Pairwise Prioritization

Pairwise prioritization solves the problem of P0/P1/P2 bucket crowding. It inherently solves ties and creates a persistent yet easily updated priority order. Efficiently manages backlogs of any size, short and long term.

The Tradeoff

The prioritization system tradeoff is: Scaleable, Accurate, Complete: pick two.

1. The P0/P1/P2 system: Scaleable and Accurate but not Complete. It doesn’t answer the “What’s next?” question.
2. Stack ranking: Accurate and Complete but not Scaleable. It answers the “What’s next?” question but breaks down at scale and is cognitively taxing.
3. Selection (just picking something from the list): Scaleable and Complete but not Accurate. It answers the “What’s next?” question by giving a suboptimal answer.

Scaleable, Accurate, Complete: pick three

The smallest element, the atom of prioritization is a comparison. It’s a decision about which one, between two tasks, is higher in priority. Other systems represent comparisons implicitly, for example “all P1 tasks are higher than P2”.

What if a prioritization system used comparisons as the first class citizen? Make N-1 pairwise comparisons to find the top priority: now you know what’s next. Start executing. Add a new item, and a single comparison will decide if it’s the new #1 or not. If you want you can also find the top-5 tasks just to see what the main choices are.

• It is Scaleable because no matter how many tasks are in the system, you only compare two items at a time.
• It is Accurate because nothing is left to chance.
• It is Complete because it unambiguously identifies the top priority.

So what are the tradeoffs?

• System assumes priority transitivety. Humans will sometimes respond in an inconsistent way: “A > B”, “B > C” but “C > A”.
• You can’t easily visualize the entire state of the system. The underlying structure is a potentially massive directed acyclic graph. Instead you can interrogate parts of the system.
• Non-priority tasks are intentionally left in an undetermined state. The system might not be able to say whether a given task is in the top 20% or on the bottom 20% of the task pool. (But neither can the bucket system.)

Why it works cognitively

Humans are inherently better at comparing than at absolute judgments. It is true for vision, hearing, and for cognitive faculties. When items are similar, absolute judgment collapses and relative judgment holds¹.

Stack-ranking is cognitively taxing because it requires holding in mind a large context. Comparing items two at a time is better aligned with the size of human working memory.²

Prior art / implementations

Non-exhaustive list of existing implementations.

• PriEsT (University of Manchester, 2011-2017) – Rooted in Analytic Hierarchy Process (AHP). Serious tool: inconsistency detection, Pareto-optimal solutions, judgment visualization. Last updated 2017, effectively abandoned. Failure mode: too heavy, assumed exhaustive upfront comparisons.
• All Our Ideas (Princeton, 2010) – Collaborative “wiki survey” using adaptive pairwise comparison. NYC’s PlaNYC 2030 used it to surface citizen ideas that outranked official proposals. Civic success.
• Flickchart — Consumer app for ranking movies via pairwise comparison. Demonstrated the mechanic can be engaging and habit-forming for entertainment. Not a task manager, but proof the UX can work at scale.
• TransparentChoice – B2B project prioritization tool using pairwise comparison and AHP. Enterprise context, team decisions, not personal backlog management.
• “Stacks” (HN Show HN, 2026) Positioned as “prioritize anything.” Appears to be a simpler implementation without the incremental update problem solved; community reception was mixed.
• Binary (Product Hunt) – Hiring software using binary/pairwise comparison for candidate evaluation. Domain-specific, not general task management.

Task Compass

Pairwise prioritization for personal tasks for iOS, macOS and Android.

Addresses failure modes of previous other implementations:

• incremental updates
• integrates instead siloing (Apple Reminders for iOS; Google Tasks, Microsoft To Do, Todoist, GitHub Issues for Android)
• fewer comparisons required: you don’t need to order the entire list. Finding the #1 needs at most N-1 comparisons. Task Compass does not require N(N-1)/2 comparisons like the other tools.