How to Turn Too Many Good Ideas Into a Focused Portfolio

Most strategy problems do not begin with a shortage of ideas. They begin with too many reasonable ideas, each backed by someone who cares about it.

That is where lists stop helping. A list tells us what we could do. Strategy must tell us what deserves attention now, what needs more evidence, who should own it, and what we are choosing not to do.

I use a simple opportunity architecture to make those choices clearer. It is not a formula that makes judgment disappear. It is a system that makes judgment visible.

Start with the changed condition

Raw ideas often arrive as activities:

• Run a program
• Build a platform
• Launch a community
• Add an AI feature

An activity is not yet an opportunity. First state what should become meaningfully different, for whom, and why the proposed mechanism might cause that change.

A useful frame is:

For a specific group facing a material constraint, there is an opening to achieve a changed condition through a plausible mechanism, because the timing makes action useful now.

“Run an AI workshop” is an activity. “Help a product team identify and test three valuable AI workflows it can own after the workshop” describes a changed condition.

This distinction removes many attractive ideas before scoring begins.

Put boundaries before scores

Some questions should never be hidden inside a weighted spreadsheet.

Before comparing opportunities, ask:

1. Is the intended outcome aligned with the mission?
2. Do the likely means preserve dignity, trust, rights, and safety?
3. Who benefits, and who carries the risk?
4. Whose knowledge is missing?
5. Are we the right actor, or would we duplicate or weaken better work?

A failed boundary is not repaired by a high total score. The idea must be reframed, referred to someone better placed, or stopped.

Build a bounded evidence packet

Research can become another way to avoid a decision. Give each opportunity the same small evidence packet instead:

• The scale and severity of the problem
• Existing work and previous attempts
• What practitioners and affected people say
• Why the timing matters
• The distinctive contribution we could make
• The assumptions most likely to be wrong
• The quality of the evidence

The point is comparability, not completeness. Contradictory evidence belongs in the packet. So do unknowns.

Separate score from confidence

A score can help a group compare unlike opportunities, as long as everyone remembers what it cannot do.

I usually examine questions such as:

• How material is the potential outcome?
• Does this build capability that enables later work?
• Is there a credible owner?
• Is the opening neglected?
• Can a focused team make progress with attainable resources?
• Can the result survive one energetic person?

Then I give the evidence a separate confidence label: high, medium, or low.

A high-potential opportunity supported by weak evidence is not a flagship. It is a candidate for discovery. A more modest opportunity with strong evidence and a ready owner may deserve action first.

Keeping score and confidence separate prevents enthusiasm from pretending to be certainty.

Build a portfolio, not a winner

A single ranked list creates false precision. Real organizations need different kinds of commitments.

I use four practical lanes:

Commit

A high-value, high-confidence opportunity with a credible owner and enough support to move.

Foundation

Shared infrastructure, knowledge, talent, or relationships that make several later efforts easier.

Experiment

A promising opening with an uncertain mechanism. It receives a bounded test, not a permanent program.

Watch

An important signal with no responsible path to action yet. It gets an owner and a review trigger, not performative activity.

The lane matters as much as the rank. It tells people what kind of commitment the evidence can support.

Match the work to an owner

An opportunity without a credible owner is still an idea.

Look beyond enthusiasm. Ask whether the person or team has legitimacy with the people affected, experience with comparable complexity, time, institutional support, access to missing partners, and the temperament to coordinate rather than control.

Often the right answer is not one heroic leader. It is a complementary pair: one person with domain trust and another with operating strength.

Ownership also needs a clear first ask. “Lead this initiative” is vague. “Interview five institutions, test the core assumption, and return with a recommendation in six weeks” gives everyone something real to evaluate.

Design the first credible test

The first test should reduce the most dangerous uncertainty, not imitate the final program at a smaller scale.

Ask:

• Which assumption is most likely to be wrong?
• What is the smallest result that would change our confidence?
• Who must participate for the test to be legitimate?
• What would make us continue, redesign, or stop?
• What reusable learning will remain if it fails?

A good test is allowed to disconfirm the idea. If every outcome can be presented as success, it is not a test.

Keep the decision alive

The portfolio should change when reality changes.

Record why each decision was made, what evidence supported it, what remained uncertain, and who owns the next step. Review new evidence regularly. Move opportunities between lanes. Stop work that no longer earns attention.

Accountability is not only reporting activity. It is being able to say:

• What changed?
• What did we learn?
• What did we get wrong?
• What should stop?
• What became easier for the next team because this work existed?

That last question matters. The strongest projects do more than produce one result. They leave behind infrastructure, knowledge, relationships, or trust that makes future work more possible.

The point

A focused portfolio does not come from finding a perfect scoring model. It comes from making several kinds of judgment explicit: mission, evidence, confidence, ownership, timing, and learning.

The system should help capable people disagree clearly, decide honestly, and act without pretending uncertainty has disappeared.

That is enough structure to choose. Then the real learning begins.