Measurement and Attribution: Stop Pretending the Numbers Are Certain

Measurement and attribution are the foundation of accountable marketing, but most businesses are building on sand. Attribution models assign credit to touchpoints, measurement frameworks track outcomes, and together they are supposed to tell you what is working. In practice, they tell you a version of what is working, shaped by the model you chose, the data you collected, and the assumptions baked in before you ran a single campaign.

That is not a reason to stop measuring. It is a reason to measure more honestly.

Why Attribution Is a Model, Not a Mirror

When I was running an agency and managing significant media budgets across multiple clients, one of the most uncomfortable conversations I had regularly was this: a client would look at their last-click attribution report, see paid search taking most of the credit, and cut their display and social spend. Within two quarters, their paid search performance would soften. The leads were still coming in, but the cost per acquisition was climbing. What had changed? The awareness activity that was warming the audience had gone. Last-click attribution had told them a story that was coherent, internally consistent, and wrong.

This is the central problem with attribution. Every model is a simplification. Last-click says the final touchpoint deserves all the credit. First-click says the first one does. Linear distributes credit evenly. Time-decay weights recent touches more heavily. Data-driven models use algorithmic weighting based on observed patterns. None of them are wrong exactly. None of them are right exactly. They are all approximations of a customer experience that is messier, more human, and less trackable than any model can capture.

Forrester has written clearly about why sales and marketing measurement need to be aligned but not identical, and the distinction matters. Marketing measurement is not the same as sales reporting. It operates over longer time horizons, involves channels that resist clean attribution, and requires a different tolerance for uncertainty. Collapsing those two things together is where a lot of measurement frameworks go wrong.

What Most Measurement Frameworks Are Actually Built to Do

I have reviewed a lot of measurement frameworks over the years, both in agencies and as an Effie judge seeing behind the curtain of how brands present their effectiveness cases. And there is a pattern that repeats. Most measurement frameworks are built to justify spend, not interrogate it. They are constructed after the strategy is set, designed to show the metrics that will perform well, and reported in ways that make the results look coherent even when they are not.

This is not always cynical. Sometimes it is just habit. Marketers learn early that the job involves selling ideas internally as much as externally, and measurement gets drafted into that process. The dashboard becomes a defence mechanism rather than a diagnostic tool.

The consequence is that marketing investment decisions get made on data that was never designed to challenge the assumptions behind them. If you set up your measurement to confirm your strategy, it will. That is not insight. That is confirmation with extra steps.

If you are building or rebuilding your measurement capability, the wider context on marketing analytics and GA4 at The Marketing Juice covers the tools, frameworks, and practical decisions involved across the full analytics stack.

The GA4 Transition and What It Exposed

The migration from Universal Analytics to GA4 was significant for most marketing teams, and not just technically. It forced a lot of organisations to confront how much of their measurement infrastructure was held together with assumptions they had never examined. Session-based reporting was replaced with event-based reporting. Bounce rate disappeared and was replaced with engagement rate. Historical data did not carry over cleanly. Reports that had been running for years suddenly needed rebuilding from scratch.

For some teams, this was a genuine opportunity to build something better. For others, it was a scramble to recreate what they had before, as quickly as possible, without questioning whether what they had before was actually useful. Moz covered some of the GA4 features most teams overlook, and the gap between what GA4 can do and what most teams are using it for is significant.

The GA4 transition also exposed something more fundamental: most marketing teams do not have a clear measurement philosophy. They have tools. They have reports. But they do not have a coherent view of what they are trying to understand, what level of certainty is achievable, and what decisions the data is supposed to support. Without that, even the best analytics platform is just producing numbers in search of a narrative.

Getting the setup right matters before any of the analysis does. Semrush has a solid walkthrough on how to set up Google Analytics properly, which is worth going through methodically rather than treating as a one-time task.

UTM Parameters: The Simplest Thing Most Teams Get Wrong

In a previous agency role, I inherited a client account where the UTM tagging was a disaster. Different team members had been adding UTMs with inconsistent naming conventions for two years. Some campaigns were tagged, some were not. The same channel appeared under three different names depending on who had set up the link. The result was a reporting environment where you could not trust any channel-level data because the data was contaminated at source.

UTM parameters are not glamorous. They are also not optional if you want to understand where your traffic and conversions are actually coming from. A consistent, enforced naming convention, applied across every paid link, every email, every social post, is the unglamorous foundation that makes everything else in your attribution model worth something.

Semrush has a thorough guide to UTM tracking codes and how they work in Google Analytics that covers the mechanics clearly. The mechanics are not complicated. The discipline required to apply them consistently across a team, over time, across multiple campaigns, is where most organisations fall down.

The fix is not technical. It is governance. Someone needs to own the naming convention, document it, and enforce it. Without that, you are measuring noise.

The Difference Between Marketing Metrics and Business Metrics

One of the clearest things I saw when I was turning around a loss-making agency was how disconnected marketing activity metrics were from the business outcomes that actually mattered. The team could tell you click-through rates, engagement rates, cost per click, and impression share. They could not tell you which campaigns had contributed to revenue, which had not, or what the margin looked like on the clients that came through paid channels versus organic ones.

Marketing metrics and business metrics are not the same thing, and conflating them is one of the most common ways measurement frameworks fail. Click-through rate is a marketing metric. It tells you something about creative performance and audience relevance. It does not tell you whether the campaign made money. Cost per lead is a marketing metric. It tells you about acquisition efficiency at one stage of the funnel. It does not tell you whether those leads converted, at what margin, or whether they stayed as customers.

HubSpot made this distinction well in their piece on why marketing analytics is not the same as web analytics. Web analytics tells you what happened on your site. Marketing analytics is supposed to tell you whether your marketing investment is generating business value. Those are related but distinct questions, and most measurement frameworks blur them.

Mailchimp’s overview of core marketing metrics is a useful reference for understanding what each metric actually measures, and more importantly, what it does not. The discipline of being precise about what a metric can and cannot tell you is underrated.

Honest Approximation vs False Precision

There is a particular kind of measurement theatre that I find more damaging than having no measurement at all. It is the dashboard that reports to three decimal places, with attribution percentages that add up to exactly 100%, presented to the board as though it represents a settled account of marketing’s contribution to revenue. It looks authoritative. It is not.

The precision is cosmetic. The model that produced those numbers made assumptions about cross-device journeys, about offline touchpoints, about the time lag between exposure and purchase, about what would have happened without the marketing. Each of those assumptions introduced uncertainty. The dashboard hid that uncertainty behind clean formatting.

The alternative is not to abandon measurement. It is to present measurement honestly. To say: based on our attribution model, which weights recent touchpoints more heavily, paid search appears to be contributing approximately 40% of tracked conversions. We know this model does not capture brand awareness activity or offline touchpoints. Our best estimate, accounting for those gaps, is that the true contribution is somewhat lower. We are making budget decisions based on that range, not on a single number.

That kind of honest approximation is more useful than false precision. It builds credibility with finance teams and leadership who are, in my experience, more sophisticated about uncertainty than marketers tend to assume. The CFO who has built financial models knows that projections involve assumptions. What they distrust is when the marketing team presents data as though it does not.

Where the Buyer experience Breaks Attribution

Forrester has written about how marketing measurement can undermine the buyer experience by optimising for the touchpoints that are easiest to track rather than the ones that actually matter. This is a structural problem with digital attribution, and it does not have a clean solution.

The buyer experience is not a funnel. It is not a linear sequence of trackable events. It involves conversations, word of mouth, brand memory formed over months or years, content consumed without clicking, ads seen without being recorded, and decisions made in contexts that no analytics platform can observe. Attribution models work on the fraction of that experience that leaves a digital trace. They then extrapolate from that fraction as though it represents the whole.

This is particularly acute in B2B, where buying committees involve multiple people, sales cycles run for months, and the decision to shortlist a vendor might have been influenced by a conference talk, a LinkedIn post, and a recommendation from a former colleague, none of which appear in any attribution report. I have worked with B2B clients who were making major budget decisions based on attribution data that was capturing perhaps 20% of the actual influence on their pipeline. The other 80% was invisible to the model.

The honest response to this is to triangulate. Use your attribution data as one input. Supplement it with customer surveys asking where they first heard of you, with sales team intelligence about what influenced deals, with brand tracking data, and with your own commercial judgment about what the business was doing when performance improved or declined. No single data source is sufficient. The combination, interpreted with appropriate humility, gets you closer to something useful.

Building a Measurement Framework That Earns Trust

When I grew an agency from 20 to 100 people and moved it from loss-making to a top-five market position, one of the things that changed was how we reported to clients. Early on, the reporting was defensive. It showed the metrics that looked good and buried the ones that did not. As the agency matured and the client relationships deepened, we shifted to reporting that was more honest about uncertainty, more explicit about what we did not know, and more focused on the business questions the client actually cared about.

That shift was uncomfortable at first. It felt like admitting weakness. What it actually did was build trust. Clients who had previously been skeptical of our numbers started engaging with them more seriously, because they could see we were not just presenting the best version of the story.

A measurement framework that earns trust has a few consistent characteristics. It is clear about what it measures and what it does not. It is consistent in its methodology over time, so trends are meaningful. It connects marketing metrics to business outcomes, not just activity. It presents ranges and confidence levels rather than false precision. And it is used to make decisions, not just to fill slides.

That last point is the one most often missed. Measurement that does not change decisions is not measurement. It is reporting theatre. If your analytics process never produces a finding that causes you to stop doing something, or to shift budget away from a channel, or to challenge a strategy that seemed to be working, then the framework is not doing its job.

The broader analytics and GA4 content on The Marketing Juice goes into the practical side of building these frameworks, including the tool choices, the GA4 setup decisions, and the reporting structures that actually hold up under scrutiny.

The Practical Steps That Actually Move the Needle

None of this requires a sophisticated technology stack or a team of data scientists. Most of the measurement problems I have seen in agencies and in-house teams were not technology problems. They were discipline problems and philosophy problems.

Start with the business question. Before you build any report or set up any tracking, be explicit about what decision this measurement is supposed to support. If you cannot answer that question, the measurement framework will drift toward whatever is easy to track rather than whatever is worth knowing.

Fix your UTM hygiene. Audit every paid link, every email, every social post. Establish a naming convention and enforce it. This is unglamorous work that pays dividends for years.

Choose your attribution model deliberately, not by default. Understand what the model assumes, where it is likely to over-credit and under-credit, and what that means for the budget decisions it will inform. Document those assumptions somewhere visible.

Supplement digital attribution with qualitative data. Ask customers how they found you. Ask your sales team what influenced the deals they closed. Build that intelligence into your view of what is working, even if it cannot be expressed as a percentage in a dashboard.

Report uncertainty. Present ranges. Distinguish between what you know, what you estimate, and what you do not know. This is not a sign of weakness. It is a sign that your measurement framework is honest enough to be trusted.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions