Against roadmaps, briefly
Why a list of dated promises is usually the wrong artifact.
A roadmap is a list of dated promises about a future you cannot see. Most of the work that turns out to matter is not on it; most of what is on it turns out not to matter. The roadmap survives because it is easy to make and easy to read, not because it is accurate.
What you actually want is two simpler artifacts: a register of open questions, ordered by how badly an answer is needed, and a log of what was done last week. The first orients the work; the second proves it moved.
A roadmap pretends to do both jobs and does neither well. It is also much harder to throw away than it should be, because by the time it is clearly wrong, several people have already pointed at it in meetings.
The remedy is not a better roadmap. It is fewer of them.
A small caveat
There is one case where a roadmap is the right artifact: when an outside party — a customer, a regulator, a partner — needs a calendar in order to plan against you. In that case the roadmap is a contract, and it should be written and reviewed as one. It should not be confused with the internal document that decides what to work on next.