A roadmap is a promise about a future you cannot see. Most of the time you make it anyway, with dates, with confidence, with a slide that has little colored bars marching neatly into the next four quarters. Then reality arrives and the bars lie there looking embarrassed.
The problem is not that planning is useless. The problem is that we plan as if certainty scales with effort, as if staring harder at quarter three will bring it into focus. It will not. The future is foggy by nature, and a plan that pretends otherwise is not a plan; it is a forecast wearing a contract's clothes. This piece is about what to do instead: how to plan honestly when the future is uncertain, using rolling-wave planning and a simple now/next/later board that tells the truth about how much you actually know.
The roadmap that aged badly
Almost everyone has shipped a roadmap they stopped believing in by month two. It is one of the quiet, universal experiences of working life. You spend a week building a Gantt chart that stretches twelve months out, every task with a start date, an end date, a dependency arrow. It looks authoritative. It gets presented to the leadership team, who nod, because a chart that detailed must be backed by something solid. Then a customer changes their mind, a key hire slips, a dependency you forgot about surfaces, and the whole elegant structure starts shedding accuracy by the day.
The deep error is treating a guess as a commitment. Henry Gantt designed his chart in the 1910s for factory work, where the steps were known, repeatable, and physical: you genuinely could say how long it took to cut a sheet of steel. Knowledge work is the opposite. Half the time you are figuring out what the thing even is while you build it. Borrowing a tool built for certainty and pointing it at deep uncertainty does not make the uncertainty go away. It just hides it behind a precise-looking line.
And precision, when it is fake, is expensive. A roadmap with twelve months of detail invites everyone downstream to plan against month nine. Marketing schedules a launch. Sales promises a client. Finance models the revenue. Each of those becomes a small debt secured against a number you invented under uncertainty. When the number moves, and it will, you are not just replanning; you are renegotiating a web of promises that should never have been made at that resolution.
The cone of uncertainty
There is a well-known idea in software estimation called the cone of uncertainty, and it deserves to be better known outside it. The insight is simple and almost obvious once stated: your ability to estimate anything improves as you get closer to it. Early in a project, an estimate might be off by a factor of four in either direction. By the time you are halfway through, the range has tightened dramatically. The error does not vanish, but it narrows as knowledge accumulates.
The practical lesson is to let plan detail mirror the cone. If your confidence about month nine is genuinely four times worse than your confidence about next week, then your plan for month nine should be roughly four times coarser. Anything sharper is decoration. The cone is not an excuse to stop planning; it is an instruction about resolution. Near things get drawn in pencil with fine lines. Far things get drawn as fuzzy regions, and that fuzziness is the honest part, not a failure to try hard enough.
Rolling-wave planning: detailed near, coarse far
Rolling-wave planning is the discipline that takes the cone seriously. The name comes from project management literature, and the picture it paints is exactly right: a wave of detailed planning that rolls forward with you in time. You plan the immediate horizon in fine grain, the middle distance in medium grain, and the far horizon in broad strokes. As time passes and the fog clears, the far horizon moves into the middle, the middle moves into the near, and you add detail at the moment you actually have the information to add it well.
This is not laziness dressed up as method. The near-term plan is genuinely rigorous: named tasks, owners, estimates, dependencies, the works. The difference is that you refuse to invent that same rigor for work you cannot yet see clearly. A common shape is to keep the next two to four weeks at task level, the next quarter at the level of grouped initiatives or epics, and anything beyond that as a handful of named themes with no internal structure at all. Taiichi Ohno and the Toyota Production System made a related point decades ago: producing detail before it is needed is a form of waste, and a fully specified plan for distant work is inventory that will spoil before you use it.
The cadence is the engine. Rolling-wave planning only works if the wave actually rolls. That means a standing rhythm, say every two weeks, where you pull the next slice of the future into focus, break the nearest theme into real tasks, and let the far edge stay deliberately vague. Skip the cadence and you are back to a static roadmap, just one with blurry edges. The recurring act of re-planning is the whole point; it is where new information enters the plan instead of bouncing off it.
The now/next/later board
If the Gantt chart is the symbol of false certainty, the now/next/later board is its honest cousin. It is the simplest planning artifact I know that actually respects the cone. Three columns. Now holds the small number of things you are genuinely working on. Next holds what you have committed to pick up soon, shaped but not yet started. Later holds everything else: the ideas, the maybes, the someday work, captured so it is not lost but explicitly not scheduled.
The genius of the board is what it refuses to say. It never claims that the third item in Later will ship in week thirty-eight. It does not assign dates to things it cannot date. By dropping the false precision, it earns trust where it matters: when something is in Now, people believe it, because Now is small and real. Compare that to a roadmap where everything has a date and therefore nothing's date is believed. The board trades the illusion of total knowledge for credibility about the near term, which is the only term you actually control.
It also makes priority a verb instead of a document. Moving a card from Later to Next is a visible, deliberate act. It forces the conversation, "are we really doing this before that?", to happen at the moment of decision rather than being buried in a chart nobody re-reads. Stakeholders can see the whole field of possibility in Later without being promised any of it, and that turns out to defuse a lot of anxiety. People do not actually need a date; they need to know their thing has not been forgotten and to understand what stands ahead of it.
How to run it without losing the plot
Plan the cadence, not just the work. Pick an interval and protect it. Every couple of weeks, refill Now from Next, reshape Next from Later, and prune Later of things that no longer matter. This recurring re-plan is where the rolling wave rolls; it is the most important meeting you will keep skipping if you let yourself.
Keep Now genuinely small. A board with eleven things in Now is a to-do list lying about being a board. Limit it, the way a kanban system limits work in progress, so that Now means in progress, not in theory. The discomfort of having to finish something before starting the next is the feature, not the bug.
Give Later a memory and a graveyard. Capture everything so nothing is lost, but periodically delete. An idea that has sat untouched for six months while everything around it changed is rarely the idea you should resurrect; it is more often clutter posing as optionality.
Resist the request to add dates back. Someone, usually senior, will ask for the timeline. The honest answer is a range tied to the cone: near work, here is a tight estimate; far work, here is a theme and a rough quarter, and we will sharpen it when it reaches Next. Hold that line. Kahneman and Tversky showed how badly the planning fallacy bends our forecasts toward optimism; the antidote is not more confident dates but more honest ones.
The point of planning is not the plan
The goal was never a beautiful chart. It was to make good decisions under uncertainty and to coordinate people without lying to them. A rigid roadmap fails at both the moment the world moves, which is immediately. Rolling-wave planning and a now/next/later board fail more gracefully, because they were built to bend.
So draw the near future in pencil and the far future in fog, on purpose. Commit hard to a few things, loosely to the next few, and not at all to dates you cannot honor. The fog will not lift because you demanded a timeline from it; it lifts as you walk into it, one cleared step at a time. Plan like someone holding a lantern, not someone holding a crystal ball, and you will be wrong far less often, and far less expensively, than the person with the gorgeous twelve-month chart.