shape-up

name

shape-up

description

Scope and bet on product work — four schools: Basecamp Shape Up (appetite, fat-marker shaping, betting table, 6-week cycles, scope hammering), Linear craft (no handoff/PMs, taste over data), Evil Martians sprints, metrics-led strategy (Perri/Klein/Lean UX). Use when shaping/pitching work, betting, cutting scope, or picking a school. Triggers: Shape Up, appetite, betting table, pitch, Linear method, design sprint, build trap, outcome roadmap.

Shape Up

Product development methodology from Basecamp built around fixed-time, variable-scope cycles.

Sources/gaps: references/sources.md maps Basecamp, Linear, Evil Martians, and metrics-led source schools; references/coverage-gaps.md keeps school conflicts and weak process claims visible.

Core Principle

Fixed time, variable scope. Don't ask "How long will this take?" Ask "How much time is this worth?" The appetite constrains the solution, not the other way around.

Phase Selection

Phase	When to Use	Reference
Shaping	Before betting, when defining what to build	shaping.md
Betting	During cool-down, when deciding what to commit to next	betting.md
Building	During the cycle, when executing on bet projects	building.md

Read the appropriate phase reference based on user context.

The Cycle Rhythm

6 weeks building → 2 weeks cool-down → 6 weeks building → ...

• Shaping happens continuously, producing pitches ready for the betting table
• Betting happens during cool-down, selecting what to build next
• Building happens during the 6-week cycle, with teams owning complete projects
• Budget refinement deliberately (MailChimp, The UX Reader): bugs and confusing workflows erode trust and support capacity, so some cycles must be cleanup by design, not leftovers. Their practice: at least once a year, "an entire release focused solely on cleanup and maintenance," fed by support, social feedback, analytics, interviews, surveys, and account-closing reasons.

Design sprints & design-build overlap (Evil Martians extension)

Sources: "MVPs, prototypes, results: how to win with a Martian Design Sprint" (Shamin & Turner, 2024) and "'Design first, then build': let's bury this myth forevermore" (Shamin & Turner, 2023). https://evilmartians.com/chronicles/mvps-prototypes-results-how-to-win-with-a-martian-design-sprint · https://evilmartians.com/chronicles/design-first-then-build-lets-bury-this-myth-forevermore

A two-week, fixed-appetite design cycle that pairs with the betting model:

• Stage-matched goals: pre-seed → an interactive prototype on a web link ("much more persuasive than relying solely on a pitch deck" — and a founder mirror that can force rethinking the idea); seed → "determining the product's essence: the core features that matter to users so much that they're willing to pay or switch" (the prioritization IS the MVP roadmap); post-PMF → hypothesis/ship/measure cycles.
• The cadence: kickoff narrows scope to what's valuable AND completable in two weeks; ~2 days exploration before designing; syncs every two days; designer-engineer alignment before handoff.
• The four anti-waterfall mantras: (1) "Design only what you need at the moment" — most startups pivot, so "designing just what you need gives you the power of the pivot"; (2) implement the design immediately — "hidden hiccups" only surface on implementation; keep design "a couple of weeks ahead of development," no more; (3) break big features into small chunks at the design stage (a release every ~2 weeks = healthy rhythm); (4) continuous two-way design↔dev communication, designer-owned.

Tension with Shape Up to route, not blend: Shape Up shapes with fat-marker sketches and resists high-fidelity handoff; the Martian sprint produces state-complete screens + handoff specs. Pick per project: fat-marker when the building team owns design decisions, sprint-fidelity when design is a deliverable to a separate build team or investor audience.

The Linear school (craft-first extension)

Sources: Karri Saarinen, "Primer on how we design at Linear" + tech-debt follow-up (X, 2023) · "Karri Saarinen's 10 Rules for Crafting Products That Stand Out" (Figma blog) · the Linear Method, esp. "Manage design projects" · Tuomas Artman, "Rethinking the startup MVP" (2024) · "Inside Linear" (Lenny's Podcast, interview digest). https://x.com/karrisaarinen/status/1715085201653805116 · https://www.figma.com/blog/karri-saarinens-10-rules-for-crafting-products-that-stand-out/ · https://linear.app/method/manage-design-projects · https://linear.app/blog/rethinking-the-startup-mvp-building-a-competitive-product · https://www.lennysnewsletter.com/p/inside-linear-building-with-taste

The third process school — it removes structure rather than adding it, betting on senior taste. Linear shares Shape Up's fixed-appetite instinct (cycles, scope-down) but rejects separating shapers from builders.

• Design is a reference, never a deliverable: "the design is only a reference… so the way it's constructed doesn't really matter." The mechanics: screenshot the app and design on top; a minimal design system ("mostly colors, type and some basic components"); one shared design file per quarter, a page per person/project; layer naming, auto layout, components all optional; "once the design is good enough and we start building… the real design is the app" — then loop. The tech-debt rebuttal: "No-one is going to create a new button component because it's 1px larger than what exists in the code" — the code design system stays the source of truth.
• The project loop: verify the root cause first ("users will ask you to build a specific feature X to fix their problem Y" — solve Y); open an "Explore designs" placeholder issue and "explore the solution freely and without judging whether something is feasible" (hours to days); seek feedback on direction, not polish — "don't just pay attention to what they say but ask them why they said it"; then "make a list of the design pieces… and get them done one by one," with engineers in the project from spec onward.
• No handoff, no PMs: "There's no 'handoff to dev.' You're never off the hook." Project leads (designer or engineer) replace PMs; Linear runs without metrics, PMs, or A/B testing. Small teams hired for product sense via paid work trials — "the more people you have, the more opinions and deliberation you introduce."
• Remote system as product system: Linear's 2025 remote-work note makes the same school operational: two-to-four-person project teams, rotating project leads, one product roadmap, short context/intent specs, feature flags for internal users and selected customers, weekly written updates that auto-post to Slack, triage/goalie rotations for unplanned work, and recurring Quality Wednesdays/feature roasts. Remote-first works because context, ownership, and feedback paths are designed, not because meetings are merely removed.
• Quality stance: the spec is the floor ("It's easy to meet the spec. It's harder to do the craft"); quality ≠ perfection — polish before customers see it, iterate after; the quality lever is scope ("the simplest way to increase quality is to reduce scope"); and the MVP rethought (Artman): "no longer about validating a novel idea as quickly as possible… its aim is to create a compelling product that draws in the early users" — narrow the audience, be substantially better for them.
• Taste over data: no A/B tests; "you must develop and trust your intuition"; "to provide the best experience, you must surprise users" — and no experiment produces a surprise. "We establish values and principles so that team members think about what they're building and why" instead of fixed process.
• Anti-roadmap + release rings (Ryo Lu, Cursor — plan-mode interview, youtube.com/watch?v=bdh8k6DyKxE): the same school at Cursor. Most companies do "planning theater… they pretend to have like a one-year road map or like a north star Figma" — a waste; with models and user mindsets shifting constantly, keep a directional bearing plus the present state and flexible aligned next steps, "talk to the users," iterate. QA runs as release rings: ambassadors and "nightly users" absorb bugs first (every build ships nightly — lots of bugs, but also the good surprises), and "The enterprise people get the slowest one."

The metrics-led product strategy school (Perri / Klein / Lean UX)

Sources: Melissa Perri, Escaping the Build Trap (2018); Laura Klein, Build Better Products (2016); Gothelf/Seiden, Lean UX, 1st ed. (2013). This school conflicts with Linear's taste-over-data stance; route by org maturity, traffic volume, and whether the team is discovering demand or refining a craft-led product.

The fourth process school treats shaping as strategy deployment plus learning loops.

• Strategy is a decision framework, not a plan: "Good strategy isn't a detailed plan. It's a framework that helps you make decisions." Use intent levels: company vision → strategic intent → product initiative → option/experiment. One intent is enough for a small company; three is plenty for a large one. More becomes Perri's "peanut-buttering" failure: too many initiatives spread over too few people.
• Quantified initiative shape: We believe that doing X for audience Y will achieve outcome Z, with a potential value estimate and a measurement plan. The example form is revenue-bearing and explicit, not a feature list. Pair every initiative with current-state data: Perri's Product Kata starts "What is the goal?" and "Where are we now in relation to that goal?"
• Done means outcome, not shipped: "We are done developing or iterating on a feature only when it has reached its goals." Perri's warning that "Version 2 is the biggest lie in software development" is the same trap Gothelf calls Phase II: don't bet on later cleanup unless the outcome gate forces it.
• Roadmaps explain strategy and stage: use Experiment / Alpha / Beta / GA, not a date-by-feature Gantt. Sales can only sell what has crossed the agreed stage boundary: anything GA, or far enough into Beta by explicit agreement.
• Fund product work like a VC: allocate budget to product lines and stages, not a fixed project list. Keep options alive while evidence is cheap; double down when the learning justifies it.
• Scope hammering from Klein: Finding the Core asks, "What has to happen for this feature NOT TO BREAK"; What-Happens-Next keeps asking until user intent is met; input-output matching cuts screens that do not transform user input into needed output. In constrained environments, build only what is absolutely necessary.

Routing among the four: Shape Up's guardrails (betting table, fat-marker, cool-down) protect mixed or scaling teams from overreach; the Martian sprint serves design-as-deliverable situations; the Linear school assumes a small, senior, taste-dense team dogfooding its own product daily, with leadership that treats craft as a business priority; the metrics-led school serves orgs stuck in output thinking, teams with enough traffic/data to learn from behavior, or bets where demand/risk is unknown. Without data maturity it degrades into vanity metrics; without taste it degrades into local-maximum optimization. Cursor runs the Linear school taken to its endpoint — designers shaping the product directly in code with AI agents, roles deliberately "muddy" — covered in design-prototyping's agents tier.

Team shape — "ship the org chart you want" (Nan Yu, Linear — Config 2024)

Org design as product design — the layer above betting/appetite. Decision rules:

• Conway's law as a design tool: make an org chart you actually want to ship; be suspicious of perfect symmetry ("where's the main thing?").
• Over-provision the main thing: Linear puts >50% of staff on the core product — that surplus is the quality bar.
• The scoping question: give a team the largest scope in which it can still make reasonable tradeoffs; slice narrow teams off only on demonstrated need, with explicit disband conditions.
• Zero-default rule: the default headcount on unimportant work is zero; narrow scopes breed kingdom-building.
• Interrogate borrowed numbers (the 5–8-person-team meme is far from its 1983–2012 sources), and don't patch a broken structure with OKRs.
• Big-bet acceptance bar (Duolingo, Config 2025): portfolio per pillar — incremental A/B tests plus a few big bets a year that need only metric-neutrality, buying strategic repositioning.

Decisiveness predicts quality (Linear "Conversations on Quality" series, 2024)

Stated independently by five guests: how fast a team decides — and un-decides — predicts its quality bar.

• The two-week reframe (Costolo): ask "what would have to be true for it to be two weeks?" — enumerates the assumptions to delete.
• Rolling 80/20 (Modisett, Perplexity): 80/20 everything weekly and circle back; investing in velocity makes more things good more of the time. One named decider per surface.
• Unshipping is a skill (Feener, Browser Company): dogfood-loop fast, call "this feels wrong — unship it" fast; interrogate literal feature requests instead of building the bug report verbatim.
• Set the bar, then stop (Nadler, Browser Company): do everything required to meet the bar, try not to do more; features make people switch, stability makes them stay.
• Bet-doc sculpting (Eismann, Slack): short bet doc, engineer starts immediately, team molds it to GA.
• Comms enables craft (Tome, Fellow): clear comms is what enables the focus craft needs; knowing the priority lets you say no.

Killing products — the sunset side of betting (Subham Agarwal, Config 2025)

The betting table decides what to build; this decides what to stop.

• Five legitimate death reasons: never took off · terrible unit economics · traction-but-strategy-mismatch · technology obsolete · less-is-more.
• RADAR: Read the signals → Assess (customer-value × business-value 2×2) → Decide (pause / iterate / kill) → Act → Review.
• Decide by identifying the most critical assumption and acting with confidence, not by long pro/con lists; ask "are we holding on out of inertia or ambition? Could this be a 10x?"
• Four forces block the call — inertia, hope, fear, love; indulging them cost the source team two years.
• Sunset comms: internal first (what/why/celebrate/learned), then external (tell them, why, time, alternatives). Component deprecation mechanics → design-systems.

Quote banks and case texture for these three sections: references/field-notes.md.

Relationship to other skills

• product-decision-making — pre-bet product judgment: problem clarity, evidence quality, optimization-vs-new-behavior classification, proposal review. Run it before shaping/betting; this skill owns the process once the bet is worth considering.
• user-research — discovery evidence that feeds shaping; pitch-shaped problem definitions consume its output.
• design-prototyping — the pitch/demo artifact itself (fidelity choices, presenting design work); this skill decides what to bet, that one builds the persuasive artifact.

Staleness note: the Basecamp cycle mechanics and shaping/betting/building references are durable book material. The Linear, Cursor, Evil Martians, Duolingo, and "Conversations on Quality" layers come from 2023–25 X threads, blog posts, and podcasts — company practices and roles change fast; verify before citing them as those companies' current process.