Measuring Communication Without Losing Meaning

Philosophy first taught me to distrust the obvious. A data point is never reality itself: it is a trace of reality, produced from a question, a method, a choice of scope, and a set of categories. Before measuring, we must therefore ask what we are actually measuring—and what the instrument necessarily leaves out.

I like to say that my profession deals in numbers and letters, but letters come before numbers.

In communication, data captures behavior fairly well: exposure, audience, clicks, recall, reputation, or self-reported engagement. But it has more difficulty capturing the meaning individuals give to a message, their hesitations, contradictions, silences, or the way words resonate with their personal history. Two people can give the same answer to a questionnaire for radically different reasons. And an indicator can improve even as the relationship deteriorates.

That is why, in my work, we most often conduct qualitative research phases (open-ended questions) alongside quantitative studies (closed-ended questions).

Philosophy has also taught me to distinguish effectiveness from legitimacy. Just because a message works does not mean it is right, desirable, or sustainable. Communication can be visible without being credible, and perform well without building trust. Not everything that can be measured necessarily deserves to be optimized.

This lies at the heart of the thinking I have developed around the broad issue of responsible communication and marketing.

Data therefore tells me what is happening, on what scale, and among whom. Philosophy helps me understand what it means, why it matters, and what we should do about it. Without data, we risk arbitrary intuition; without philosophy, blind precision.

Yes, I believe so. But that irreducible part must not become an excuse for abandoning evaluation.

Many things can be measured in communication: exposure, understanding, recall, changes in perception, trust, the intention to act or buy, and even certain behaviors. The condition is to start from the objectives being pursued, rather than trying to translate everything artificially into “euros.”

A campaign designed to reassure, make a decision understandable, or preserve the quality of a relationship cannot be judged solely by its immediate financial return. That said, let us try each time, on the basis of this measurement, to establish a benefit, including a financial one: how much is one point of awareness worth? What is a “good” reputation worth? A well-informed employee who can see a future for themselves in the company is talent the company will not need to recruit; that also translates into avoided costs.

What partly escapes measurement is, first of all, the silent transformation produced by communication. A sentence can permanently change the way someone views an institution without provoking any observable reaction. A story can become a reference several years later. A public statement can help establish a climate of trust even though it is impossible to isolate its precise contribution from among all the experiences people have had.

There is also a fundamental difficulty: we can measure present effects more easily than effects that have been prevented.

How exactly do we evaluate a crisis that did not happen, distrust that did not take hold, or a misunderstanding that was cleared up in time? Communication often acts as relational infrastructure: its value becomes especially visible when it disappears or deteriorates.

Measuring absence is one of the challenges of my profession. What risk arises from not having a stand at a particular trade show? What is the value of avoiding a damaging hit piece through intelligent media relations?

Finally, some dimensions have less to do with effectiveness than with appropriateness. Respect for an audience, the sincerity of what is said, and the dignity with which an organization addresses its employees or acknowledges a mistake cannot be reduced to an indicator. We can observe their consequences, but measurement cannot put a number on their value.

I would therefore say that communication is not entirely measurable, but it is almost always evaluable. Measuring consists of producing numbers, whereas evaluating consists of making a reasoned judgment based on objectives, indicators, facts, context, and sometimes a qualitative assessment. The point is not to turn everything into numbers, but never to leave vague what can be objectified.

This is one of measurement’s great paradoxes: as soon as an indicator becomes a target, it risks ceasing to be a good indicator. The organization learns to produce the expected number, sometimes without producing the desired effect. It increases the number of posts, views, or interactions, but not necessarily understanding, trust, or the quality of the relationship.

The first safeguard is always to build a basket of indicators, with each indicator acting as a check on the others.

We must therefore resist the temptation of a single figure. As soon as a complex performance is summarized by one KPI, its optimization is mechanically encouraged. A robust evaluation must combine several dimensions: the production of communication, its exposure, its reception, its effects on perceptions and, where possible, its behavioral consequences.

The second safeguard is always to connect the indicator to the objective: visibility is not influence, recall is not agreement, and media tone is not reputation. If indicators and objectives are not meticulously aligned, there is a serious risk of poor decision-making.

Sometimes, for the sake of convenience, one may say, “This indicator does not correspond exactly to the objective, but it is easy to quantify…” That is where the trouble begins.

I would add that we need to include warning indicators. A campaign can attract a great deal of attention while eroding trust. A statement can generate strong engagement because it polarizes. An internal initiative can achieve an excellent participation rate because employees feel compelled to respond. Every performance indicator should therefore be accompanied by a question about its side effects: what might we have damaged by improving this figure?

The governance of measurement ultimately matters as much as the method. Communications professionals should not be the only ones defining the criteria by which their work will be judged. Objectives and indicators must be discussed with the relevant departments—the “internal clients”—tested against audience expectations, and periodically reviewed. A dashboard that never changes almost always ends up measuring how the system operates rather than whether it remains relevant.

Good measurement is therefore not the kind that supplies the most figures. I call this illness “quantophrenia”: a frenzy for quantification, and it is common. The best antidote to the tyranny of indicators is to treat the dashboard like a GPS: indispensable for finding your way, but always poorer than the territory it represents.

It changed both.

In my personal life, I found myself on a war front, on the front line.

I discovered that producing figures could generate thousands of articles, tens of thousands of insults, two serious death threats, and the filing of a complaint.

I thought I worked behind the scenes—in the kitchen, below deck—and found myself in the shop window.

Counting public demonstrations taught me that a figure can be technically sound and politically contested, not necessarily because the method is flawed, but because it takes a position within a conflict of narratives. For the organizers, the number measures mobilization and contributes to its legitimacy. For public authorities, it helps assess the balance of power. For the media, it becomes a framing device. Measuring a demonstration is therefore never simply a matter of counting people: it means introducing a figure into an already charged symbolic space.

That experience profoundly changed my conception of objectivity. I no longer define it as a position above the fray, one that would make it possible to escape all subjectivity or conflict. Objectivity is instead a discipline of method: making the calculation rules explicit, stabilizing the protocol, documenting the conditions under which the result was produced, accepting contradiction and, as far as possible, enabling verification by third parties. It does not consist in saying, “Trust me, I am neutral,” but rather, “This is how we proceeded; you can examine, criticize, and reproduce our approach.”

That was the case for nearly thirty participating media organizations—from AFP to La Croix, via Libération and Mediapart. We embraced transparency and allowed all stakeholders, including the organizers, to join us in carrying out and checking the count.

On several occasions after demonstrations, we also offered to recount all or part of the event using video footage.

Very few people took us up on that offer, which proves, in a sense, that the objections to our figures were always purely ideological rather than simply technical or methodological.

Ideology is sometimes stronger than facts and stronger than the desire to know.

It also made me more attentive to uncertainty. In public debate, people often expect a single, simple, definitive figure. Yet every measurement involves a margin of error, technical limitations, and conventional choices: where does the procession begin, and how should lateral flows, people moving back and forth, or interruptions be handled? Acknowledging these limitations does not weaken the measurement; it makes it more honest. False precision is often more seductive than rigor, but it is also more fragile.

And yet we developed precise solutions for every one of these particularities.

Above all, I understood that an accurate figure is not enough to create trust. Trust is built beforehand through the consistency of the protocol, the perceived independence of whoever is measuring, and the transparency of the process. When stakeholders think the thermometer belongs to one of the camps, they dispute the temperature before they have even read it. The technical quality and social legitimacy of measurement are therefore inseparable.

We thought we had avoided that problem by surrounding ourselves with a group of media outlets representing different ideological leanings, but we had not. Our opponents reversed the logic: media equals bias. Each person would cherry-pick the outlet that did not match their ideology in order to disqualify the figure: “That particular outlet is part of the group… therefore…”

Finally, that experience transformed the way I present results. I now try more carefully to separate three levels: what was observed, what the method allows us to conclude, and the political interpretation that each person may then give it. The role of the person measuring is not to eliminate debate, still less to decide the political meaning of an event.

It did not lead me to relativism. Not all figures are equal, not all methods are equivalent, and the truth certainly does not lie halfway between two estimates—the organizers’ figure plus the Prefecture’s figure, divided by two. But it taught me that objectivity is never an acquired state: it is a demanding practice built on transparency, modesty, and the possibility of challenge. In short, the experience did more than change the way I explain measurement; it taught me that measuring also means organizing the conditions under which a result can be believed without demanding blind acceptance.

Yes. Most often, the poor decision does not stem from a false indicator, but from an accurate indicator being asked to answer the wrong question.

For example, I have seen organizations conclude that communication was effective because its visibility had risen sharply. The number of media mentions, share of voice, or impressions increased: the dashboard was green. Yet the organization was becoming more visible for the wrong reasons, or its message remained misunderstood. The measurement did indeed show an increased presence in the public sphere, but it was made to mean that reputation had improved. Visibility measures exposure, however, not necessarily consideration, and still less trust.

The same phenomenon exists with digital engagement. A heavily commented-on piece of content may be deemed successful even though it primarily generates anger, mockery, or polarization. The indicator records the intensity of the reaction, while the organization may interpret it as support. It can then repeat a style of communication that maximizes short-term attention while gradually damaging the quality of its relationship with its audiences.

I have also seen internal surveys lead to poor decisions when too much meaning was attached to an overall average. An apparently stable engagement score can conceal the collapse of a strategically important group, wide disparities between professions, or silent resignation. By looking only at the aggregate figure, the organization reassures itself at precisely the moment when it should be concerned. The average is not false; it is simply incapable, on its own, of describing the structure of the problem.

I often speak of “data telling”: telling the story of a situation through a set of figures rather than through a single figure.

The danger therefore appears when an indicator ceases to be a signal and becomes a verdict. Its conditions of production, its margin of interpretation, and above all the phenomena it does not measure are then forgotten. A Net Promoter Score does not summarize the entire relationship between consumers and a brand. An open rate does not tell us whether a message was understood. Advertising value equivalency does not measure the true value of media coverage.

A sound decision requires each indicator to be placed within a chain of causality: what did we do, whom did we actually reach, how was the message received, what did it change, and what other causes could explain the observed result? The further we move from the communication activity toward an economic or behavioral effect, the more difficult attribution becomes. Presenting correlation as causation often gives a leader comfortable certainty, but it is methodologically fragile.

I would even say that some indicators are dangerous precisely because they are simple. They give the impression that complexity has been mastered. Yet the role of a dashboard is not to replace judgment, but to discipline it. It should force us to ask better questions, not artificially provide a definitive answer.

The question to ask when faced with any indicator is therefore not so much, “What does this figure say?” as, “How far are we entitled to make it speak?” A measurement becomes misleading not because it is imperfect—they all are—but because the organization forgets the boundary between what it actually establishes and the narrative built around it.

An imperfect measurement is useful only if its imperfection is known, made explicit, and proportionate to the decision it is intended to inform. An approximation presented as such can be valuable; an approximation disguised as certainty can be more dangerous than the absence of a figure.

Measurement creates a reality effect. As soon as a dimension is given a score, it appears more objective, more comparable, and often more important than dimensions that do not have one. The risk is therefore not only measuring badly, but shifting the organization’s attention toward what can easily be counted. A fragile indicator may then legitimize a decision, prematurely close a discussion, or marginalize experiences that are nevertheless essential.

Like Sherlock Holmes, our profession sometimes involves gathering a body of clues even though we do not possess absolute proof.

More generally, almost anything can now be measured, including the emotions elicited by an advertising film. We can also conduct eye-tracking studies that measure and record eye movements in real time, for example. The question then becomes: “What for?”

But beyond “Can we measure everything?” there is also a moral question: “Do we have the right to do so, and at what human cost?”

I therefore prefer to distinguish between what should remain unquantified and what should remain unevaluated. Very few realities should escape every form of evaluation. But evaluating does not always mean assigning a score. It can involve interviews, observations, reasoned debate, case studies, or informed collective judgment. Qualitative inquiry is not incomplete measurement; it is sometimes the fairest method for approaching what numbers would oversimplify.

Before producing an indicator, we should ask three questions: what decision is it meant to inform, what part of reality will it necessarily reduce, and what might it transform merely by existing? If the answers are unsatisfactory, refraining from measurement can be a form of rigor.

The right maxim is therefore not “It is better to measure imperfectly than not to measure,” but rather: “Acknowledged ignorance is better than false numerical knowledge.”

As in the hard sciences: knowing what we know and knowing what we do not know.

The point is never to quantify everything; it is to know when a number illuminates, when it simplifies usefully, and when it begins to blind us.

The experience is unsettling because it separates things we were accustomed to holding together: words, a voice, a face, and a presence. The avatar can reproduce the outward signs of my embodiment with great fidelity even though I am not in front of the camera and, at times, did not directly originate every sentence being spoken.

I would first draw the line at intention. When the avatar delivers a text that I have written, approved, and intended for a specific context, I still regard it as technical delegation. Like a book, a recording, or a video, it allows my words to circulate in my absence—and sometimes in languages I do not speak at all. The technology changes the scale and realism of that delegation, but not necessarily its nature.

The situation changes when the avatar begins producing answers that I have neither formulated nor explicitly approved. It can then speak “like me” without truly speaking on my behalf. It statistically extends my positions, vocabulary, and reasoning, but plausibility does not amount to intention. A machine can imitate a person’s coherence; it shares neither their awareness of the situation, nor their hesitations, nor the consequences they will personally have to bear.

For me, that is where the essential limit lies: embodiment is not only a matter of appearance or voice. It involves being “fully present” in a situation, being affected by what is at stake, and being able to change what one says through contact with the other person. In a genuine conversation, I do not merely deliver content: I listen, perceive a reaction, gauge the effect of my words, and can reconsider, clarify, or even decide to remain silent. An avatar can simulate these adjustments; it does not experience them.

Embodiment also involves an element of risk. When I speak in person, I put my reputation, my relationship with the other person, and sometimes my own convictions on the line. I am not exactly the same after certain conversations because they can change me. An avatar can create an appearance of presence without experiencing that vulnerability. Yet speech without vulnerability can quickly become assertion without responsibility.

The boundary must therefore also be explicit. Viewers or listeners must know that they are encountering an avatar, distinguish what was approved in advance from what was generated, and understand how far that representation is authorized to speak on my behalf. The more autonomy the avatar has, the more essential traceability, the right of control, and the possibility of withdrawal become.

To be transparent, I have not yet experimented with giving my avatar that degree of autonomy. I do not know whether it is technically possible today.

I therefore regard my avatar as a limited proxy, not as a double. It can convey, explain, repeat, translate, or make accessible content whose authorship I assume. But it should not decide on its own what I think in a new situation, make a commitment in my place, or turn an algorithmic probability into a personal position. The intersection of agentic AI and avatars will raise precisely this issue: the creation of digital chimeras.

Ultimately, I think what remains to us is the science of questioning: questioning the world and questioning ourselves.

When it comes to answers, the contest is lost. We have been surpassed.

That is, incidentally, one of the benefits of my philosophy studies: questioning lies at the heart of knowledge.

More precisely, as the cost of producing content tends toward zero, as you suggest, value shifts toward discernment. The communications professional will no longer simply be someone who knows how to make a message, but someone who can determine whether that message is necessary, legitimate, timely, and worthy of the attention it will demand.

For a long time, scarcity lay on the production side: writing, filming, illustrating, or adapting content required time, skills, and resources. Artificial intelligence reverses that economy. It makes it possible to produce dozens of versions, in every format and for every audience, almost instantaneously. But it creates neither additional time to receive them nor additional attention to understand them. The more abundant content becomes, the scarcer human attention becomes.

The dilemma, then, is who creates value—and who will pay for that value?

If production is no longer worth very much, are we prepared to pay much more for ideas?

An organization can now feed its social networks without interruption, personalize its newsletters, multiply its public statements, and respond to every trend. But this hyperactivity can create a form of communication pollution. Every additional piece of content competes not only with content from other organizations, but also with the organization’s own genuinely important messages.

The communications professional’s responsibility therefore increasingly consists in protecting the hierarchy of meaning.

Not everything deserves a campaign, not every event deserves a public statement, and not every news story requires a brand to express an opinion. Knowing when not to produce can preserve the credibility of something said in the future. An organization that speaks constantly risks no longer being heard when it truly has something to say.

The communications director must sometimes also become the Director of Silence.

That choice of silence must not, however, be confused with inaction or excessive caution. Not communicating is also a decision that can have consequences. The aim is not to make silence an absolute virtue, but to recover genuine editorial rigor. Why are we speaking? To whom? What useful effect are we trying to produce? And what would actually happen if this content did not exist? That final question will probably become one of the most important in our professions.

Artificial intelligence will also make an essential distinction between production and authority more visible. A machine can generate impeccable text, but it cannot decide on its own whether the organization has the right to say it. It can imitate empathy but cannot guarantee that words correspond to actions. It can propose a position, but it bears neither the reputational risk, the moral responsibility, nor the political consequences of that position. The communications professional will remain the person who connects the message to an intention, a context, and a responsibility.

Their value will therefore lie less in the ability to add content than in the ability to make choices: choosing what deserves to be said, what should be reformulated, what requires a human presence, what can be automated, and what should remain unsaid. They will become more of an editor, arbiter, and guardian of coherence than a mere producer.

But I do not believe the ability to produce will disappear. To choose well, one must have an intimate knowledge of how content is made and understand words, images, narratives, and their effects. AI may generate a thousand possibilities, but someone will still have to recognize the right one—or understand that none of them should be published.

The value of the communications professional will ultimately lie in a new form of scarcity: no longer a scarcity of content, but a scarcity of judgment. In a world where anything can be produced, the decisive skill may be knowing what deserves to exist.