Project Portfolio Management and Priorities

The manufacturing group’s investment committee had approved ninety-four technology projects over three years and rejected none. At the review date, nine were complete and eighty-five remained open: forty-one active and forty-four not yet started. Every proposal had a favorable business case, a sponsor, and a link to one of six broad strategic themes. The plant-data platform needed for the group’s proposed predictive-maintenance program had waited eleven months for staff. Its plan required four engineers at three-quarters of their time each. Each engineer already spent a quarter on each of three active projects and a quarter on continuing support: all capacity was allocated. The committee had never reviewed whether those commitments should change. When the new CFO asked what the group had decided not to do, the chair answered: we approve; we have never had to choose.

The group is fictional. Its approvals evaluated proposals separately without resolving their combined demands on money, skills, or affected teams. Richard Rumelt’s discussion of bad strategy emphasizes choosing where to concentrate resources and which goals to set aside. A portfolio review applies that discipline to a feasible set of investments. It examines strategic contribution, value, risk, dependencies, and capacity, then records which work proceeds, waits, changes, or stops. Approval can express that choice when it includes these comparisons; a favorable standalone case is insufficient.

Approval is not choice

The committee asked whether each proposal was worthwhile on its own terms. It also needed to compare what the same resources could support together. OMB’s 2013 Federal Enterprise Architecture Framework (FEAF v2) describes business classifications as support for portfolio alignment and the Enterprise Roadmap as a connection between strategy, projects, budgets, dependencies, and risks. These are historical federal planning examples. For this manufacturer, the relevant practice is to test the combined plan, including existing commitments and continuing operations. Project counts identify work to investigate; they do not, without effort and capacity estimates, prove overload.

Question	Approval process	Portfolio process	The group
Is it worth doing?	The standalone case clears a hurdle	Compare feasible alternatives and the work each would displace	Ninety-four approvals without an explicit resource comparison
Does it fit the strategy?	It cites a theme	Examine the proposed contribution, evidence, assumptions, and obligations	Broad themes did not establish which proposals addressed the need
Can we do it?	The sponsor expects capacity	Compare dated demand with skills and available capacity, including support and change work	Four engineers had no unallocated time
What depends on what?	Dependencies stay inside each case	Define readiness milestones and compare joint investment options	Plant rollout needed a validated data path; design could begin earlier
Should it continue?	Approval ends the investment review	Compare remaining choices using current evidence while preserving the original case	No continuation decision had been recorded

Strategic fit is a line of sight, not a citation

FEAF v2 uses line of sight to connect detailed investments and activities to agency outcomes. Its 2013 discussion of Exhibit 300 proposals asks for links to agency goals, target architecture, and performance measures. That scoped reporting example is not a statement of current requirements for every investment. The manufacturer can use a similar chain to investigate a proposal: what output it should produce, which outcome it is expected to affect, and which strategic objective it supports. Naming the capability involved helps locate the work. The Forty-Seven Initiatives and No Strategy supplies the diagnosis and intended direction; the proposal must explain and support its expected contribution. A written connection makes the claim inspectable, not proven.

Ask what evidence supports each link, what else must change, how large the contribution could be, and which alternatives might achieve it. Improvement of a weak capability is one purpose; preserving a critical service, extending an existing strength, reducing exposure, or learning about an uncertain opportunity can also justify investment. Use a measure suited to that purpose and retain uncertainty where an effect is not yet known. A portfolio can report traceability coverage, for example the share of the coming year’s proposed spending with an accepted rationale and measurement plan. State the denominator, period, evidence standard, and treatment of missing records. Coverage describes documented reasoning; it is not the percentage of spending that will create value.

Value, risk, and dependency are weighed together

SAFe’s weighted shortest job first (WSJF) divides relative cost of delay by relative job duration. The numerator combines user and business value, time criticality, and risk reduction or opportunity enablement. Job size can proxy duration only when staffing and delivery assumptions make that reasonable; it is not another component of delay cost. Intercom’s RICE instead multiplies reach over a defined period, impact, and confidence, then divides by effort in person-months. Both depend on evidence and estimates. The Vote Always Went to the Forty explains their use on a platform backlog. Neither formula alone chooses the best portfolio under all budget, capacity, and dependency constraints.

Separate a required outcome from a proposed way to achieve it. A regulation, support change, or security finding calls for assessment of scope, authority, deadline, and acceptable remedies. The distinction between mandatory gates and preferences prevents a preference score from excusing an unmet binding condition. It does not automatically approve a particular replacement project. Compare permitted remedies, such as an upgrade, supported alternative, narrower scope, or stopping an activity. Allocate resources so the selected response can meet the obligation. If the plan is infeasible, escalate the conflict to the appropriate authority; internal reprioritization cannot waive an external requirement.

A dependency constrains when a particular activity can proceed. Funding approval, development, testing, and production use may have different dependencies. The manufacturer’s plant rollout requires a validated data path, but consumer design and bounded trials can overlap with platform development. Compare the smallest useful foundation and its consumers as a package against alternatives, counting shared benefits and costs once. MIT CISR’s account of architectural maturity concerns developing reusable organizational capabilities; it does not establish a universal funding order for individual projects.

Scoring and selecting a set are different problems. In an illustrative fixed-budget choice, suppose independent, indivisible items A, B, and C cost 6, 5, and 5 units and deliver estimated values of 10, 8, and 8. With a budget of 10, choosing the highest value or value-to-cost ratio first selects A and yields 10; B and C together yield 16. This is a portfolio-selection example, not a refutation of WSJF under its sequencing assumptions. Record the decision objective, feasible combinations, uncertainty, and sensitivity instead of treating a sorted list as the decision.

Capacity and changing priorities also need separate evidence. DORA’s 2024 summary reports adverse relationships between unstable organizational priorities, productivity, and burnout. It does not establish that the number of concurrent projects or this committee’s approval rate caused those outcomes. Investigate priority changes, interruptions, rework, workload, and available support locally. Several stable workstreams and one repeatedly redirected project create different conditions.

Dimension	What it measures	How it is judged	Constraint or uncertainty to preserve	The plant-data platform
Value	Expected outcomes and consequences of delay over a stated horizon	Compare supported estimates, ranges, alternatives, and shared effects	Do not count the same enabled outcome for every contributing project	Evaluate the data path together with the maintenance uses it enables
Risk	Delivery, operating, transition, and non-action exposure	Keep owners and remedies visible; use justified quantitative estimates or scenarios where useful	A risk-adjusted score cannot waive a binding obligation	Assess novelty and alternatives to the proposed platform
Dependency	The capability needed at each milestone	Joint funding and overlapping development where feasible; readiness before dependent use	Do not confuse a production prerequisite with a ban on parallel design	Validate data quality and interfaces before each plant uses predictions
Capacity and change load	Demand by skill, team, individual, period, and affected group	Include support, training, coordination, and transition work; agree any reassignment	Budget and usable capacity both constrain the set	Four engineers need 75 percent each; retain their 25 percent support allocation

Balancing current commitments and future opportunities

Three horizons, managed at once

McKinsey’s account of the three horizons framework distinguishes the established core business, emerging growth businesses, and options for future growth. Its main planning lesson is to attend to all three concurrently rather than postponing future opportunities until current work is finished. These are growth horizons, not direct equivalents of run, grow, and transform technology budgets. Improving a current business can belong to horizon one; a shared platform may support several horizons. The table below is the manufacturer’s separate spending classification, chosen to expose present commitments and future investment.

Finance reconstructs the previous full fiscal year’s technology spending on one consistent cost basis, indexed to 100 units. It includes continuing operations and project work. The next-year plan also totals 100 on that basis. Each cost is assigned once to a primary spending purpose, with separate tags for shared capability, obligation, and supported growth horizon. The figures are invented for this case; obligations can occur in any category, and a project can have cost lines in more than one.

Bucket	What belongs in it	How it is judged	Typical failure	Previous actual / next-year plan
Run	Sustain agreed service, support, and control requirements	Compare acceptable operating arrangements and improvement options	Protecting an incumbent cost without checking alternatives	60 units / 55 units; planned reductions through validated retirement, right-sizing, and contract changes
Grow	Improve established business capabilities and processes	Compare contribution, cost, uncertainty, timing, and capacity	Approving every positive case without resolving combined demand	39 units / 35 units; cost scope includes existing and revised work
Transform	Explore or enable material changes to business capabilities and ways of working	Stage investment around assumptions, learning, and decision points	Committing to a large solution before its uses are validated	1 unit / 10 units; includes the initial data-platform increment in this case

Set allocations and revisit them with evidence

Leadership agrees a provisional allocation alongside required commitments and feasible investment options. It may reserve capacity for uncertain opportunities so short-term demands do not absorb everything, then revisit the allocation when evidence, costs, obligations, or strategy change. Compare marginal choices across categories as well as within them, with reasons and approval for material changes. The operating model informs requirements for shared processes and data, but does not calculate a funding percentage. Forty-Seven Initiatives and No Strategy explains why shared foundations, local initiatives, and renewal costs need explicit funding and counting rules. A lower run share is not itself a saving: if a total doubled from 100 to 200 while its run share fell from 60 to 55 percent, run spending would rise from 60 to 110. In this case the equal totals make the planned five-unit reduction visible, but actual savings still need verification.

Stopping is a decision with the same standing as starting

FEAF v2 applies lifecycle assessment to existing assets as well as new investment. For each project, preserve the original rationale, benefit hypothesis, assumptions, and review dates, using the evidence discipline in Live for a Year, Used by a Third. At review, compare that history with current findings and remaining alternatives. Continuing unchanged can be a sound decision. Completion, a smaller scope, a pause with restart conditions, or cancellation may be preferable when the evidence changes. Record what was considered and why, without requiring a minimum number of stopped projects.

Exclude past expenditure that cannot be recovered from the forward comparison, but include costs that differ between the remaining choices. Termination charges, retained records, handover, replacement services, restart effort, and effects on dependent work can matter. If continuation has a future net value of 20 units and switching offers 25 before an additional exit cost of 10, switching yields 15, so continuation is better on those assumptions. None of the past sunk spending enters that calculation. Name who retains operating and legal obligations, verify what resources can actually be released, and only then reassign them. GOV.UK’s legacy guidance identifies competing priorities as one barrier; a documented funding and review arrangement is a possible response, not a guarantee that a protected bucket solves every case.

Judging a portfolio

A useful portfolio review can show the rationale and evidence for its investments, the obligations it must meet, feasible resource allocations, and dependencies at the relevant milestones. It distinguishes planned spending from actual costs and observed outcomes from attributed benefits. Check that the latest evidence was considered, alternatives were compared, and decisions and reconsideration conditions are recorded. A justified continuation is as reviewable as a cancellation. Project counts, hurdle-rate results, and traceability coverage are useful records, but none alone establishes that the portfolio is valuable or feasible.

What a reviewer should ask

1. What outcome, obligation, or learning purpose does each item serve, and what evidence supports the expected contribution?
2. Which period and cost basis define the allocations, how are shared costs counted once, and who may revise them?
3. Which requirements bind, what are their deadlines, and which permitted remedies have been compared?
4. At which milestones do dependencies constrain work, and which development or funding decisions can overlap?
5. What demand by skill and period meets the available capacity after operating and support commitments?
6. What are the consequences of delay, how uncertain are they, and how do they change the feasible portfolio?
7. Why was each reviewed item retained, changed, paused, or stopped, and what future costs and obligations remain?
8. Which alternatives were set aside, and what evidence or date would prompt reconsideration?

The committee’s revised plan keeps the nine previously completed projects separate from the eighty-five open items. Of those open items, eighty-one had been labeled growth, with thirty-eight active and forty-three not started. The other four were three active renewal projects and the waiting data platform. After reviewing evidence and alternatives, nineteen growth items remain separate projects: ten active and nine waiting. Forty-four leave the current delivery plan: forty-two are canceled and two paused with recorded restart conditions; twenty-five of these forty-four had been active. Twelve are merged into the retained projects, including three previously active items. Six not-started items are reclassified as run. Thus 19 + 44 + 12 + 6 accounts for all eighty-one; merger and cancellation are not reported as completed delivery.

The three projects sharing the four engineers are among the twenty-five removed from active delivery: two are paused and one canceled. After handover and closure work, each engineer’s three quarter-time allocations are released to the platform, while their remaining quarter stays with support. The platform receives four people at 75 percent each, or three full-time equivalents. The ten retained active growth projects and three active renewals continue, and the platform starts: active project count changes from 41 to 14 by removing 25, merging 3, and adding 1. Those counts describe the chosen work structure, not a universal capacity limit.

The next-year plan uses the 55/35/10 allocation above and retains operating owners for paused, merged, and canceled work. Eleven of the fourteen active projects meet their revised delivery criteria within the following year. The nine completions in the prior three years are reported separately: scope, remaining work, and project grouping differ, so the counts do not establish the effect of the portfolio process. A first-plant predictive-maintenance pilot starts after data and operating-readiness checks. Whether maintenance improves still requires outcome and cost evidence. The committee now records those unresolved claims alongside the choices it made and the conditions for reviewing them.

References:

• OMB, Federal Enterprise Architecture Framework Version 2 (2013), portfolio planning and Performance Reference Model
• Richard Rumelt, The perils of bad strategy, McKinsey Quarterly (2011)
• Steve Coley, Enduring Ideas: The three horizons of growth, McKinsey Quarterly (2009)
• Scaled Agile, Weighted Shortest Job First, publicly available definition
• Sean McBride, RICE: Simple prioritization for product managers, Intercom
• MIT CISR, 50th—50 Facts, enterprise architecture maturity
• DORA, Accelerate State of DevOps Report 2024, official summary
• GOV.UK, Managing legacy technology (Open Government Licence v3.0)
• GOV.UK, Managing technical lock-in in the cloud (Open Government Licence v3.0)