on data privacy will only tighten.
Real user data becomes scarcer and legally hazardous with each passing year. Synthetic data offers one path forward. AI models trained on consenting users generate statistically identical “digital twins” that mimic behaviour without containing personally identifiable information. You can run simulations and A/B tests in virtual environments before spending a pound in the real world.
Next-generation MMM tools integrate generative AI to automate the tedious parts of modelling. Feature engineering, hyperparameter tuning, and data cleaning. This makes sophisticated econometrics accessible to mid-market businesses that couldn’t previously afford consultants.
But tools alone won’t save you. 😉
The fundamental shift is philosophical. We spent two decades pretending that more data meant more truth. We collected everything, tracked everyone, and built towering dashboards of impressive metrics. Most of it was noise dressed up as a signal.
The companies that thrive in this new environment will be those that embrace disciplined uncertainty over comforting illusion. They’ll run experiments to discover what actually works. They’ll triangulate multiple imperfect sources rather than trusting any single one. They’ll focus on business outcomes such as profit and customer lifetime value rather than on vanity metrics like clicks and impressions.
Digital without trust doesn’t mean digital without strategy. It means strategy without delusion.
And honestly?
That’s probably healthier for everyone. 😎
Leave a Reply