Measurement is More Like a Hospital Than a Club Sandwich
A modern measurement strategy is often described as stacking methods on top of each other: MMM, experiments, brand surveys, dashboards, MTA, maybe a few more for texture.
That image is tidy. It is also misleading.
Measurement is not a club sandwich. More methodological layers loosely stacked do not make it better. At some point, you end up just trying to take a bite without embarrassing yourself. Plus, stacking more capabilities like a Jenga tower increases the risk of collapse, not more flavorful insights.
A better metaphor is a hospital.
In a well-run hospital, each floor has a purpose. The emergency department is not the radiology lab. Surgery is not recovery. Physical therapy is not the pharmacy. The elegance is knowing which floor fits the diagnostic question in front of you.
If the question is “Is the ankle broken?,” you start with an X-ray. You do not begin with surgery, recovery, and a full specialist tour just because an administrator recently read a white paper on “holistic care.”
Do Not Send Every Patient to Every Floor
Large advertisers are not suffering from a shortage of instruments. The problem is usually more basic: No one has agreed on the analytic protocol.
So, every important question gets sent on the full hospital tour. Start in radiology. Stop by the lab. Call in surgery. Maybe consult physical therapy, just to be safe. By the end, everyone has more data, more meetings, and a very familiar sense that the decision somehow became less clear. The outcome suffers despite more specialist attention.
We curse “analysis paralysis” but fail to build the process mapping that inoculates our teams against it.
Lift studies or incrementality tests are increasingly popular (and helpful) tools, but they are only one part of the larger diagnostic suite. The map for large advertisers uses three or more floors or measurement layers:
- If the decision is about enterprise-level budget allocation, modern marketing mix modeling (MMM), advanced through Commercial Analytics, provides the holistic foundation and should be your first stop. It helps leaders understand the broader commercial picture: Media, pricing, promotion, distribution, competition, seasonality, operations, macro conditions, and demand dynamics.
- If the decision is about a specific intervention — a store activation, offer design, media tactic, customer experience change, staffing model, or rollout sequence — an experiment (or lift study) can pressure-test that choice. They are particularly valuable for testing a new tactic, creative, audience, or noticeably different support level on a response curve.
- If the decision is about in-flight optimization, attribution and other forms of tactical measurement, including A/B testing, can help monitor and adjust activity quickly. These tools help catch execution issues while there is still time to act: Audience overlap, broken targeting, local market, or other signals that suggest the campaign looks unhealthy.
A hospital is efficient and useful because it has a protocol, not because every specialist gets invited to every diagnosis. Each method/layer/floor has a distinct role.
The discipline is to know which floor owns the question before the organization starts ordering tests. A mature measurement system does not confuse activity with intelligence. It does not send every patient to every floor because the hospital needs help to justify the investment in high-priced machines.
Many Business Decisions Belong on One Floor Only
If decisions have a natural first stop, then knowing which floor or department owns the question is the critical discipline. For example, if the decision is about enterprise-level budget allocation, Commercial Analytics—an advanced evolution of marketing mix modeling (MMM)—is the first stop. But applying this simple rule of using Commercial Analytics for cross-channel budget decisions is trickier than it looks.
Take a seemingly simple question: Should we shift spend within Google — for example, from Search into Video, Performance Max, or Display? On the surface, that can look like an attribution or platform-planning question. Google Ads Performance Planner can help forecast how budget or bid changes may affect performance within selected Google campaigns, and that is useful for executional planning. But the bigger decision may not actually live inside Google. Different Google channels can have very different halo effects with media outside Google, response curves, and sensitivity to category demand, competition, seasonality, and macroeconomic pressure. A platform tool can help optimize inside the room, or in this case, within the Google YouTube room. Commercial Analytics helps determine whether you are in the right room.
That distinction matters. If YouTube helps create demand that later shows up in paid search, a search-heavy read will make search look like the hero when it is partly harvesting demand another channel planted. And if consumer demand is shifting because of inflation, weather, competitive pricing, or supply constraints, reallocating within Google based only on platform signals can feel safe while quietly missing the full commercial story. Those performance changes you see in Performance Planner might have very little to do with what is going on with your Google media.
This example shows why building a layered measurement approach that elegantly integrates multiple measurement methods takes effort. Well-intentioned, good-sounding approaches can mislead. This is what layered measurement clarifies: keeping each tool in its lane while making the decision better, faster, and less vulnerable to false confidence.
Google Ads tools still add value in this example. Experiments can test alternative keyword strategies or campaign settings, and Smart Bidding can optimize execution in real time. What narrows is the role of Performance Planner. It can help advertisers manage activity within Google Ads, but investment decisions belong to a different area of the hospital.
Not Every Question Belongs in The Test Wing
Many organizations fall into the habit of treating experiments as the gold standard for every important question. That sounds disciplined until someone starts scheduling exploratory surgery for conditions that only require a diagnostic X-ray.
Experiments create the most value when they own the question. If the decision is about a specific intervention, a new creative approach, an offer design, a store activation, a customer experience change, or whether a tactic should scale, the test wing deserves a visit.
Some of the most useful tests are not designed to validate ROI measures from other methods. They are designed to inform decisions. For example, deliberately varying spend across meaningful investment levels can help determine whether an estimated response curve and diminishing returns relationship is credible. For many budget decisions, that is more valuable than (re-)proving a tactic generated lift under a single set of conditions.
But every test, even the most carefully designed, hits an insight ceiling.
A test may estimate incremental impact in a specific geography, audience, time period, and spend range, yet still tells us very little about what happens elsewhere. Marketplace contamination, audience spillover, competitive reactions, halo effects, and the simple reality that people (especially store or branch managers) behave differently when they know they are being observed can all bend results in unexpected ways. Many tests never capture longer-term effects such as adstock, delayed consumer response, brand building, or interactions among media, pricing, promotion, distribution, and competitive activity.
The goal is not to test everything; that’s inefficient. The goal is to test the questions that deserve a room in the test wing and resist admitting every patient who simply wants a second opinion.
Integrate With Elegance
The best measurement organizations do not treat experiments as side projects. They do not confuse the speed of tactical optimization methods with accuracy. They do not average conflicting methods and call it triangulation because the word sounds technical and reassuring.
They build a layered measurement system with integrity and elegance.
Integrating with elegance means knowing which method belongs to which decision. It means designing experiments thoughtfully. It means using Commercial Analytics to understand the full commercial system. It means bringing results into scenario planning instead of leaving them stranded in recap slides.
Most importantly, it means remembering that measurement exists to improve and accelerate decisions. Not to produce more numbers. Not to create a prettier dashboard. Not to force every method to agree. To help leaders decide where to go next.
Experiments can and should be part of that system. But they create the most value when they are integrated with elegance: not jammed into models, not averaged into mush, and not isolated from the decisions they are supposed to improve.
A mature measurement system does not ask every method to answer every question. It knows which floor to send the patient to. And it does not schedule surgery for a sprained ankle just because the hospital has a really impressive operating room.
