Concept

GIST Framework

conceptproductevidence-guidedgistprioritisationexperimentation

GIST Framework

GIST (Goals, Ideas, Steps, Tasks) is Itamar Gilad’s meta-framework for evidence-guided product development. Introduced in Evidence-Guided: Creating High-Impact Products in the Face of Uncertainty (2023). It integrates lean startup, design thinking, product discovery, agile delivery, and OKRs into a single coherent model. Each layer addresses a characteristic failure mode in conventional plan-and-execute product organisations.


The core problem GIST addresses

Opinion-based development: teams decide what to build based primarily on conviction, strategic narrative, or seniority — without empirical evidence of user behaviour or business impact. Itamar contrasts two Google products:

  • Google+ (~1,000 people, 2011–2019): plan-and-execute mode; hypothesis untested at scale until years of investment were sunk. Shut down 2019 with no measurable impact on competitors.
  • Gmail tabbed inbox (2012–present): zero-code validation first (Wizard of Oz test); discovered that ~85–88% of passive inbox users loved it — a finding internal “power user” bias would have suppressed. Now used by 1.8B users.

The difference was operating mode, not talent. Both projects involved the same organisation.


The four layers

G — Goals

What goes wrong: goals as planning (“what do we build by when?”); siloed functional goals that pull teams in different vectors.

What GIST proposes:

  1. North Star metric — measures value delivered to users (WhatsApp: messages sent; Airbnb: nights booked; Amplitude: weekly active learning users).
  2. Top business KPI — measures value captured (revenue, profit, market share).

Both together define the value exchange loop: the organisation creates value for users and captures value back; growing both in balance drives compounding growth.

Build a metrics tree from each:

  • Decompose each top metric into sub-drivers hierarchically.
  • The two trees overlap in the middle — these are the highest-leverage sub-metrics (moving them shifts both trees).
  • Teams own sub-metrics as their area of responsibility.
  • Trees expose the quantitative impact of sub-metric changes on top-level metrics.
  • Team topology can be rationalised around tree structure rather than functional hierarchy.

OKR integration: metrics trees + mission populate OKRs. Max four key results per team.


I — Ideas

What goes wrong: ideas evaluated by conviction, strategic theme (“it’s about AI”), or HiPPO (highest-paid person’s opinion). Confidence is self-assigned and almost always inflated.

What GIST proposes: ICE scoring + confidence meter.

ICE (Impact, Confidence, Ease; created by Sean Ellis):

  • Impact: effect on stated goals — hard to estimate; best-case backed by test data; at minimum, structured estimation.
  • Confidence: how sure should we be about impact and ease? The hardest and most often inflated dimension.
  • Ease: inverse of effort — also a guesstimate, but asking the question improves discussion.

Confidence meter — a tiered calibration model (0–10):

ScoreEvidence classExamples
0–1OpinionSelf-conviction, pitch decks, strategic themes
1–2SocialStakeholder review, colleague feedback
2–3EstimatesBack-of-envelope modelling, business case
3–4Anecdotal dataA few interviews, competitor has the feature
4–5Market dataSurveys, competitive analysis, field research
5–7Low-fidelity testsFake door, Wizard of Oz, usability study
6–8Rough buildsEarly adopter programme, fish food
8–10ExperimentsAB test, multivariate, staged rollout

Key principle: investment in an idea should scale with confidence level. Start cheap; earn the right to invest more.


S — Steps

What goes wrong: teams equate “build an MVP” with “run an experiment.” The MVP is usually a near-complete beta, expensive and too late to course-correct cheaply.

What GIST proposes: a full spectrum of validation methods, each a learning milestone:

LevelMethodsCost
AssessmentICE, assumption mapping, stakeholder 1:1s, business modellingNegligible
DataUser interviews, surveys, competitive analysis, observationLow–medium
Fake/low-fidelityFake door, smoke test, Wizard of Oz, conciergeLow
Rough buildFish food, early adopter programme, longitudinal studyMedium
Near-complete buildDog fooding, preview, beta, labsMedium–high
ExperimentsAB test, multivariate, hold-backs, staged rolloutHigh

Each step generates evidence. After each: continue, pivot the idea, or kill it and promote the next ICE idea. “You don’t have to start at the right-hand side, which is expensive.”


T — Tasks

What goes wrong: two disconnected worlds. Managers in roadmap/strategy mode. Developers in Jira/story-points mode. PMs exhausted as translators. Developers disengaged from users and outcomes.

What GIST proposes: the GIST board.

GIST board (per team):

  • Goals: max 4 key results for the quarter
  • Ideas: active ideas with ICE scores
  • Steps: next learning milestones for each idea

Reviewed every two weeks. Discussion: Are we on the right ideas? How are we tracking against goals? What’s blocking the most important steps?

“This middle layer discussion — what are we actually trying to achieve and how well are we doing — is the one that doesn’t happen. Most discussion is at the roadmap level or the task level.”


Outcome vs. release roadmaps

Release roadmapOutcome roadmap
Commits toFeatures + datesGoals + dates
SolutionPre-determined (low confidence)Left open until confidence is gained
Effect on discoverySuppresses it (race to ship)Enables it
When features appearUpfrontAfter validation (promoted to dated milestone)

The shift to outcome roadmaps is culturally disruptive — it requires executive alignment because it removes the certainty that feature roadmaps provide to stakeholders.


Stage calibration

StageRecommended approach
Pre-PMF startupFocus on finding PMF. Metrics trees and heavy OKRs are overkill. Iterate fast; goals = find PMF.
Series A–BStart building North Star metric and business KPI. Lightweight GIST board and ICE useful.
Scale-upFull GIST warranted. Cost of opinion-based development is highest here.

Relationship to other frameworks

  • JAM Model (Gibson Biddle): JAM force-ranks Growth/Engagement/Monetisation. GIST’s North Star metric + business KPI is roughly equivalent; GIST adds the full tree decomposition and the decision framework.
  • DHM Model (Gibson Biddle): DHM evaluates product strategy quality; GIST evaluates the process for generating and validating that strategy.
  • OKRs: GIST is OKR-compatible. The key results in OKRs map directly to GIST goals; the metrics tree populates them.

See also