Trade-off Sliders

The situation link

Someone proposes a design and someone else objects. The argument goes in circles because both people are optimising for different qualities and neither has said which one.

The fix is boring and works: force the trade-off to be named out loud, in advance, with a ranking. Not “we value all of these” — a ranking.

The default link

Pick the top three, in order, before evaluating options. Write them down. The rest are constraints to satisfy, not goals to maximise.

The value of the table is not the pairings — those are debatable. It is that having a list in front of two arguing engineers converts an unfalsifiable argument into a resolvable one.

Doing it in a review link

A design review that produces “looks good to me” produced nothing. One that produces “we are buying availability with consistency, and here is where that shows up for the user” has done its job.

flowchart LR
    A["Proposal"] --> B{"Top 3 qualities<br/>named and ranked?"}
    B -->|No| C["Stop.<br/>Rank them first."]
    B -->|Yes| D{"What does<br/>each option spend?"}
    D --> E["Pick.<br/>Record the currency<br/>in the ADR."]
    C --> B

When the default is wrong link

Rankings are situational and expire. The ranking that was right during a launch push (time to market first) is actively harmful eighteen months later when the same team is drowning in the operability debt they knowingly took on.

Revisit the ranking when the phase changes: pre-product-market-fit, scaling, maturity, and sunset are four different rankings for the same system. A team that never re-ranks ends up defending decisions with reasons that stopped being true two years ago.

What it costs link

Explicit trade-offs create accountability, which people avoid. Saying “we chose speed over operability” in writing means someone can point at it during the incident review. That discomfort is the entire point — it is also why teams quietly resist the practice.

The other cost is false precision. These qualities are not measurable on a common scale, and pretending otherwise (weighted scoring matrices with made-up numbers) produces confident nonsense. Rank them; do not score them.

See also link

• Constraints First
• The Boring Baseline
• ADRs — where the ranking gets recorded.