The decisions that must be made before any ranking is legitimate.

When a team gathers to rank items without first establishing the Frame, they are engaged in the illusion of alignment. The room produces consensus. Executives nod. A spreadsheet is published. The meeting felt productive.

But the ranking is hollow. It has no structural legs. It cannot survive its first encounter with execution pressure — the moment a VP returns from a competitor briefing, or a board asks why we’re building this instead of that, or a team lead needs to know what to protect when resources contract.

The rationale was never captured. The trade-offs were never recorded. The criteria were never formalized.

The Frame is the structure that makes ranking mean something.

What the Frame establishes

Every defensible prioritization outcome requires five foundational decisions before the ranking can occur:

1. Participation

Who has a voice? Who decides? Who observes?

Without explicit participation, the ranking becomes a political act dressed as an analytical one. Late vetoes appear. Illegitimate outcomes are produced by people who were never at the table — or by people who were at the table when they shouldn’t have been.

Priorities.ai enforces explicit participation through role-based governance: Cycle Owners set scope, Facilitators run sessions, Decision Authorities hold final approval, Contributors and Subject Matter Experts provide the judgment inputs. Every person’s role in every decision is on the record.

2. Pool Formation

What is actually eligible to be ranked?

Without explicit pool formation, the option set is arbitrary. Items appear because someone remembered to include them. Others are absent because no one thought to ask. The ranking reflects the memory of the room, not the actual landscape of possibility.

The platform separates Pool Formation from the ranking as a distinct, explicit step. The candidate pool is assembled, normalized, and locked before any scoring begins.

3. Criteria Selection

What dimensions of value matter, and how are they weighted?

Without explicit criteria selection, the scoring is corrupted before it begins. If criteria are chosen informally — or, worse, adjusted mid-process to favor a predetermined winner — the outcome has no integrity regardless of how rigorous the spreadsheet appears.

Criteria are selected in a dedicated session step and locked before participants score items. Any subsequent criteria change is an admin-level action with a required disposition (continue with updated data / invalidate and restart / defer until session closes).

4. Scale Definition

How are criteria expressed and measured?

Value ranges, scoring scales, and measurement approaches are defined and confirmed before value assignment begins. This prevents the common failure where two participants interpret “high impact” differently and the aggregated scores are meaningless.

5. Value Assignment

How does each item perform against each criterion?

Expert and stakeholder judgments about each item’s performance are gathered explicitly, with provenance. Every accepted value carries a record of who provided it, when, and in what context.

Comparability Geometry

The Frame does not just establish who and what — it establishes which comparisons are valid.

In Priorities.ai, items are only comparable when they occupy the same Track Cell — the intersection of Item Category and Altitude.

Track Cell = Item Category × Altitude

Example Track Cells:
  Strategic Initiative (Category: initiative, Altitude: 3)
  Customer Feature    (Category: feature, Altitude: 2)
  Technical Debt      (Category: technical, Altitude: 1)

A strategic initiative should not be ranked against a bug fix. A corporate-level bet should not be ranked against a team-level task. Forcing those comparisons produces rankings that collapse under any scrutiny.

Track Cells enforce that items are only compared against their actual peers. This is the mechanism that answers the “comparing apples to oranges” problem — the one that appears in every organization’s list of reasons why their current process doesn’t work.

Item Relationships Within the Frame

The Frame also governs how item relationships affect the comparison. The platform maintains a typed relationship graph on every item:

Relationship	What it means in the Frame
Bloc	These items must be selected together or not at all — atomicity governs commitment
Dependency (hard)	This item should not receive a commitment-quality ranking until the dependency is explicitly addressed
Dependency (soft)	Warning: selecting this without the other may reduce expected value
Aggregation	When a parent is committed, the system surfaces all hard dependencies of its children
Cluster (duplicate)	These items must be resolved before either can enter a comparison

None of these are passive labels. Each carries enforcement logic that the Frame applies at the relevant decision point.

Protection Rings

Once items enter a prioritization context, they are protected from comparability-affecting changes:

Ring	Context	What it governs
Ring 3	Not in a track or session	Fully editable
Ring 2	In a track, not in an active session	Comparability-affecting changes require approval and audit
Ring 1	In an active session (locked comparison context)	Comparability-affecting changes require admin approval + session disposition declaration

The most-restrictive-ring rule applies when an item participates in multiple contexts simultaneously.

This prevents the common failure where someone edits an item mid-session — changing its category, its altitude, or a scoring attribute — without the session being aware that the comparison context has shifted.

What becomes possible when the Frame is locked

When the Frame is established and locked before the ranking runs:

The ranking can be defended. Not because it was made by the right people or with the right tool — because the decisions that preceded it are documented. Who participated. What criteria were applied. What trade-offs were acknowledged. The answer to why exists on the record.
The ranking travels. When the output reaches the execution team, they understand not just what they are building but why — and they can evaluate whether a proposed change is a legitimate response to new information or an erosion of a commitment that still holds.
The ranking accumulates. Over multiple cycles, the organization builds institutional memory about what types of decisions hold under pressure, which criteria matter most by altitude and category, and which items tend to generate dependency conflicts before they are resolved.

Developer access

The Frame is exposed through the API as:

POST /v1/cycles — create a governance container
POST /v1/sessions — configure a session with criteria and participants
GET /v1/items/:id/relationships — read item relationships and their enforcement state
PATCH /v1/items/:id/attributes — upsert attribute frames (RICE, sizing, custom)
POST /v1/sessions/:id/transition with to_state: CRITERIA_FINALIZED — lock the criteria