How I Think

7 min read

decision-making, sensemaking, measurement, storytelling, strategy

Good decisions live between story and measurement. These are the principles I use to find hidden tensions, test intuition, and turn analysis into action.

During global campaign reviews, a familiar moment kept repeating. A concept would be on the screen, and the conversation would turn on two questions.

Will this resonate?

How would we know?

The first question belongs to story. It asks what people care about, which tension they recognize, what they might believe after hearing something they did not believe before.

The second belongs to measurement. It asks what changed, compared with what, for whom, and whether the signal is strong enough to justify a decision.

Marketing often separates these questions. The creative team protects the story; the analytics team protects the numbers. A campaign moves between them as if meaning and evidence were different stages of a production line.

I have never been able to think about them separately. A story without measurement can become a beautiful explanation for something that is not happening. Measurement without story can become a precise account of behavior nobody understands.

My way of thinking lives in the movement between the two. Story makes ambiguity coherent. Measurement pushes back on the story. The work becomes useful when that tension changes a decision.

The Story Finds the Tension

When a problem is ambiguous, I rarely start by asking for more ideas. I look for the tension that the obvious framing has hidden.

Working on communication products across markets, I learned that the same product could not be explained the same way everywhere. The facts might travel. The meaning did not.

What sounded aspirational in one place could feel distant in another; what looked like a clear benefit from inside the company could be irrelevant to the person hearing it. More distribution did not solve that problem. A sharper story sometimes did.

I think of story as a working model of how a person moves from one understanding to another, not as the campaign wrapped around the strategy. It decides which facts matter, what causes what, who is acting, and why the consequence deserves attention.

Jerome Bruner's essay The Narrative Construction of Reality gave language to something I had encountered in practice: people do not simply receive reality as a list of observations. They organize events into accounts that make motives, sequence, and consequence legible.

Karl Weick's work on sensemaking makes a related point inside organizations. Under ambiguity, people act from the explanations they can construct, revise, and share.

The story is already there, whether it is useful or not. A team saying "customers want simplicity" has a story. A dashboard organized around conversion has a story about what progress means. A roadmap that gives one segment priority over another has a story about where value will come from.

The first job is to make that story visible enough to question.

I look for distinctions because they expose the hidden story. Translation and localization sound like neighboring words until a technically correct translation feels wrong to the person reading it. Activity and learning can share the same dashboard until a team ships ten experiments and cannot explain what it knows now that it did not know before.

A useful distinction does more than rename the problem. It reveals the choice the old framing concealed.

Measurement Pushes Back

The most uncomfortable thing about a good story is how quickly it begins to feel true.

Coherence has persuasive force. Once the pieces fit, contrary evidence starts to look like noise.

Measurement earns its place by resisting the story, not by decorating it with percentages. It should show where the explanation breaks.

I used to read aggregate lift as evidence that a story had traveled.

Market-level results changed that interpretation. The apparent win was sometimes carried by the audience already easiest to persuade, while the audience the business needed next remained unmoved.

The measurement was correct. The conclusion was too broad.

That distinction changed what I looked for. A campaign could perform well overall and still fail to move the market, segment, or behavior the strategy depended on.

Lift was easy to report. Strategic reach was harder, and it mattered more.

Metrics always carry a theory, even when the theory is buried in implementation. A completion rate defines where the journey ends. An attribution window decides which actions count as causes. An average hides the distribution it summarizes.

A dashboard also gives visual priority to some behaviors and makes others expensive to notice.

This does not make measurement suspect. It makes measurement designed.

Donald Campbell warned about one consequence in Assessing the Impact of Planned Social Change. When a quantitative indicator becomes important to decision-making, pressure builds to optimize it, often at the expense of the process it was meant to represent.

The warning is usually repeated as a reason to distrust targets. I take it as a reason to inspect the relationship between the target and the story around it.

What would have to be true for this metric to mean what we say it means? Who disappears inside the aggregate? Which behavior could improve the number while weakening the actual outcome? What would we notice in conversation, observation, or qualitative feedback before it appeared in the data?

Data is most valuable when it creates friction, not certainty. It should make the favored explanation work harder.

The Test Is the Bridge

When intuition and data disagree, teams often choose a side. The experienced person defends context; the analyst defends the instrument. The meeting becomes a referendum on whose way of knowing has more status.

I would rather turn the disagreement into a test.

A good test begins with the mechanism inside the story. If starting with a user's motivation should improve completion because it makes the experience feel personally relevant, the test should measure completion.

It should also protect the outcome the new step might damage: qualification, retention, trust, or whatever sits downstream.

The guardrail matters because most stories are locally convincing. They explain the effect we want and ignore the system around it.

A conversion improvement can lower lead quality. A lower acquisition cost can hide weaker retention. A message can perform well with the easiest audience while making the product less legible to the audience the business needs next.

The decision rule matters for a different reason. Without one, a test can produce data forever.

Teams inspect segments, debate significance, ask for another week, and preserve the original disagreement inside a larger spreadsheet. The analysis grows while the decision stays untouched.

So I want to know in advance: what result would increase our confidence? What would weaken it? Which downside would make the apparent win unacceptable? When do we stop?

Not every meaningful question can be settled experimentally. Senior operators routinely decide with incomplete instrumentation, small samples, slow feedback, or consequences that cannot be cleanly isolated.

Judgment still has to carry the remainder. I am not sure that remainder can ever be fully formalized without becoming another metric to game.

Even when a perfect test is impossible, the discipline survives: state the belief and the evidence, then identify what would change your mind. Uncertainty belongs in the decision instead of behind a confident tone.

Thinking Ends in a Decision

Marketing measurement is often most persuasive just before it becomes insufficient.

I have built marketing mix models to understand which investments were associated with growth. The output could rank channels, estimate contribution, and give budget discussions a common language. It could not decide what the business was trying to become.

A channel can look efficient because it captures demand already in the market. Another can look weaker in the short term while creating awareness among the people the business needs next.

If the decision is made on immediate return alone, the model does more than measure the strategy. It quietly becomes the strategy.

That is where strategic review begins. Is the objective to extract more from the current audience, enter a new category, change who considers the product, or learn which message can carry the next stage of growth?

The same analysis can support different choices because the objective changes what counts as good evidence. Scaling the current mix may be right when efficiency is the constraint. Rebalancing toward a less proven channel may be right when the business needs to create demand rather than harvest it.

Neither choice lives inside the model. The model clarifies the tradeoff; judgment assigns value to each side.

The work is complete when someone can decide where to invest, what not to fund, which risk is acceptable, and what evidence would justify reopening the choice. Analysis that cannot alter an allocation, audience, message, or market priority is still unfinished.

AI makes that boundary easier to see because execution can outrun interpretation. Teams can produce more messages, analyses, and experiments before deciding which market deserves pursuit or what kind of growth is worth buying. The scarce skill moves toward framing the question, choosing what evidence counts, and knowing when the result is trustworthy enough to act on.

A story gives the decision meaning. Measurement gives it resistance. Judgment is what remains after both have had a chance to change your mind.

Sources