Most data visualizations are about the past.

Even when they claim to show the future, they’re usually just extending a line forward — as if uncertainty behaves nicely, as if the world is predictable, as if a single trajectory is enough. It isn’t.

The moment you move from describing what has happened to what could happen, the rules change. You’re no longer visualizing data. You’re visualizing possibility.

And most of the conventions we rely on break immediately.

Look at almost any forecast chart: a clean line moving forward, maybe with a shaded band around it.

It looks precise. Controlled. Reassuring.

It’s also misleading.

That single line implies:

But real-world systems don’t behave like that.

In reality, futures branch instead of extend, uncertainty compounds over time, and outcomes are often non-linear and asymmetric.

Take hurricane forecasts. The cone is often read as the storm getting bigger, when it actually represents uncertainty in location over time. COVID projections had a similar issue. Many visualized a single curve, but what actually happened depended on behavior, policy, timing, resulting in completely different outcomes.

Alberto Cairo wrote about this misconception in his article for Nightingale, “The Day I Thought I Misled the President of the United States”, where he discussed how forecast visuals can unintentionally distort public understanding. The piece relates to an interactive hurricane forecast visualization also by Alberto Cairo for The New York Times, which explored alternative ways to communicate uncertainty more effectively.

When you’re dealing with predictions, simulations, or counterfactuals, you’re not working with a dataset — you’re working with a space of outcomes.

That changes the question from:

“What does the data say?” to “What could the data become?”

This introduces three core challenges:

Most forecast visualizations flatten these dimensions into a single trajectory because it’s easier to read, even when it’s less truthful.

Counterfactuals are especially difficult.

“What would have happened if…?”

These scenarios don’t exist in the data. They’re constructed. Hypothetical. Often model-dependent.

Yet they’re critical:

The challenge is that counterfactuals require you to show absence, a version of reality that never occurred. Most visualizations ignore this entirely or reduce it to a toggle, and that’s not enough.

If the problem is that we’re compressing possibility into a single path, the solution is to embrace structure over simplicity.

Here are three patterns that move in that direction.

Instead of a single confidence interval, uncertainty should be stratified.

Think in layers:

But more importantly, these layers should change shape. Because uncertainty isn’t uniform.

It might skew in one direction, split into multiple clusters, or collapse under certain conditions.

The goal isn’t to show “how much” uncertainty exists, but how it behaves.

Weather ensemble forecasts already move in this direction. Instead of showing one predicted path, “spaghetti plots” visualize dozens of possible outcomes simultaneously. The density and clustering of the lines reveal confidence, divergence, and instability far better than a single smoothed forecast line.

A single trajectory implies inevitability. A branching structure reveals decision points.

Instead of:

Think:

Each branch represents a condition, decision, and threshold being crossed. This makes the visualization interactive in a meaningful way where users don’t just explore data, they explore consequences.

This is especially powerful for policy simulations, climate scenarios, and model behavior under different inputs.

Because it shifts the narrative from:

“Here’s what will happen” to “Here’s what could happen depending on what changes”

Climate visualizations are one of the clearest examples of branching futures. Different emissions scenarios produce completely different warming trajectories over time. The future changes depending on policy, energy use, and collective behavior.

Instead of plotting individual futures, you visualize the space itself.

Imagine a 2D or 3D surface where each point represents a scenario, axes are variables (time, intensity, probability, etc.), and color or texture encodes outcome quality or risk.

This allows users to see regions of stability, zones of volatility, and sharp transitions between outcomes. It’s less about storytelling and more about exploration.

Earthquake hazard maps work similarly. Instead of predicting one specific event, they visualize regions of risk across combinations of magnitude, depth, and location. The goal isn’t to predict a single earthquake, but to understand the landscape of possible impact.

And it acknowledges something most visuals ignore: The future isn’t discrete, it’s continuous.

The hardest part of these approaches isn’t technical, it’s cognitive.

When you show multiple futures, users can feel overwhelmed since patterns become less obvious and Interpretation requires effort.

But simplifying the visual to make it “easier” often makes it wrong.

So the goal shifts from reducing complexity to structuring it.

This is where interaction matters:

You’re not just designing a visualization, but you’re designing a way to think about uncertainty.

If we want to design visualizations for things that don’t exist yet, we need to stop pretending they behave like things that do.

That means:

Because the goal isn’t to predict the future, but to help people understand how many futures are possible and what shapes them.