How to prioritize when every request sounds urgent

Every request sounds urgent because it arrives with a local story. A customer is blocked, a seller has a live deal, support sees repeated pain, engineering sees accumulating risk, and an executive sees a market shift. Each claim may be true. Ranking the claims one at a time still produces a roadmap shaped by who can make the most vivid case.

Prioritization becomes possible only after the team states what kind of progress matters now. Diagnose the constraint limiting the desired outcome, choose a policy that says which work should receive concentrated effort because of that constraint, and collect comparable evidence before discussion changes people’s judgments.

Good Strategy Bad Strategy supplies the relationship between diagnosis, guiding policy, and coherent action. The Wisdom of Crowds explains why input quality depends on diversity, independence, local knowledge, and a real aggregation mechanism—not merely inviting many people to the meeting.

Urgency is a property of the requester’s position

Urgency usually answers “Who experiences a consequence soon?” It does not automatically answer “Which work changes the system most?” A request can be genuinely urgent for one account and weakly connected to retention, growth, safety, or the current strategy. A quiet infrastructure risk can have a larger cost of delay while generating no forceful advocate.

Do not dismiss urgency; decompose it. Identify the affected group, event or deadline, current workaround, consequence of waiting, reversibility, and whether the request reflects one case or a recurring mechanism. The story becomes evidence when another person can inspect those elements.

Importance is also not the opposite of urgency. Work can be both. The distinction exists to prevent emotional immediacy from becoming the ranking rule by default.

The same applies to executive requests. Seniority may bring broader context or merely more influence. Preserve the information by requiring the same causal account rather than either obeying automatically or treating hierarchy as proof the request lacks merit.

Diagnosis creates the basis for comparison

Rumelt’s diagnosis names why progress is difficult. A target such as “increase retention” is an outcome, not a diagnosis. The useful explanation identifies the mechanism currently limiting it: customers may not reach first value, one incomplete workflow may make the product replaceable, or the wrong segment may be entering through acquisition.

Without diagnosis, requests are compared through generic attributes such as reach, effort, confidence, or revenue. Those attributes can help, but the weights have no strategic meaning. A large-reach item may deepen a part of the system that is not constraining results.

Treat the diagnosis as a testable working model. State the evidence supporting it and the observation that would make the team revise it. Concentrating on a false diagnosis can be worse than dispersed work, so the reasoning must remain visible.

The appropriate planning horizon matters. A severe reliability incident may temporarily become the constraint even when the quarter’s growth diagnosis remains valid. Explicitly name whether an interruption replaces the strategic focus, consumes a reserved response budget, or is handled by a separate operating lane.

A guiding policy must reject plausible work

The guiding policy states how the organization will address the diagnosed constraint. It is more specific than a value such as “customer first” and broader than a list of features. It should create a reason some attractive work will wait.

Test the policy against the current request set. If advocates can plausibly rewrite every request to fit, the policy is not restrictive enough. A useful policy clarifies which segment, behavior, risk, or capability receives priority and which adjacent goals will be underachieved for now.

Priorities should also be coherent. Ask whether completing one item makes the others more effective. A set of individually high-scoring requests can remain strategically weak if each addresses a different theory of progress. Concentration creates leverage when actions reinforce one another around the constraint.

Reserve capacity explicitly for maintenance, mandatory work, incidents, and small customer commitments if those categories are real. Hiding them inside a strategic score makes the model dishonest. A portfolio can contain multiple lanes while remaining clear about the purpose and limit of each.

Collect independent evidence before the prioritization meeting

Surowiecki’s wise-crowd conditions are easy to destroy in group discussion. Once a senior person names a favorite, independent estimates become socially expensive. Vivid anecdotes become available to everyone, while quiet local evidence disappears. The meeting can converge quickly and still become less accurate.

Before discussion, ask each relevant function to submit the request, underlying observation, affected group, frequency, workaround, cost of delay, confidence, and connection to the diagnosis. Have participants evaluate the evidence or rank the small set privately. Preserve the raw observations separately from the proposed solution.

The meeting should focus on disagreement. A large spread can reveal different access to evidence, incompatible diagnoses, or uncertainty hidden by an average. Discussion is valuable for exposing those reasons; it is weak as the first moment judgments form.

Aggregation should match the decision. A median estimate can reduce the effect of extremes, but authority and accountability may still belong to a named decision-maker. Collective intelligence supplies a better information structure; it does not eliminate responsibility for the choice.

Cost of delay needs a specific causal path

“We will lose the deal” or “customers will churn” may be important and still insufficient. Identify which event occurs, how likely it is, when it becomes irreversible, and whether an alternative response exists. A workaround may lower urgency; a regulatory deadline may make delay discontinuously expensive.

Also examine the cost of interrupting current work. Switching priorities delays the existing coherent set, introduces coordination cost, and may teach the organization that every escalation resets strategy. The incoming request should be compared with what it displaces, not evaluated in isolation.

Requests that repeatedly arrive as emergencies may indicate a missing policy, service level, or product capability. The immediate item needs a decision, while the recurrence deserves a diagnosis of its own.

The next move is one evidence-first priority round

Write the current constraint and guiding policy at the top of a one-page request form. For the next planning round, require every candidate to include the observation behind it, affected group, frequency, workaround, cost of waiting, confidence, and the specific way it addresses the diagnosis. Collect the forms and private judgments before opening discussion.

In the meeting, first challenge the diagnosis with contrary evidence. Then discuss candidates with the widest disagreement and select the smallest coherent set that fits actual capacity. Record which attractive requests were excluded and why; an unranked backlog hides the tradeoff the policy was created to make.

Define an update condition for the chosen set. If the expected leading behavior does not change, or evidence shows a different constraint, revisit the diagnosis rather than simply adding more requests. The success condition is not that every stakeholder agrees their item is important. It is that the team can explain why the chosen work acts on the current constraint, what it displaced, and what evidence would change the choice.