The value is in the sequencing rather than the list. Most technology decisions have dependencies, and doing them in the wrong order means redoing work. Customer records before marketing automation, because automation needs somewhere to read from. Conversion tracking before advertising spend, because otherwise you are buying unmeasurable clicks. Documented processes before hiring, because the tools serve the process rather than replacing it. Writing the order down surfaces those dependencies while they are still cheap to respect.
Keep the horizon to about a year and hold anything beyond six months loosely. Roadmaps extending three years are fiction for a business this size, and treating them as commitments produces resistance to changing course when the evidence says you should. What you are producing is a considered order of operations, not a contract with your future self.
Tie each item to a problem rather than to a capability. A line saying implement a customer system is a purchase. A line saying we lose track of enquiries that are not ready to buy, and this is how we will stop, is a problem with a proposed solution, and it can be evaluated afterward against whether it worked. It also lets you notice when a cheaper answer would do.
Include what you will stop or remove, which almost nobody does. Year one accumulates tools, subscriptions, and half used systems, and a roadmap that only adds is a roadmap that increases cost and complexity indefinitely. Naming what gets retired is frequently where the budget for the additions comes from.
Review it quarterly and expect it to change. The point is not adherence, it is that decisions get made deliberately rather than reactively, and that when something urgent arrives you can see what it displaces rather than simply adding it to everything else already underway.
Share it with anybody who works with you, including contractors, because the most common cause of duplicated or wasted technical work is somebody solving a problem that was already scheduled to be solved differently. A roadmap kept in the owner's head produces exactly that, and a single page shared costs nothing while preventing the situation where two partial solutions exist for the same problem.