Create Real Options

An option is a specific action that the decision owner could actually take. If the option set is weak, accurate forecasts and sophisticated scoring merely identify the least bad version of an avoidable choice.

People often compare the first idea with the status quo and call that a decision. “Launch now or cancel,” “hire Alex or hire nobody,” and “build or buy” may each be a false dilemma—a frame that presents two choices when other feasible paths exist.

Generating options is not brainstorming for entertainment. Each proposed action must differ on a consequential lever—timing, scope, provider, process, commitment, or resource allocation—and be plausible enough to evaluate.

Start by banning evaluation for ten minutes. Early criticism makes people defend the first acceptable idea instead of searching for a better one.

The three-option minimum is a simple guard against binary thinking. Before evaluation begins, require at least three distinct, feasible actions, including one that changes the design of the choice.

Use these roles:

Role	Question	Launch example
Leading option	What action currently appears best?	Launch the full product in September
Strong alternative	What would a thoughtful dissenter choose?	Delay to November for reliability work
Redesigned option	Can scope, sequence, or commitment change?	September launch to 10% of users with a stop rule

“Do nothing” counts only if it is specified. “Keep the current product for six months while measuring renewal and building a smaller replacement” is an option. “Wait” without a date, learning plan, or consequence is avoidance.

Check distinctness by asking whether two options would create meaningfully different consequences. Renaming the same plan or changing a trivial feature does not expand the set.

A decision lever is a feature of the action that can be changed. Moving systematically across levers produces more useful options than waiting for inspiration.

For a customer-support redesign:

Lever	Variations
Who answers	Internal team; specialist partner; community; self-service
Which demand	All tickets; routine questions; one language; one customer tier
When	Immediate; phased over three months; after product fixes
Channel	Email; live chat; callback; in-product guidance
Commitment	Full contract; 60-day pilot; cancellable month-to-month
Ownership	Fully internal; outsourced; split by complexity

Combine one or two variations into coherent options. Avoid producing a catalogue of fragments that nobody could execute.

Use a constraint creatively: “We cannot add headcount” can prompt prevention, tooling, prioritization, a partner, or a temporary transfer. A boundary should guide design before it kills imagination.

A hybrid option combines useful parts of two approaches while separating the conditions under which each is used. Good hybrids resolve a tradeoff structurally; weak hybrids merely add both costs.

Common hybrid patterns:

• segment: use one model for high-value customers and another for routine demand;
• sequence: run a service first, then build internally after volume is known;
• threshold: automate low-risk cases and route exceptions to people;
• portfolio: fund several small experiments instead of one full commitment;
• core-and-edge: own the differentiating capability and rent commodity infrastructure.

Hybrids are not automatically superior. Count the interface cost: duplicate systems, handoffs, governance, training, and failure ambiguity. “Best of both worlds” can become “both bills and neither simplicity.”

Write the allocation rule precisely. “Use people for complex cases” is vague; “route any refund above £500 or any identity mismatch to a trained agent” can be implemented and audited.

Opportunity cost is the value of the best alternative you give up by choosing an option. It matters because most decisions consume scarce time, money, attention, or capacity even when no invoice appears.

If a team spends six engineer-months building an internal analytics tool, the cost is not only salaries. It is also the most valuable product, reliability, or customer work those engineers cannot do.

Use this comparison:

Option	Direct cost	Scarce resource consumed	Best forgone use	Opportunity-cost question
Build internally	£180,000	Six engineer-months	Improve checkout conversion	Is owning analytics worth more than the expected checkout gain?
Buy a service	£90,000 yearly	Procurement and integration time	Another vendor integration	Does speed justify recurring cost and dependence?
Stay with current tool	£20,000 yearly	Analyst time on manual work	Deeper customer analysis	What insight is lost while analysts clean data?

Do not add opportunity cost twice. If the benefit of the forgone project already appears in the comparison, label it clearly rather than counting it again as a separate penalty.

Ask a revealing question: If this option disappeared tonight, what would we do with the same people, money, and time? The answer identifies the true alternative, which may not be the official status quo.

Option generation must eventually become a shortlist. Remove an option because it violates a real constraint, is dominated by another option, or cannot plausibly achieve the required outcome—not because it feels unfamiliar.

A dominated option is worse than another on every relevant criterion and no better on any. If supplier A costs more, takes longer, has lower capacity, and carries equal risk compared with supplier B, A need not survive.

Use a transparent elimination table:

Option	Keep or remove	Stated reason	Evidence needed
Full internal build	Keep	Highest control; cost uncertain	Engineering estimate
Five-year vendor lock-in	Remove	Violates approved two-year commitment constraint	Contract term
Two-month pilot	Keep	Buys usage evidence with limited downside	Pilot proposal
Vendor X	Remove	Dominated by Vendor Y on price, capability, and exit terms	Comparable quotes

Do not eliminate on an uncertain assumption. Mark “conditional” and define a test: “Keep outsourcing if an audit confirms the provider meets the data-residency requirement.”

Record who removed each option and why. This creates an audit trail and lets new evidence restore an option without restarting the whole process.

The value of information is the expected improvement in a decision that comes from learning before committing. An experiment, prototype, reference call, survey, or delay is worthwhile only when it could change the action and its expected benefit exceeds its cost.

Consider a retailer deciding whether to spend £100,000 on a new format. Without research, it estimates:

• 50% chance the format succeeds and produces £220,000 of value;
• 50% chance it fails and produces £20,000 of recovery value;
• expected value before cost = (0.50 × £220,000) + (0.50 × £20,000) = £120,000;
• net expected value = £120,000 − £100,000 = £20,000.

A £5,000 pilot will not reveal the future perfectly, but suppose analysis estimates it would improve expected decision value by £18,000 by preventing some bad launches and rescuing some good ones. Its expected net information value is £18,000 − £5,000 = £13,000, before counting delay. The pilot is attractive if the delay does not destroy more than £13,000 of value.

Do not test what is merely easy to measure. Test the uncertain assumption most likely to reverse the decision.

A worked example shows how the option set changes the answer. Imagine a software company believes it must hire a senior data analyst because reporting is slow.

Frame: choose how to give product teams reliable weekly decision metrics within twelve weeks, with no more than £120,000 of first-year spend.

The first binary is “hire or do not hire.” Changing the levers produces:

1. hire a permanent senior analyst;
2. contract a specialist for twelve weeks to repair pipelines and train staff;
3. assign an internal analyst and buy a managed data-quality tool;
4. narrow the metric set, automate collection, and postpone recruitment until usage data arrives;
5. use a hybrid: contractor repairs the system, internal analyst owns the stable process, and a recruitment trigger is tied to request volume.

Opportunity cost changes the picture. Recruiting permanently uses nearly all the budget and three months of leadership attention. The contractor option is faster but may leave dependence. The managed tool adds recurring cost. The narrow-and-automate option may fail if the core problem is unclear definitions rather than collection.

The team selects a six-week contractor diagnostic with explicit deliverables, an internal owner, and a recruitment decision at the end. This is not indecision. It is a learning option designed around the uncertainty most likely to change the long-term commitment.

An option workshop is a bounded method for expanding the choice before comparison begins. It works best with people who understand operations, affected users, risk, and finance—not only senior advocates of the first proposal.

Minutes	Action	Output
0–5	Read the frame, outcomes, and hard constraints	Shared target
5–10	Write options independently	Less anchoring on the loudest person
10–20	Change who, what, when, where, how, and scale	Broad option list
20–28	Create hybrids, staged choices, and experiments	Redesigned options
28–34	Name opportunity cost for each serious option	Real resource tradeoffs
34–40	Remove only violations and dominated choices	Transparent shortlist
40–45	Identify the uncertainty worth testing	Information option or action

Quality checklist:

• [ ] At least three options are distinct and feasible.
• [ ] One option is a hybrid, staged commitment, or experiment.
• [ ] The status quo has a date, consequences, and opportunity cost.
• [ ] Options change meaningful levers rather than labels.
• [ ] Hybrids include coordination and interface costs.
• [ ] The best forgone use of scarce resources is named.
• [ ] Every removed option has a documented reason.
• [ ] Uncertain options are tested rather than dismissed by assumption.
• [ ] Information gathering is tied to a choice it could change.
• [ ] The shortlist remains small enough for serious consequence analysis.

The goal is not the largest list. It is a set in which at least one option is better because you deliberately designed it, not because it happened to arrive first.