Fund the improvement portfolio your operation can actually finish.

Create one view of improvement demand

Bring daily-improvement escalations, RCA actions, maintenance changes, quality CAPA, automation, digital initiatives, and capital proposals into a common portfolio view without forcing identical delivery methods. Classify by value stream, problem, outcome, risk, resource, horizon, and decision status.

Keep ideas separate from authorized projects. An idea needs enough evidence to compare; a project needs an owner, outcome, boundary, capacity, controls, and next decision. Recording every suggestion as active work creates invisible overload and long cycle times.

• Problem and affected customer or stakeholder
• Value stream, constraint, and strategic connection
• Expected outcome, evidence confidence, and uncertainty
• People, downtime, capital, data, and specialist capacity
• Safety, quality, regulatory, cybersecurity, and change risk
• Dependencies, earliest learning point, and stop criteria

Prioritize with comparable evidence, not inflated savings

Use a small set of decision dimensions: system impact, urgency, confidence, effort, risk, strategic fit, and learning value. Convert claimed benefit into the mechanism that creates it. Separate cost avoidance, released capacity, cash, risk reduction, and booked financial benefit.

Avoid one weighted score making the decision automatically. Scores expose assumptions; leaders still need to balance mandatory work, short and long horizons, value streams, and uncertainty. Preserve the rationale so later results can improve the model.

Limit portfolio WIP and fund the next learning point

PMI treats portfolio management as a continuous process of selection, authorization, monitoring, and change. Apply the same logic to improvement work. Set WIP limits by constrained resource and review new work against capacity before authorization.

Fund uncertain work in stages: problem validation, causal evidence, pilot, scaled change, and sustainment. Each stage should answer a decision. This reduces sunk-cost pressure and lets the portfolio redirect people when the evidence weakens.

1. Inventory

   Create a common view of proposed and active improvement demand.

2. Screen

   Remove duplicates, unsupported solutions, and work outside strategy or authority.

3. Balance

   Compare value, risk, horizon, dependencies, and constrained capacity.

4. Fund learning

   Authorize only to the next evidence-based decision gate.

5. Verify and rebalance

   Reconcile results, release resources, and update priorities.

Choose an initial project limit from the shared work

Illustrative planning case: one specialist can give improvement projects 12 hours per week after routine duties. Each active project needs about four specialist hours that week, so three simultaneous projects would consume all 12 hours. Start a trial with two, reserving four hours for variation and urgent work. This is a local starting hypothesis, not a universal two-project rule.

For four weeks, record ready projects waiting, specialist hours actually required, completed learning steps and age of the oldest item. If two projects repeatedly need more than 12 hours, split or pause work before starting another. If the reserve stays unused and waiting grows, test a third only after checking the expected demand and dependencies.

Stop the trial if safety work is delayed, quality checks are bypassed or overtime exceeds the agreed limit. Revisit the estimate when the specialist workload changes. A weighted priority score cannot create those missing hours. The workflow bottleneck guide follows the shared review queue in more detail.

Govern dependencies and operational change

Map shared equipment, shutdown windows, products, validation resources, data sources, vendors, and standards. Two individually sound changes can conflict when they alter the same control logic, operator routine, schedule, or metric.

Assign portfolio, value-stream, and project decision rights. Define who can start, pause, stop, change scope, accept risk, and verify benefit. The review cadence should match the work: urgent containment is not governed like a capital expansion.

Close the loop from forecast to sustained benefit

At closure, compare baseline, prediction, actual result, implementation cost, operating cost, side effects, and confidence. Continue observation through a representative sustainment window. Do not count released minutes as cash unless the operation actually converts the capacity.

Use forecast accuracy, recurrence, benefit decay, project cycle time, late-stage cancellation, and capacity overload to improve portfolio decisions. A project that disproves an expensive assumption early can be a valuable success if the learning is retained.

Five priority levels need local decision rules

Critical, high, medium, low and optional can be five useful priority labels, but there is no universal numbering convention. Some systems make 1 highest; others make 5 highest. Write the meaning beside the code.

For this portfolio, agree what requires immediate risk escalation, what protects a committed customer outcome, what merits the next capacity slot, what can wait and what should not start. This is a proposed local policy, not a published industry scale. Record who can change a priority and which existing work must pause. Ten “critical” projects do not create ten available engineers.

Give each benefit a record that survives project closure

PMI separates identifying, delivering and sustaining benefits. An improvement team can connect those stages in one short record: the owner, process boundary, baseline, expected mechanism, measure, implementation and running costs, observation period, actual result and sustainment review.

There is no universal requirement to split benefits into exactly four types. Cash savings, cost avoidance, released capacity and reduced risk answer different questions. Use categories your finance and process owners can reconcile, and retain uncertainty rather than forcing every result into a currency value.