The Roadmap as a Negotiation
Overview
Section titled “Overview”You have a sequence justified by reversal cost, learning value and capacity. It is defensible, and it is about to meet several people who want a different one.
They are not being obstructive. Each is right about their own driver, and order is the one dimension where their drivers cannot all win.
Order is where the conflict goes
Section titled “Order is where the conflict goes”Scope conflicts are negotiable by addition — both features, eventually. Order conflicts are not. Someone is in month three and someone is in month nine, and no amount of goodwill changes the arithmetic.
This is why stakeholders who agreed enthusiastically about the destination become difficult about the plan. Nothing changed in their position; the roadmap simply made the trade-off visible for the first time. That is the roadmap working correctly, not a sign you presented it badly.
graph TD
P["Product lead<br/><i>revenue feature first —<br/>the date is promised</i>"] --> N{"One sequence<br/>has to ship"}
C["Compliance lead<br/><i>boundaries agreed first —<br/>or nothing ships at all</i>"] --> N
E["Engineering lead<br/><i>nothing before the<br/>migration milestone</i>"] --> N
N --> A["Agreed sequence<br/>plus what each one gave"]
classDef s1 fill:#E8A33D,stroke:#C4832A,stroke-width:2px,color:#ffffff
classDef s2 fill:#D9695B,stroke:#B4483A,stroke-width:2px,color:#ffffff
classDef s3 fill:#5B8FB9,stroke:#4A7A9E,stroke-width:2px,color:#ffffff
classDef gate fill:#7C7C7C,stroke:#5E5E5E,stroke-width:2px,color:#ffffff
classDef out fill:#7D9A4E,stroke:#5F7A38,stroke-width:2px,color:#ffffff
class P s1
class C s2
class E s3
class N gate
class A out
Three defensible orders, one sequence. The olive box is the deliverable — and the second half of its label is the part that makes it hold.
Present consequences, not a plan
Section titled “Present consequences, not a plan”The move that changes these conversations: stop presenting the sequence as a recommendation and start presenting it as a set of consequences.
A recommendation invites a verdict — approve or reject, and rejection leaves you nowhere. A consequence table invites a trade, because each party can see what they get and what they give under each ordering, and can argue about the thing they actually care about.
The format is one row per candidate ordering and one column per stakeholder, with what each gains and loses. It is unglamorous and it works, because it moves the discussion from “is the architect right?” to “which of these do we collectively prefer?” — a question the room is qualified to answer and you are not.
Three techniques that make it hold
Section titled “Three techniques that make it hold”Make the displacement visible. Carry the displacement statements from the previous topic into the room. The person whose work slips must hear it from you, in advance, in specific terms. The alternative is that they learn it later and conclude you went around them.
Offer two orderings, not one. A single option is a decree, and people resist decrees regardless of merit. Two viable options with honest consequences is a decision the room can own — and both of them can satisfy your constraints, because you built them.
Put the review date in writing. The most useful sentence in a contested roadmap is a date on which a deferred item is reconsidered. It converts “no” into “not this quarter, and here is when we look again”, which is both true and much easier to accept. It also obliges you to actually revisit it, which is how the instrument keeps its credibility.
The parallel-track illusion
Section titled “The parallel-track illusion”The failure mode specific to this conversation: under pressure from two stakeholders who want different things first, the roadmap quietly commits to both, in parallel, with the same engineers.
It resolves the meeting and guarantees the quarter. Neither track gets the attention it needs, both slip, and the displacement that was never named happens anyway — to quality, to the migration, or to both. The capacity arithmetic from the previous topic does not become false because a meeting ended pleasantly.
Naming this out loud is uncomfortable and it is the job. “We can do both, and here is what both looks like: each one takes twice as long and the migration milestone moves. If that is the trade we want, I will write it down.” Said plainly, with the arithmetic attached, it usually ends the fantasy without ending the relationship.
Feeds the workbook
Section titled “Feeds the workbook”Your roadmap should show the sequence and what each stakeholder gave to get it. A roadmap with no concessions recorded is one that was never negotiated, which means it will be renegotiated later without you.
→ These four topics were decisions you made. If they only work when you are in the room, you have built a dependency rather than a practice. — A repeatable design process