AI & Measurement
When was the last time a report changed what your team did next? Not informed a conversation. Not validated something you already believed. Actually changed a decision.
Most organizations are not using their measurement to make better decisions. Not because the data isn't there -- it's everywhere. Because data tells you what happened. It doesn't tell you why.
And without the why, you have more to look at and less to act on.
That's the gap most measurement systems were never designed to close. They were built to track and report. Who clicked, what converted, how it performed. The what, in exhaustive detail. The why -- what drove the result, what it means, what you should do differently -- that's the part that produces decisions. And it's increasingly the part that gets left out.
More data hasn't solved this. It's made it worse. Volume creates the illusion of understanding without the substance of it. You feel informed. You're not actually directed. That's not a data problem. It's a design problem. Most systems were never built to answer why. They were built to report what. Until that changes, the gap between information and action doesn't close. It compounds.
Most analytics infrastructure starts with "what can we measure?" rather than "what do we need to decide?" The result is a stack of metrics that are technically accurate and operationally useless. High visibility, low decision utility. Everyone tracks them. No one acts on them.
Even the dominant models reflect that bias.
Nearly 77% of marketers say last-click attribution is the easiest way to track performance -- but not the best way to measure it -- and it miscredits results by an average of 37% compared to incrementality testing.
The industry already knows this. The problem isn’t awareness -- it’s action.
Awareness isn’t the issue. System design is.
You feel informed. You're not actually directed. That gap between informed and directed is where strategy erodes quietly, one cycle at a time.
The organizational cost is real, and it compounds quietly. A 2025 study of 500 senior decision-makers found that 77% only sometimes or rarely question the data they rely on daily, even as 67% worry that over-reliance on those same tools is causing them to miss critical opportunities. Leaders across industries cite the same persistent challenge: years of platform investment, still struggling to turn analytics into trusted inputs for actual decisions.
The bottleneck was never the volume of data. It's always been the design.
The fix sounds simple once you say it out loud. Before you decide what to measure, identify what you need to decide. Who is making the call, how often they make it, and what would actually change their answer. Most organizations skip that conversation entirely and go straight to the dashboard. That's where the design flaw gets locked in.
The reframe is structural. Every metric should map to three things: a decision owner, a decision frequency, and a decision threshold.
Three stress-test questions worth applying to any metric your team currently tracks:
Who decides? If no one owns the decision this metric informs, it's a reporting metric. Full stop.
When do they decide? If the cadence of the data doesn't match the cadence of the decision, the metric is decorative. A quarterly report informing a weekly media call is not measurement. It's noise with a timestamp.
What triggers action? If there's no threshold that changes behavior, the metric is ambient. It fills dashboards without filling decisions.
Run this filter against your last five reports. If you can't point to a specific decision made as a direct result of each one, you have a reporting system. Not a decision system.
Most organizations measure what's in front of them rather than what's beneath it. The channels are tracked. The campaigns are reported. The quarterly numbers land on time. But the work never gets below the surface to the questions that actually drive action: what do people believe, what are they doing, and did we cause any of it.
There are three distinct layers of measurement, and they don't all serve the same purpose.
Perception: What do people believe? Brand health, positioning, message resonance. This is the leading indicator. It's also the layer companies most often underinvest in because it's harder to collect and slower to move.
Behavior: What are people doing? Product engagement, content performance, channel metrics. This is where most teams over-index. It's the easiest layer to collect, which is why it dominates reporting.
Incrementality: What did we actually cause? Investment performance, attribution, optimization. This is the hardest layer -- and the only one that answers the real question: did this work because of us?
Most measurement systems show where demand appears, not what creates it. Attribution tells you where demand was captured. Incrementality tells you whether demand was created.
Most organizations live almost entirely in the behavior layer. It's measurable, it's fast, and it feels rigorous. But behavior without perception is activity without context.
And neither answers the question that determines whether the investment was worth it.
The problems with most measurement systems are not random. They follow patterns. And once you know what to look for, you start seeing them everywhere: in the weekly report nobody acts on, in the metric that survives every planning cycle without ever informing one, in the dashboard that grows more complete every quarter while the decisions it was supposed to support get harder, not easier.
Four patterns show up consistently.
The vanity metric problem. Impressions, followers, page views. High visibility, low decision utility. Numbers like these create feel-good moments that rarely translate into meaningful outcomes and can quietly drive misguided allocation.
The committee metric problem. Designed to satisfy multiple stakeholders, actionable to none. When a metric has to mean something to everyone, it usually means nothing to anyone. Ownership requires specificity.
The lag problem. Measuring outcomes long after the decision window has closed. By the time the report lands, the call has already been made, revisited, or moved on from. Late measurement is history, not intelligence. This matters more than ever: 57% of global business executives say they are missing opportunities because they cannot move fast enough. Slow feedback loops are a strategic liability.
The completeness illusion problem. A full dashboard feels like a complete picture. It rarely is. Standard attribution assumes a neat path from ad to purchase, but real-world marketing doesn't work that way. It overlooks key drivers -- brand, offline media, pricing, market conditions -- leaving much of the true impact invisible.
The common thread across all four is the same. The system was designed around what was easy to produce rather than what was needed to decide. Once that pattern is established it's remarkably hard to break, because everyone has already organized their work, their reporting, and their stakeholder relationships around it. Recognizing it is the first step. The harder part is being willing to build something different.
This is not an abstract problem. Boards want proof. CFOs want answers. Senior leaders across industries are integrating analytics aggressively, yet many still struggle to turn them into trusted inputs for strategy.
What it looks like when it works: Expedia's CEO told investors the company improved targeting and measurement capabilities, reduced their least efficient spend, and reallocated dollars to where they saw the highest incremental return. Marketing budget rose 10% year over year. Room nights booked grew 9%. Revenue grew 11%. Critically, the CFO was brought in as an explicit partner -- accountability owned by finance and marketing together, not housed in one silo and ignored by the other.
That's the model. Not more data, not a better dashboard. A system where measurement connects directly to who decides, when they decide, and what it takes to change what they do.
Most measurement conversations treat perception, behavior, and incrementality as separate workstreams owned by separate teams. Brand tracks sentiment. Performance tracks conversion. Analytics tracks attribution. Each reports up independently, and somewhere in a leadership meeting someone tries to reconcile three different stories about what is working.
That's not measurement. That's three monologues.
The shift happens when you treat the three layers as a system designed to answer a single question: did this work, why did it work, and do enough people believe in us for it to keep working? That's triangulation. Using perception to explain what behavior alone can't account for. Using incrementality to separate what you caused from what would have happened anyway. Using all three together to arrive at a read on the business grounded enough to act on.
In practice it looks like this. A brand generating strong behavioral metrics but declining on perception is losing ground it isn't making up. A campaign that drives impressions and clicks but shows flat incrementality captured existing demand rather than creating new demand. A product that wins on incrementality but scores poorly on perception has a ceiling it hasn't hit yet. Each of those is a different decision. You only see it clearly when all three layers are in the room at the same time.
Deciding with confidence doesn't mean having a complete picture. It means having a triangulated one.
That's what it means to act with confidence. Not certainty. Not a complete picture. A triangulated read on the business that is strong enough to move on and specific enough to know what you're betting on when you do.
The dashboard above is not a technology story. It is a design story.
Every element in it was defined before a single metric was chosen. Who owns this decision. How often they make it. What number would actually change what they do. The metrics followed from that. The thresholds followed from that. The alert -- the one that says consideration has dropped below 48% and a decision window opens in six days -- exists because someone decided in advance what action that number should trigger.
That is upstream design. And most organizations never do it.
They go straight to the dashboard. They add metrics until the view is complete. They build something that looks rigorous and functions as a record. The design flaw is locked in before the first report runs, and it compounds from there.
The fix is structural and it starts with three questions asked before anything is built. Who decides? When do they decide? What would change their answer? Define those and you have the architecture of a measurement system. Skip them and you have a very expensive library of things that happened.
The best measurement systems don't give you more to look at. They give you less to second-guess.
Every metric traceable to a choice someone has to make. Every report timed to the moment it needs to inform. Every threshold explicit enough that it's clear when to act. That's the standard. Not comprehensive. Not sophisticated. Oriented.
Most organizations have it backward. The cost is not just inefficiency. It's the slow accumulation of information that never informs anything -- and the decisions that get made anyway, without it.
What's one metric your team tracks that no one has ever acted on? Start there. That's the design problem. And it's the one worth fixing.
Sources: Heidrick & Struggles 2025 Data, Analytics, and Artificial Intelligence Officers Compensation Survey; Mutinex, Is Your Organization Measurement Ready in 2026?; Deloitte 4Q 2025 CFO Signals Survey; WARC/PwC via GEISTE 2026; Expedia Q4 2025 Earnings, EMARKETER x Work Magic Incrementality to Proof.
