EC-14.3 Trade-offs, Decided on Purpose

What this is and why it exists

Every interesting engineering decision is a trade, and most are made implicitly by whoever happened to write the first version.

This topic makes them explicit. Not to make them objective, which is not possible, but to make the judgement visible so that it can be discussed and later revisited.

The recorded decision is worth more than it looks. A project that can say why it chose something can change its mind cheaply. One that cannot is stuck with an accident.

The vocabulary

• Trade-off — a choice where improving one property worsens another.
• Criterion — a property being compared across options.
• Weight — how much a criterion matters relative to the others.
• Order-of-magnitude estimate — a rough calculation good to a factor of a few.
• Reversible decision — one that can be changed later at small cost.
• Decision record — a short written note of what was decided, the alternatives and the reason.

The mental model

Name the axes before comparing anything.

Cost against performance. Power against speed. Size against serviceability. Time against risk. Writing the axes down first is what stops a comparison from being decided by whichever option was described more enthusiastically.

It also reveals when two options are not comparable at all. They are strong on different axes, and nobody has said which axis matters more. That is a question for whoever set the requirements, not for the person doing the comparison.

A weighted comparison organises a discussion and does not settle it. Score each option against each criterion, multiply by the weights, and add.

Be honest about what that produces. The weights are judgements and the scores are estimates, so the total is a judgement dressed as arithmetic. The value of the table is in making the judgements visible, not in its bottom row. A total that contradicts everyone's instinct usually means a weight is wrong rather than the instinct.

Estimating before measuring is a habit worth more than it costs. An order-of-magnitude calculation of power, cost or time takes minutes.

Being wrong by a factor of two is fine. Being wrong by a factor of a hundred is the point. It tells you that an option is not worth investigating, or that a requirement is not achievable, before anyone spends a week.

The most useful rule in this topic concerns reversibility. Some choices can be changed next week; some are fixed once a board is manufactured or a supplier is contracted.

Spend analysis in proportion to how hard a decision is to undo. A reversible choice made quickly and revisited costs less than a careful analysis that delays everything. An irreversible one deserves the meeting.

Risk follows from the same reasoning. The part you are least sure about should be prototyped earliest, not last.

Projects fail late because the risky element was deferred until there was no time left to respond to what it revealed. Doing the frightening part first converts an unknown into either a solved problem or a change of plan. It does so while changing the plan is still possible.

Finally, recording the decision. A page each: what was decided, what the alternatives were, and why.

The payoff comes months later, when someone asks why the design uses this part. With a record, the answer is read. Without one, it is re-argued from memory, usually by people who were not there, and often reversed for reasons that were already considered and rejected.

What you should now be able to explain or do

• Name the axes a decision trades between before comparing options.
• Build a weighted comparison and state honestly what it does and does not settle.
• Make an order-of-magnitude estimate and use it to eliminate options.
• Classify a decision as reversible or not and spend analysis accordingly.
• Order the work so the riskiest element is prototyped first.
• Write a decision record that answers a question asked six months later.

Check yourself