The Steering Framework

Why Steering Exists

Most interfaces just show data. A steering interface lets a person act on it. They can see where the data came from, change the assumptions behind it, and watch the consequences play out, all without losing trust in what they are looking at.

Steering is built to be data-agnostic. The same interface should hold up whether the data is climate, finance, logistics, epidemiology, energy, or operations, because the underlying job never changes. Take any dataset, show its present state honestly, and let a person explore where it could go next. Nothing in the method is tied to one domain. The interface adapts to whatever data it is given.

This matters most in an interface that does something no dashboard does. It shows the present and projects the future in the same view, and that carries a real burden: it asks people to believe something that has not happened yet. The harder the model works, the more it can start to feel like a black box, and a black box predicting an outcome is exactly the kind of thing people either distrust completely or, worse, trust blindly.

So the real design problem is not visualization. It is trust, and steering is the method for earning it, for any data.

↑ Back to contents

The Core Principles

Steering rests on five principles. Each one answers a question the user is quietly asking, and an interface earns trust only when it can answer all five.

Legibility "Where does this come from?"

Every value carries its provenance. At a glance the user can tell whether a number is measured, interpolated, or projected, with its source, recency, and uncertainty only an interaction away. Nothing is dressed up as fact when it is really an estimate.

• Show the source and date behind any value, on demand.
• Treat uncertainty as a feature rather than an embarrassment, with confidence bands on projections and error ranges on measurements.
• Let the model's key assumptions be inspected in plain language, with the technical detail sitting underneath for anyone who wants it.

Control "What if I change an assumption?"

The future is steerable, not fixed. When the user adjusts the inputs that drive a projection, the interface re-runs in plain view, so they shape the outcome instead of passively receiving it.

• The core levers, meaning the assumptions that matter most, are directly adjustable.
• Every change is framed as a what-if, never as a rewrite of the truth, and the baseline always stays in view.
• Constraints stay honest. If a combination of inputs is implausible, the interface says so rather than quietly producing nonsense.

Reversibility "Can I get back to where I was?"

Exploration has to feel safe. The user can experiment freely because nothing they do is destructive and the starting point is always within reach.

• Every scenario is saved, named, and easy to return to.
• A reset to the baseline is always one click away.
• Scenarios can be set side by side, including against the measured present.

Calm "Is this alarming me or informing me?"

Heavy data, whether it is climate, health, or financial risk, can frighten people or leave them numb. The interface has to inform without manufacturing panic. That is an ethical responsibility, not only an aesthetic one.

• Color and motion stay restrained, and intense signals are saved for a genuine signal rather than decoration.
• A predictable, consistent interaction grammar keeps the tool feeling stable even when the data is grave.
• No dark patterns and no doom-engineering. Severity is shown in proportion, and the data is left to speak for itself.

Authorship "Is this scenario mine, or theirs?"

The future a user builds belongs to them. They can keep it, name it, and pass it on, and a shared scenario carries its assumptions with it, so it stays legible and honest in someone else's hands. The person remains the interpreter. The interface informs, it never decrees.

• Saved scenarios are credited to the user who built them.
• Sharing a scenario also shares its assumptions, so two people can agree on what it actually represents.
• The interface offers projections and leaves the conclusion to the person.

How the principles relate

These are not a checklist of equals. They build on one another. Legibility is the foundation, and without it nothing else can be trusted. Control and Reversibility turn the interface into a thinking tool. Calm governs how all of it feels, and Authorship is what the user walks away with. Read together they make a single promise: you can understand it, steer it, undo it, sit with it calmly, and call the result your own.

↑ Back to contents

Multiple Levels of Data in One Visual

A steering interface does not scatter data across a wall of disconnected charts. It composes several kinds of meaning into one living view, the composite visualization, so a person reads them together, on a shared scale, at the same time. Because the layers are assigned by data type rather than by domain, the same composite works for any dataset you feed it.

Three base layers carry three different kinds of meaning. A fourth layer projects the future, and a fifth ties everything together.

Why one composite beats many charts

• Relationships that stay hidden across separate charts, like a category spiking at the exact moment a trend turns, become obvious once the layers share an axis.
• One shared scale removes the silent distortion that creeps in when you compare charts with different ranges.
• The user holds one mental model instead of stitching several together, which lowers the mental load. Attention is the scarce resource here, not screen space.

↑ Back to contents

The Timeline: Another Way to See the Path

The path a dataset takes, from measured past, through the present, and into projected futures, can be read in more than one way. The composite visualization shows that path in place. The timeline shows the same path along time, as a scrubber the user can move through.

Both are views of the same underlying path, and neither is the real one. The composite answers a single question: what does the whole picture look like right now? The timeline answers a different one: how did we get here, and where does each moment sit on the way to the future? Offering both lets a person pick the reading that fits the question in front of them.

What the timeline adds

• A movable now. The present sits as a marked point on the timeline, with measured history behind it and projected futures ahead, turning the present/future divide into a spot the user can stand on.
• Scrub to any moment. Dragging the cursor moves every layer to that point in time at once, so the user can replay how the path unfolded or step forward into the projection.
• The path as a shape. Stretching the trajectory along time reveals turning points, accelerations, and the moment uncertainty starts to widen, all of which are harder to read in a single composed frame.
• Compare moments. Two points on the timeline, say today's baseline against a projected future, can be held side by side, so the change across the path reads clearly as a change.

↑ Back to contents

Defining the Data That Changes the Path

At any moment, the user can define the input that changes the projected path. This is the heart of steering. The interface is not a finished picture to read, it is a live model to push on, and when you change an assumption a new future appears.

How it works

• The user selects or adjusts a parameter, whether by dragging a point, moving a lever, or editing a value right on the visual.
• The interface generates a new projected point on the path, clearly marked as modeled and kept distinct from measured history.
• The projection is wrapped in an uncertainty band, so it never reads as a hard fact, in keeping with the Anchor Rule.
• The new path sits alongside the original, so the change reads as a change rather than a silent overwrite.

Two ways the future is resolved

The same action produces a different kind of answer depending on what the user has selected, and the mode is always a visible, deliberate choice rather than something hidden.

↑ Back to contents

Switching Views: One Visual, Many Goals

Switching how you see the data has to be effortless, a single obvious action rather than a rebuild. At any moment the user can move from one view to another: composite to timeline, a comparison reading to a direction reading, an expert lens to a plain-language one. The underlying data never changes. Only the view does.

An easy way to switch

• One control, always in reach. A persistent view switcher lets the user move between views with a single tap or click from anywhere in the interface.
• State carries across. The current selection, time cursor, and active scenario follow the user from one view to the next, so switching never loses their place. It simply reframes what they were already looking at.
• No reload, no rebuild. Views are just different readings of one shared model, so switching is instant and reversible, and the user can flip back and forth freely to compare them.
• The switch is legible. It is always clear which view is active and what it emphasizes, so the change never disorients the user.

Beyond switching the form of the view, composite or timeline, the user can switch the goal of the view, the lens, without rebuilding anything. The data does not change; the lens does. A single composite can be read for very different purposes, because each goal simply re-weights which layer leads and what the interface emphasizes.

What a "goal" changes

• Which layer leads. The same data pushes the bars forward for a comparison goal, the trend line for a direction goal, or the heat map for a where-is-the-pressure goal.
• Which metric is the headline. The same scenario can be judged on cost, risk, time, or reach, depending on what the user is trying to decide.
• Which detail is surfaced or tucked away. An expert lens exposes assumptions and raw values, while a public lens leads with a plain-language summary.

Why the same data serves different people

A mixed audience, from analysts to the general public, does not need different products. It needs the same composite viewed through different goals. Progressive disclosure is what makes that work: everyone starts on one calm, legible surface, and depth unfolds only when a particular goal calls for it.

↑ Back to contents

The Steering Loop

The principles, the layers, the path-defining, and the goal-switching are not separate features. Together they form one continuous loop the user moves through.

1. Read the present. The measured, sourced state of the system, solid and trustworthy.
2. See the default future. The baseline projection, with its uncertainty made visible.
3. Switch the view. Move between the composite and the timeline, and choose the goal or lens for whatever you are trying to decide. The same data reframes instantly.
4. Define the change. Adjust the parameter that drives the path, and a new future appears, marked and banded.
5. Compare. The new scenario sits beside the baseline and the measured present, in whichever view you prefer.
6. Keep it. Save, name, and share the scenario, assumptions attached, as your own.

How to tell it is working

• A non-expert can tell the measured present from the modeled future within seconds. (Legibility)
• A user can change an assumption and understand what changed and why. (Control)
• A user explores freely and returns to the baseline without anxiety. (Reversibility)
• People leave informed rather than alarmed or numb. (Calm)
• Two people can look at a shared scenario and agree on what its assumptions were. (Authorship)

Steering turns a visualization of any data into an instrument, one a person can read honestly, see along time or in place, switch between views in a tap, steer deliberately, and trust enough to decide with.

↑ Back to contents