Startup Priorities
Score features with (Breadth x Depth) / Cost, scaled to 1 - 100.
Formula: score = max(1, round(4 * b * d / c))
| Score |
Band |
Emoji |
Action |
| 1 - 4 |
Skip |
⚪ |
Backlog or kill it |
| 5 - 12 |
Low |
🔵 |
Only if nothing better exists |
| 13 - 24 |
Moderate |
🟡 |
Schedule in a future sprint |
| 25 - 49 |
High |
🟠 |
Current or next sprint |
| 50 - 100 |
Do it now |
🔴 |
Stop what you're doing |
Modes
- Autonomous (default): Score all three variables yourself using the litmus tests below and the context available. Output the one-liner immediately. No user interaction unless you hit uncertainty (see below).
- Collaborative: Walk through each variable with the user one at a time. Enter this mode when:
- The user explicitly asks to collaborate, discuss, or decide on the variables
- You lack sufficient context to confidently answer a litmus test (e.g. no info on user base size, no way to gauge cost)
- Two adjacent levels seem equally valid and the difference would change the band
When entering collaborative mode, explain which variable you're uncertain about and why, then use AskUserQuestion for that variable only. Return to autonomous for the rest.
Input spectrums
Each level has a litmus test — a yes/no question that pins the score and removes ambiguity.
Breadth (b) — reach across current users + validated TAM
Score based on current users or TAM expansion — whichever is higher. TAM claims require evidence (waitlist, inbound requests, competitor data, LOIs). "We think the market is big" without evidence stays at b=1.
| b |
Label |
Current users |
OR TAM expansion (with evidence) |
| 1 |
Niche |
<5%, named accounts |
No TAM signal |
| 2 |
Narrow |
5-20%, one segment |
A few inbound requests from non-users |
| 3 |
Moderate |
20-50%, multiple segments |
Validated demand: waitlist, competitor proof |
| 4 |
Broad |
50-80%, most weekly actives |
Large reachable TAM segment via existing channels |
| 5 |
Universal |
>80%, all users incl. new signups |
Unlocks dominant share of TAM |
Depth (d) — observable user behavior, not your guess
| d |
Label |
Litmus test |
| 1 |
Cosmetic |
If you shipped it silently, would anyone open a ticket or mention it? (No) |
| 2 |
Convenience |
Do affected users have a workaround they tolerate without complaining? (Yes) |
| 3 |
Meaningful |
Do users bring this up unprompted in feedback calls or surveys? (Yes) |
| 4 |
Critical |
Is it a top-10 feature request? Have sales deals stalled or required workarounds because of it? (Yes) |
| 5 |
Existential |
Have users churned explicitly citing this gap, or is a new market segment blocked without it? (Yes) |
Cost (c) — scope of the engineering work
| c |
Label |
Litmus test |
| 1 |
Trivial |
Can one person ship it today — config change, copy fix, flag flip? (Yes) |
| 2 |
Small |
One person, 1-3 days, isolated to one component, standard PR? (Yes) |
| 3 |
Medium |
Needs a design discussion, touches 3+ files across 2+ modules, may need QA beyond unit tests? (Yes) |
| 4 |
Large |
Requires cross-team coordination, a migration plan, or staging env testing? (Yes) |
| 5 |
Massive |
Would you create a project plan with milestones and track it weekly? (Yes) |
Workflow
Autonomous (default)
- Read the feature/task/goal from context.
- Score each variable using the litmus tests. For each, identify which litmus test applies.
- If confident on all three, compute and output the one-liner directly:
<emoji> Priority Score: <score>(<band>) - <archetype>: <brief why>
- If scoring multiple features, output a ranked list of one-liners.
Collaborative (on uncertainty or user request)
- Present the spectrum tables so the user sees the litmus tests.
- For each uncertain variable, explain the uncertainty and use
AskUserQuestion.
- Score confident variables autonomously — only ask about the ones you can't pin.
- Compute and output the one-liner.
Showing your work
When outputting the one-liner, always follow it with a brief breakdown so the user can challenge the score:
<emoji> Priority Score: <score>(<band>) - <archetype>: <brief why>
b=<b>(<label>) d=<d>(<label>) c=<c>(<label>)
Key rules
- Early-stage caveat: pre-PMF, Depth > Breadth. 100 devoted users beat 1000 lukewarm ones. Call this out when presenting results.
- This is a thinking tool, not a replacement for judgment.
- When uncertain about a variable, ask — don't guess. A wrong score is worse than a slow score.
See REFERENCE.md for archetypes, worked examples, and band analysis.
1---2name: priority-score3description: Score feature priorities using Geoff Ralston's (b*d)/c formula — Breadth, Depth, and Cost on 1-5 scales yielding a 1-100 priority score. Use when user wants to prioritize features, decide what to build next, rank a backlog, compare initiatives, or mentions "priority score".4---56# Startup Priorities78Score features with **(Breadth x Depth) / Cost**, scaled to **1 - 100**.910Formula: `score = max(1, round(4 * b * d / c))`1112| Score | Band | Emoji | Action |13|-------|------|-------|--------|14| 1 - 4 | Skip | ⚪ | Backlog or kill it |15| 5 - 12 | Low | 🔵 | Only if nothing better exists |16| 13 - 24 | Moderate | 🟡 | Schedule in a future sprint |17| 25 - 49 | High | 🟠 | Current or next sprint |18| 50 - 100 | Do it now | 🔴 | Stop what you're doing |1920## Modes2122- **Autonomous (default):** Score all three variables yourself using the litmus tests below and the context available. Output the one-liner immediately. No user interaction unless you hit uncertainty (see below).23- **Collaborative:** Walk through each variable with the user one at a time. Enter this mode when:24 - The user explicitly asks to collaborate, discuss, or decide on the variables25 - You lack sufficient context to confidently answer a litmus test (e.g. no info on user base size, no way to gauge cost)26 - Two adjacent levels seem equally valid and the difference would change the band2728When entering collaborative mode, explain *which* variable you're uncertain about and why, then use `AskUserQuestion` for that variable only. Return to autonomous for the rest.2930## Input spectrums3132Each level has a **litmus test** — a yes/no question that pins the score and removes ambiguity.3334### Breadth (b) — reach across current users + validated TAM3536Score based on current users **or** TAM expansion — whichever is higher. TAM claims require evidence (waitlist, inbound requests, competitor data, LOIs). "We think the market is big" without evidence stays at b=1.3738| b | Label | Current users | OR TAM expansion (with evidence) |39|---|-------|--------------|----------------------------------|40| 1 | Niche | <5%, named accounts | No TAM signal |41| 2 | Narrow | 5-20%, one segment | A few inbound requests from non-users |42| 3 | Moderate | 20-50%, multiple segments | Validated demand: waitlist, competitor proof |43| 4 | Broad | 50-80%, most weekly actives | Large reachable TAM segment via existing channels |44| 5 | Universal | >80%, all users incl. new signups | Unlocks dominant share of TAM |4546### Depth (d) — observable user behavior, not your guess4748| d | Label | Litmus test |49|---|-------|------------|50| 1 | Cosmetic | If you shipped it silently, would anyone open a ticket or mention it? (No) |51| 2 | Convenience | Do affected users have a workaround they tolerate without complaining? (Yes) |52| 3 | Meaningful | Do users bring this up unprompted in feedback calls or surveys? (Yes) |53| 4 | Critical | Is it a top-10 feature request? Have sales deals stalled or required workarounds because of it? (Yes) |54| 5 | Existential | Have users churned explicitly citing this gap, or is a new market segment blocked without it? (Yes) |5556### Cost (c) — scope of the engineering work5758| c | Label | Litmus test |59|---|-------|------------|60| 1 | Trivial | Can one person ship it today — config change, copy fix, flag flip? (Yes) |61| 2 | Small | One person, 1-3 days, isolated to one component, standard PR? (Yes) |62| 3 | Medium | Needs a design discussion, touches 3+ files across 2+ modules, may need QA beyond unit tests? (Yes) |63| 4 | Large | Requires cross-team coordination, a migration plan, or staging env testing? (Yes) |64| 5 | Massive | Would you create a project plan with milestones and track it weekly? (Yes) |6566## Workflow6768### Autonomous (default)69701. Read the feature/task/goal from context.712. Score each variable using the litmus tests. For each, identify which litmus test applies.723. If confident on all three, compute and output the one-liner directly:73 ```74 <emoji> Priority Score: <score>(<band>) - <archetype>: <brief why>75 ```764. If scoring multiple features, output a ranked list of one-liners.7778### Collaborative (on uncertainty or user request)79801. Present the spectrum tables so the user sees the litmus tests.812. For each uncertain variable, explain the uncertainty and use `AskUserQuestion`.823. Score confident variables autonomously — only ask about the ones you can't pin.834. Compute and output the one-liner.8485### Showing your work8687When outputting the one-liner, always follow it with a brief breakdown so the user can challenge the score:88```89<emoji> Priority Score: <score>(<band>) - <archetype>: <brief why>90 b=<b>(<label>) d=<d>(<label>) c=<c>(<label>)91```9293## Key rules9495- **Early-stage caveat**: pre-PMF, Depth > Breadth. 100 devoted users beat 1000 lukewarm ones. Call this out when presenting results.96- This is a **thinking tool**, not a replacement for judgment.97- When uncertain about a variable, ask — don't guess. A wrong score is worse than a slow score.9899See [REFERENCE.md](REFERENCE.md) for archetypes, worked examples, and band analysis.
mohganji