When data has not yet made its case

In Dante’s Peak, the volcano begins to speak long before it erupts.

Small earthquakes. Changes in the water. Dead animals. Affected vegetation. Temperature variations and, later, an increase in seismic activity and gases. None of these signs, on their own, proves that the volcano is about to awaken.

That is precisely the problem.

They are what we know in strategy and foresight as weak signals: early, incomplete and ambiguous information that may anticipate a significant change, but whose meaning is not yet clear.

They constantly appear in organisations. A metric that begins to behave oddly. An unexpected request from a customer. A new use of the product. Several people who, separately, identify the same problem. An apparently peripheral technology that begins to appear more and more frequently.

Each signal may have a perfectly reasonable explanation. It may be an exception, noise or simply something unimportant. The problem arises when explaining them separately prevents us from asking whether, together, they are beginning to tell a different story.

Because major transformations rarely begin with an unambiguous alert on a dashboard.

Change usually begins before we have a metric capable of naming it.

Seeing data does not mean knowing how to read it

In Dante’s Peak, several people can observe the same phenomena and yet not reach the same conclusion. The data is there. What changes is how it is interpreted.

Harry Dalton does not have access to secret information. What he brings is experience, context and the ability to relate what he is seeing to patterns he already knows. An anomaly stops being merely a data point when it can be connected to others and acquire meaning.

We could represent it like this:

data → context → pattern → interpretation

Yet in many organisations, we continue to act as though the process were much shorter:

data → response

We build dashboards, analyse user behaviour, collect feedback, conduct research and, increasingly, use AI systems capable of processing enormous quantities of information. All of this expands our capacity to observe. But observing more does not necessarily mean understanding better.

A dashboard can show us that a metric has fallen, but not why. Research can identify a recurring behaviour without yet explaining what causes it. An AI can find correlations or summarise thousands of responses without necessarily knowing which ones should change how we understand the problem.

Value emerges when we are able to place that data in context, relate it and build an interpretation that we can test.

Because accumulating information can help us see more things. But more information does not necessarily produce greater understanding.

When several signals begin to tell the same story

One anomaly may be noise. Two may be as well. But there comes a point when signals from different places begin to show a certain coherence.

In Dante’s Peak, the earthquakes, changes in the water, affected vegetation or gas emissions matter not simply because they accumulate. They matter because they begin to fit together. What seemed like independent phenomena begins to form a pattern.

This is where pattern recognition comes into play: the ability to identify meaningful relationships between elements that, viewed in isolation, might seem disconnected.

It happens constantly in an organisation. The product team detects a change in user behaviour. Sales begins to hear an objection that had barely come up before. Customer support receives new questions. Qualitative research uncovers an expectation the product was not designed to meet. Separately, each team may interpret its own findings as a local phenomenon.

The interesting question arises when we bring those observations together:

What if they are not different problems?

One of the difficulties organisations face is that signals often appear scattered across departments, tools and responsibilities. Each team has part of the reality and good reasons to analyse it from its own context. But important changes do not have to respect the structure of our organisational chart.

That is why recognising a pattern is not merely about finding similarities. It also requires questioning the boundaries through which we have organised information and asking ourselves when we should stop explaining each anomaly separately.

Because sometimes, understanding what is changing begins when we discover that several apparently different signals are telling the same story.

Harry Dalton could be wrong

So far, it is tempting to think that Harry Dalton is simply seeing before everyone else something that will eventually become obvious. We also have an advantage: we know we are watching a film called Dante’s Peak. We expect the volcano to erupt.

The characters do not have that certainty.

Harry has reasons to worry, but his superior, Paul Dreyfus, also has reasons to call for caution. Unusual volcanic activity does not necessarily mean that an eruption will occur. And making a decision such as evacuating a town is not without consequences either: it has economic and social implications, causes alarm and may erode trust if nothing ultimately happens.

That nuance changes our reading of the film considerably. The conflict is no longer between someone who is right and someone who refuses to listen to him, but between two ways of managing an uncertainty that cannot yet be resolved.

Organisations constantly face similar decisions, though usually without a volcano involved. When do we invest in a technology that we still do not know will become established? When do we alter a product because of emerging user behaviour? When does a sustained drop in a metric stop being a fluctuation and require a change in strategy?

Acting too early has a cost. Acting too late does too.

That is why waiting until we have all the data is not always the most rational option. Sometimes, when the evidence is sufficient to eliminate uncertainty, part of our ability to respond has also disappeared.

The challenge is to define our decision threshold: what combination of evidence, risk and impact justifies acting even though we may still be wrong.

Because the criterion is not about achieving certainty. It is about deciding what level of uncertainty we are willing to accept before acting.

The problem is not only interpreting. It is getting an organisation to interpret

Harry can observe the signals, relate them and build a hypothesis. But that does not mean the organisation will automatically reach the same conclusion.

Between an individual interpretation and a collective decision, new variables emerge: hierarchies, responsibilities, economic interests, reputation, communication and, of course, disagreement between specialists. It is no longer simply a matter of asking what the data says, but of constructing a shared reading of what is happening.

This is where organisational sensemaking comes into play: the process through which an organisation tries to make sense of an ambiguous situation in order to decide how to respond.

And this has an important consequence. A good organisation is not one in which the expert always wins the argument. Harry may have experience and correctly identify a pattern, but that does not automatically make his interpretation the only possible one. As we have seen, he could also be wrong.

What matters is that there is room to put interpretations to the test: to test hypotheses, bring in different perspectives, challenge existing explanations and determine what new evidence might confirm or weaken each reading.

The problem arises when the structure does precisely the opposite: when a signal is dismissed because it comes from the wrong department, contradicts an earlier decision, threatens certain interests or simply does not fit the story the organisation has already told itself about itself.

At that point, the risk is not a lack of information. It is having built an organisation incapable of discussing what that information might mean.

Because interpreting well should not depend on finding the person who is right, but on creating the conditions for an interpretation to be examined before it is accepted or dismissed.

When everyone agrees, it is already too late

Eventually, the volcano erupts.

At that point, disagreement disappears. There are no longer signals to interpret, hypotheses to test or decision thresholds to discuss. Everyone observes the same reality and reaches the same conclusion.

Certainty is at its highest.

The scope for decision-making, however, has narrowed drastically.

This is perhaps one of the most interesting paradoxes of Dante’s Peak: when it becomes easier to know what is happening is also when there are fewer options left for responding.

Something similar happens in organisations. When a technology has already transformed an industry, when consumer behaviour has evidently changed, when a competitor has established a new model or when a cultural problem begins to show up in turnover, results and the loss of talent, reaching consensus becomes much easier.

But by then, we are no longer deciding whether to anticipate change. We are trying to adapt to it.

This does not mean we should react to every signal or constantly try to predict the future. It means accepting that some of the most important decisions will have to be made while disagreement, incomplete information and a real possibility of being wrong still exist.

Because there is an uncomfortable relationship between certainty and the ability to act: the greater the certainty, the narrower our scope for decision-making may be.

Waiting until everyone agrees may feel reassuring. The problem is that, by then, reality may already have decided for us.

Conclusion

Dante’s Peak was released in 1997. Almost thirty years later, some of its special effects have aged and several of its scientific liberties are obvious. Yet the dilemma it presents remains surprisingly relevant.

Today, we have access to an amount of information that Harry Dalton could scarcely have imagined. Real-time dashboards, advanced analytics, user research, predictive models and artificial intelligence systems capable of finding patterns across enormous quantities of data.

And yet, we still face the same question:

What does what we are seeing actually mean?

Perhaps that is why judgement remains so important. Not because it allows us to predict the future or eliminate uncertainty, but because it helps us decide what deserves attention when we do not yet have all the answers.

Sometimes, the most important signal is not even the appearance of something inexplicable. It is discovering that the explanation we have used until now no longer sufficiently explains what we are seeing.

That is where the difficult work begins: observing, relating, interpreting, testing and deciding while knowing that we may still be wrong.

Because the value of judgement exists precisely before certainty.

What signals are we dismissing today because we can still explain them separately?

When reality makes the answer obvious, perhaps we will no longer be interpreting signals. We will be managing consequences.