Pairwise showdown

Scoring frameworks — RICE, weighted grids — ask you to put numbers on things nobody believes to one decimal place, then inherit that false precision into the ranking. A pairwise duel asks the one question people can actually answer: between these two, which first? Do enough of them and a full order falls out (Copeland: rank by wins minus losses), but you never had to invent a single number. The tool is stingy with your attention — it serves the duels that carry the most information (closing triangles, splitting near-ties), so a dozen taps usually settle a field of eight.

The real payoff is where the order won't settle. If you say A beats B, B beats C, then C beats A, there is no consistent ranking — and that's not a mistake, it's a signal. Each of those three judgements was made on a different criterion (A is cheaper, B ships sooner, C matters more), and the loop is the moment those unshared criteria surface. The tool finds every such loop (strongly-connected components of the beats-graph) and asks you to name the criterion on each edge; once you do, "three criteria pretending to be one" is the honest read.

So use the order when it's clean, and use the loops when it isn't — they're the argument worth having, found before anyone anchored on a spreadsheet. A position that “needs direct comparison” means two neighbours were never directly compared; the tool will offer you exactly those duels if you want to firm it up. This is a priority discussion record, not an expected-value model, a decision receipt, or a delivery plan: use Tree for quantified choices, Paths for conditional investment, and Roadmap for delivery commitments.