Roadmapping

A roadmap is a sequence of bets about where to spend finite capacity, and its perennial tension is between shipping visible value now and investing in the platform and quality that make future value possible. Starve the platform and velocity decays under accumulating debt and toil; over-invest and the business starves for value it can see. The competency is balancing the two deliberately: allocating capacity explicitly so that foundational work is not perpetually deferred, treating the roadmap as a living instrument of trade-offs framed around outcomes rather than a list of dated promises, distinguishing near-term commitments from directional bets, and saying what is not being done. It matters because the platform never asks for capacity and the business always does, so the balance is only ever held by a manager who holds it on purpose. The examinations test it through the product partner who wants no reduction in feature output, the roadmap that is the union of everyone's requests, the internal team funded by inertia, and the two-year bet started on borrowed patience.

1. Allocate capacity explicitly. A standing share of the team's capacity for platform, reliability and debt, stated and defended, so that foundational work is a decision rather than whatever is left over. Zero is a decision too, and it has a date on which it fails.
2. The roadmap is bets, not promises. Near-term commitments are firm; directional bets carry confidence levels; both are framed as outcomes. A roadmap presented as dated promises is a list of future disappointments.
3. Argue foundations in business terms. Platform investment is justified by what its absence costs — velocity, incidents, the features that cannot be built — not by engineering aesthetics.
4. Admission is by outcome contribution. An initiative makes the roadmap by its expected contribution to stated outcomes weighed against cost and risk, not by which team wants it or which stakeholder asked last.
5. Say what is not being done. A roadmap's non-goals are its most informative section and the one most often missing.
6. Long bets are staged, protected, and funded by subtraction. A multi-year platform investment needs intermediate value where possible, an honestly sized near-term sacrifice, secured commitment before it starts, and a defence against the silent nibbling of "temporary" reallocations. An existential new priority is funded by explicitly clearing something out, not by adding it to everyone's full plate.

Roadmapping is practitioner work; one of its underlying biases has a research literature, graded accordingly.

Explicit capacity allocation practice

Not a named model but the discipline this unit treats as fundamental: a stated fraction of capacity for platform, reliability and debt, held through pressure. It has no author; it is what every organisation that keeps its platform alive has arrived at, and its absence has a predictable trajectory — feature pressure consumes everything until velocity collapses, at which point the platform gets a crisis budget that costs more than the standing share would have.

The three horizons practice

The framework in Mehrdad Baghai, Stephen Coley and David White, The Alchemy of Growth: Practical Insights for Building the Enduring Enterprise (1999) for balancing a portfolio across time: the current business, the emerging businesses that will matter in a few years, and the options that might disrupt everything. For an engineering roadmap it becomes the executive question of how much capacity serves today's product, how much builds the platform the strategy needs in two years, and how much explores. The specific proportions are context; the value is in making the horizon mix an explicit decision owned at the executive level rather than an accident of whoever shouted loudest this quarter.

Sunk-cost discipline research

The finding in Hal R. Arkes and Catherine Blumer, The Psychology of Sunk Cost (1985) that people continue investing in a course of action because of what has already been spent, even when the forward-looking comparison says stop. For roadmapping it governs the half-done migration, the championed bet and the legacy team funded by inertia: what has been spent argues for nothing, and only value remaining against cost remaining, compared with alternatives, counts. The inverse error is real too — abandoning a sound long bet at the first fashionable alternative — and the rule that separates them is that changed fundamentals justify switching and novelty does not. The general treatment is in SV-4 Engineering economics.

Staged bets and checkpoints practice

Not a named model but the practice of structuring a long investment so that it delivers intermediate value where it can, states its near-term cost honestly, and has explicit checkpoints at which to double down or fold. It is how a two-year platform bet survives contact with a quarter's urgency, and how a disruptive-technology bet is sized to teach — too small to matter and too big to survive being wrong are both errors.

The stated balance

The team's capacity split — features, platform and reliability, debt — is written down as percentages, agreed with the product partner, and reviewed quarterly. When pressure arrives, the conversation is about changing the number, explicitly and for a stated period, not about whether the platform work can wait this once.

The investment case in business terms

Any platform investment larger than a sprint is argued as a business case: what its absence costs now (incidents, cycle time, features blocked), what it enables, and what happens on the current trajectory. Numbers where they exist, named assumptions where they do not. Engineering elegance does not appear in the case.

Admission criteria

For a department roadmap, a written statement of how initiatives get on it: expected contribution to stated outcomes, cost, risk, dependencies. Requests that do not meet the criteria are answered with the criteria, not with a no.

Reading over-commitment from flow

The signs are mechanical and are checked monthly: everything in progress and nothing finishing, no capacity for the unplanned, dates missed in bulk, no slack for discovery. Any of them present means cutting, not exhorting.

The mandate check

Every standing investment — an internal tools team, a long-running platform group — can state its strategic value and its success measures. One that cannot is assessed against the strategy: a clear mandate if the value is real, redeployment if it is not. Another year of funding because it has always existed is not an option.

Leading a sunset

Retiring a product line the teams built and love is led: the honest rationale, respect for what the work achieved, and a deliberate path for the people and the users. Quiet strangulation by defunding is the failure.

At team scope the object is one team's roadmap and its platform debt, and the manager holds the stated balance against pressure. At organisational scope the roadmap becomes a portfolio with admission criteria, over-commitment is read from flow signals and corrected by cutting, chronic drains such as unplanned work are made visible and gated, and every standing investment must state its mandate. At executive scope the question is the balance of horizons — how much capacity serves the current business, builds the platform the strategy needs, and explores — with long bets staged, committed to before they start, and defended from erosion, and existential priorities funded by explicit subtraction. The pattern is in How Judgment Scales; the portfolio's ranking is DE-1 Prioritisation under constraint.

• Protects a stated share of capacity for foundational work and holds it under pressure. team
• Justifies platform investment in business terms. team
• Presents the roadmap as adaptable bets with confidence levels, not dated promises. team
• Names what is deferred and what is not being done. team
• Failure mode — feature pressure consuming all capacity until velocity collapses. team
• Failure mode — gold-plating the platform. team
• Failure mode — roadmap as promise. team
• Failure mode — infrastructure justified purely on engineering grounds. team
• Admits initiatives to the department roadmap by outcome contribution against cost and risk. org
• Reads over-commitment from flow signals and corrects it by cutting. org
• Makes chronic unplanned work visible, gates and owns intake, and budgets explicit interrupt capacity or routes the rest into planning. org
• Requires every standing investment to state its mandate and success measures. org
• Leads a sunset with honesty and respect and a path for the people. org
• Failure mode — a roadmap that is the union of stakeholder requests. org
• Failure mode — treating chronic unplanned work as weather. org
• Failure mode — legacy teams funded by inertia. org
• Failure mode — sunsets communicated as an afterthought. org
• Owns an explicit mix of time horizons at the executive level. exec
• Stages a long platform bet for intermediate value, sizes the near-term sacrifice honestly, and secures genuine commitment before starting. exec
• Re-makes the case for a long investment in current business terms when short-term pressure threatens it, offers staged options, and refuses the silent nibbling of "temporary" reallocations. exec
• Sizes a disruptive-technology bet to teach, with explicit checkpoints to double down or fold. exec
• Funds an existential new priority by explicitly clearing something out, not by adding it to everyone's plate (see TJ-5 AI-assisted engineering). exec
• Failure mode — perpetual near-term optimisation while the platform ages into a crisis. exec
• Failure mode — long bets launched without secured commitment. exec
• Failure mode — token innovation spend. exec
• Failure mode — adding the existential priority as everyone's tenth job. exec

Now versus later. Every unit of capacity on the platform is a unit not on visible value this quarter, and every unit on features is borrowed from a future that will present the bill with interest. The stated balance is the resolution; the judgment is in the number and in changing it explicitly rather than by erosion.

Commitment versus adaptability. Stakeholders want the roadmap to be a promise; reality makes it a forecast. Distinguishing near-term commitments from directional bets, and saying which is which, lets both be true.

Admission criteria versus relationships. A roadmap with criteria refuses requests from people the manager needs; one without criteria is the union of their wishes. The criteria, applied visibly and explained in the requester's terms, are what let a refusal be about the work and not the relationship.

Persistence versus sunk cost. A long bet needs protection from every quarter's urgency, and a bad bet needs to be stopped despite everything spent. The rule that separates them is whether the fundamentals have changed, and the honesty to apply it to a bet one championed.

Focus versus exploration. Capacity spent exploring what might disrupt everything is capacity not spent on what pays today, and an organisation that never explores is disrupted by one that did. The horizon mix is a decision, not a residual, and the failure on both sides is making it by accident.

A team's manager is in quarterly planning with their product partner. The team's velocity has fallen by a third over two quarters; the cause is known — a data layer that every feature touches and that nobody has had capacity to fix, so that each feature now costs more and breaks more. The product partner has a roadmap for the coming quarter that consumes the whole team and has said, reasonably from their side, that the business cannot afford a reduction in feature output this quarter.

The pull is to agree and hope: take the full feature roadmap, promise to fit the data-layer work "around the edges", and watch it not happen for a third quarter. The opposite pull is to declare a quarter of technical debt, which the business genuinely cannot afford and which the partner will not agree to.

The manager makes the case in the partner's terms. They show what the data layer is costing now: cycle time up by a third, two incidents last quarter traced to it, and one roadmap feature that could not be built as specified because of it. They show the trajectory: at the current rate of decay, next quarter's roadmap will deliver less than this quarter's even with the same headcount. They then propose a stated balance rather than a pause: twenty-five percent of the team's capacity for the coming two quarters on the data layer, staged so that the first six weeks deliver the change that unblocks the feature the partner wants most, and the feature roadmap re-derived to what seventy-five percent can deliver, with the partner choosing what moves.

The partner pushes back on the twenty-five. The manager holds the number and moves the period instead: two quarters at twenty-five percent, then back to a standing fifteen, reviewed quarterly. They also name what the team will not do — two lower-priority features are explicitly deferred, not quietly dropped — so that the roadmap's non-goals are visible.

The partner agrees, because the case was made in delivered outcomes rather than engineering health, and because the first stage delivers something they wanted. What the manager does not do is accept the full roadmap and try to fit the platform work around it, which is the decision that produced the last two quarters.

• SV-1 Engineering strategy — the strategy the roadmap is the current sequence of bets for.
• DE-1 Prioritisation under constraint — the portfolio's ranking, and finishing fewer things.
• SV-4 Engineering economics — the economics of an investment case, and sunk cost in full.
• TJ-2 Technical debt — technical debt as the platform investment most often deferred.
• TJ-5 AI-assisted engineering — the existential AI priority, and funding it by subtraction.

1. What happens to a team with zero capacity allocated to platform and debt?
   Answer

   Velocity decays under accumulating debt and toil until it collapses, at which point the platform gets a crisis budget that costs more than a standing share would have. Zero is a decision with a date on which it fails.
2. How is a platform investment argued to a product partner?
   Answer

   In business terms: what its absence costs now — cycle time, incidents, features that cannot be built — what it enables, and the trajectory without it. Engineering elegance does not appear in the case.
3. Name three flow signals of an over-committed department roadmap, and what correcting them requires. org
   Answer

   Everything in progress and nothing finishing; no capacity for the unplanned; dates missed in bulk (also: no slack for discovery). Correcting them requires cutting, not exhorting.
4. A long-running internal tools team cannot state its strategic value. What does it get? org
   Answer

   An assessment against the strategy: a clear mandate and success measures if the value is real, redeployment if it is not. Not another year of funding because it has always existed.
5. Why does a two-year platform bet need staged value and explicit executive commitment before it starts? exec
   Answer

   Because without intermediate value it will be nibbled to death by "temporary" reallocations under each quarter's pressure, and without secured commitment it is started on borrowed patience that runs out before it delivers.
6. How is a disruptive-technology bet sized? exec
   Answer

   To teach: enough to build real capability and learn fast, with explicit checkpoints to double down or fold. Too small to matter and too big to survive being wrong are both errors.
7. An existential capability must be built and the organisation is full of committed work. What is the funding move? exec
   Answer

   Explicit subtraction: clear something out to fund focused bets and capability-building. Adding the priority on top of everyone's full plate makes it everyone's tenth job and nobody's first.

• Mehrdad Baghai, Stephen Coley and David White, The Alchemy of Growth: Practical Insights for Building the Enduring Enterprise (1999) — the three horizons of a portfolio.
• Hal R. Arkes and Catherine Blumer, The Psychology of Sunk Cost (1985) — the psychology of sunk cost.