John Cutler's Post
John Cutler · article · 2026-09-12 source ↗ Reader 🖥️
Feedback LoopsGoodhart's LawProduct Discovery & ExperimentationProxy Validity (Construct Validity of Measurement Proxies)Strategy & Direction-Setting
Shifting from outputs to outcomes fails not because the idea is wrong but because teams adopt the language of outcomes while keeping the psychology of outputs — chasing certainty, easy measurement, and reputational credit instead of committing to the qualitative judgment, tolerance for ambiguity, and honest attribution that outcome-orientation actually demands.
practical · expository — An enumerative list of failure modes, each stated as a diagnosis-plus-implicit-prescription. It is practical (aimed at what teams should stop doing) but delivered expository/aphoristically rather than as an argued case — there is no sustained chain of reasoning, just eleven independently standing observations unified by a shared underlying claim about what 'outcomes over outputs' actually requires.
Structure
| Part | Role |
|---|---|
| Framing sentence: teams have good intentions but 'miss the point' of the outputs-to-outcomes shift. | context |
| #1 Leading vs. lagging confusion (chasing outcomes with no near-term feedback loop, or ignoring qualitative foundational work). | evidence |
| #2 and #5: Metrics-perfectionism — treating metrics as needing to be perfect before acting, and favoring easy-to-measure ideas over powerful ones. | evidence |
| #3: Analysis paralysis — endless framing debates that prevent shipping anything. | evidence |
| #4, #9, #10: Outcome theater — performative outcome-claiming, retrospective rationalization, quarter/promo-timed cherry-picking, evangelism over efficacy. | evidence |
| #6: Uneven application — outcome-centricity only where outcomes are visible (near customers/revenue), feature-factory behavior everywhere else. | evidence |
| #7: Loss of strategic conviction — incremental, easily-attributable improvements crowd out bets that shape the future. | evidence |
| #8: Functional balkanization — Engineering, Product, and Design each define their own convenient 'outcome' instead of converging on one. | evidence |
| #11: Outcome/attribution confusion — teams feel they must prove causal ownership of an outcome, which pushes them toward safe work or fabricated causal stories. | evidence |
Key terms
- outcomes
Changes in customer or business reality (behavior, value delivered) that a team's work is meant to produce — as opposed to the artifacts shipped. Cutler treats 'outcomes' as inherently harder to pin down, slower to confirm, and contestable in their causal attribution to any one team's work. - outputs
Shipped artifacts/features — the thing 'outcome-centricity' was supposed to move teams away from privileging, but which persists in disguised form (e.g. #4's performative outcome-claiming is output-thinking wearing outcome language). - leading vs. lagging outcomes
A distinction within outcomes themselves: leading outcomes are near-term and closable within a feedback loop a team can observe; lagging outcomes are foundational, qualitative, and only visible over a longer horizon. Cutler's #1 claims teams routinely fail to hold both in view at once. - metrics as hypotheses
Metrics are provisional instruments for testing a belief about value, not a target to be perfected before acting — perfection-seeking (#2) is itself evidence of misunderstanding what a metric is for. - claiming outcomes (performatively)
The act of declaring credit for an outcome through narrative framing rather than through a defensible causal link — 'retrospective rationalization' that makes outcome-claiming as gameable as output-shipping ever was.
Propositions
- explicitTeams fail to distinguish leading (near-term, closable) outcomes from lagging (foundational, qualitative, slow-to-confirm) outcomes, and as a result either chase outcomes they can't realistically close the loop on, or chase only near-term outcomes while neglecting the qualitative work that produces durable ones.conf: high
- explicitMetrics are hypotheses to be tested, not targets to be perfected before a team is allowed to act; demanding perfect metrics is a misunderstanding of what a metric is for.conf: highextendsTBM 437: Tokens, Hours, Points, and Other Curious Proxies · p13 — Both treat a business/ROI claim as a falsifiable hypothesis to be tested and retired rather than a fixed pitch to be perfected; Cutler generalizes the same epistemology from AI-ROI business cases to team metrics in general.extendsAre You Getting Your Money's Worth from AI? · p14 — Cutler's 'metrics are hypotheses, not perfection targets' generalizes the shore-ai piece's Goodhart's-Law-motivated rule that AI-impact metrics should be used to evaluate an approach then retired, not treated as an ongoing performance target.
- explicitOver-investment in framing the 'perfect' outcome is itself a failure mode, because it substitutes for shipping — the analysis-paralysis version of outcome-centricity is functionally indistinguishable from never trying.conf: high
- explicitOutcome-claiming can become just as performative and gameable as output-shipping once the spirit of experimentation is lost — there are many ways to 'get credit' for something that looks like an outcome but isn't, and this collapses into retrospective rationalization.conf: highrefinesThe 8 software engineering metrics AI broke · p9 — The AI-broke piece claims outcome/value measures are the metrics AI 'cannot inflate or fake' because they measure results rather than activity; Cutler's #4/#9/#10 refine this by showing outcome-claiming has its own gaming mechanism (performative claiming, retrospective rationalization, quarter-end cherry-picking) — outcomes are harder to fake by volume but not immune to narrative gaming.
- explicitTeams systematically favor metrics that are easy to measure over ideas that are powerful but only imperfectly measurable, inverting the priority that should hold between idea quality and measurability.conf: high
- explicitOutcome-orientation is applied unevenly across an organization: work visible to customers and revenue gets outcome treatment, while work further down the stack (platform, infrastructure, internal tooling) keeps running as a feature factory.conf: high
- explicitAn outcomes framing that only rewards attributable, incremental improvement erodes strategic conviction and the willingness to shape the future, leaving a team vulnerable to competitors who make bigger, less immediately measurable bets.conf: medium
- explicitEngineering, Product, and Design each default to defining outcomes that flatter their own function's contribution rather than converging on a single shared outcome that reflects what actually matters.conf: high
- explicitOutcomes reported by teams cluster suspiciously around quarter-end and promotion cycles because teams cherry-pick favorable results instead of reporting the normal, ongoing pattern of setbacks and reversals — the timing itself is evidence the metric is being managed for narrative, not learned from.conf: high
- explicitOutcome-centricity, treated as a principle to evangelize and a transformation to be seen completing, is a different thing from outcome-centricity as daily practice — evangelism can survive even after the practice (discovery, shipping, learning, absorbing setbacks) has been abandoned.conf: medium
- explicitRequiring teams to prove clean causal attribution between their specific work and an outcome pushes them toward two bad alternatives: avoiding valuable but causally-ambiguous work, or manufacturing implausible causal stories to satisfy the attribution demand.conf: highextendsTBM 437: Tokens, Hours, Points, and Other Curious Proxies · p9 — Cutler's diagnosis that attribution-demands push teams to avoidance or fabricated causal stories extends the tbm-437 prescription to judge return via validated proxies and a causal model 'rather than attributing outcomes to individual units of work' — same underlying critique of forced attribution, applied to team outcomes rather than AI spend.
- implicitThe demand for clean causal attribution is itself in tension with the outcomes philosophy: real outcomes are usually the product of many contributing factors, so an attribution requirement quietly reintroduces output-style, individually-creditable thinking under outcome language.conf: medium
- implicitAll eleven failure modes share a single root cause: organizations adopt outcome vocabulary without adopting the tolerance for ambiguity, patience, and qualitative judgment that genuine outcome-orientation requires — so the underlying psychology of output-chasing (certainty-seeking, easy measurement, individual credit) persists under new labels.conf: medium
Arguments
Outcome-centricity, as typically implemented, does not actually replace output-thinking — it re-encodes the same failure modes (gaming metrics, seeking easy credit, avoiding ambiguity) in outcome vocabulary. strength: medium
Explicit premises:
- Teams chase near-term, easily attributable outcomes over foundational, lagging ones (#1).
- Teams demand metric perfection instead of treating metrics as testable hypotheses (#2).
- Teams favor easy-to-measure ideas over powerful, harder-to-measure ones (#5).
- Outcome-claiming becomes performative credit-seeking once the spirit of experimentation is lost (#4).
- Reported outcomes cluster around quarter-end/promo cycles, evidencing narrative management rather than honest reporting (#9).
- Attribution requirements push teams toward avoidance or fabricated causal stories (#11).
Implicit premises:
- The psychological drivers behind output-chasing (certainty-seeking, desire for individual credit, aversion to ambiguity) are stable dispositions that survive a change in vocabulary unless the underlying incentive structure changes.
- An organization's stated methodology (outcomes-over-outputs) and its actual measurement/reward mechanics can diverge, and it is the mechanics, not the stated methodology, that determine behavior.
Inferences:
- If the diagnosis is correct, simply renaming output KPIs as 'outcome' KPIs (without changing what is rewarded, timed, or attributed) would predictably reproduce the same eleven failure modes under new labels.
- A team could exhibit genuine outcome-orientation only if it can tolerate a metric it hasn't perfected, hold conviction about a bet it can't yet attribute cleanly, and report a setback without professional cost — which makes outcome-orientation as much a psychological-safety and incentive-design problem as a metrics-design problem.
Demanding causal attribution of outcomes to specific work is itself a design flaw in how many organizations implement outcome-centricity. strength: medium
Explicit premises:
- Teams start believing they must prove their particular work caused the outcome (#11).
Implicit premises:
- Most valuable outcomes are overdetermined — produced by the interaction of many teams' work, market conditions, and timing — so clean single-team causal attribution is frequently impossible to obtain honestly.
- An organization that requires attribution as a condition for credit will get either avoidance of ambiguous-but-valuable work, or fabricated attribution narratives, because those are the only two ways to satisfy an impossible-to-honestly-satisfy requirement.
Inferences:
- The fix implied but not stated by the author is to fund and reward contribution to outcomes probabilistically or qualitatively (a portfolio/systems view) rather than requiring proof of individual causal ownership — which is the same corrective the corpus's proxy-validity material proposes for AI-ROI measurement.
Narratives
- The imagined C-suite executives who lived through an earlier 'product transformation,' got 'on a high horse' about outcomes, and are now rebelling because teams 'lost the thread.'
→ Supports proposition #10 (evangelism vs. efficacy) — it is Cutler's compressed anecdote for how outcome-centricity decays into a slogan divorced from the daily practice (discovery, shipping, learning, absorbing setbacks) it was meant to describe, and for why leadership sours on a good idea once it becomes theater.
Questions addressed
- ✅ Why does the well-intentioned shift from measuring outputs to measuring outcomes so often fail to change team behavior in practice?
Cutler resolves it by enumerating eleven concrete failure modes, converging on the implicit claim that teams adopt outcome vocabulary without adopting the underlying tolerance for ambiguity, patience, and qualitative judgment the shift actually requires. - ⭕ What should an organization do instead, given these eleven failure modes?
The post is entirely diagnostic; it names the ten (eleven) ways teams go wrong but does not prescribe a positive replacement framework, incentive redesign, or measurement structure that would avoid the failure modes. - ⭕ How should attribution of outcomes actually be handled if strict causal proof is neither honestly obtainable nor desirable to demand (per #11)?
Gaps
- incompleteThe post diagnoses eleven failure modes but offers no positive alternative: no proposed measurement cadence, incentive structure, or organizational practice that would let a team hold 'metrics as hypotheses' and 'strategic conviction' simultaneously without falling into paralysis (#3) or perfectionism (#2). Cutler names the trap but not the exit.
Read next: Read Cutler's own longer-form writing (e.g. his 'North Star Framework' and 'Outcome/Output/Opportunity Tree' pieces) or product-discovery literature (Cagan, Torres) for the positive framework this post assumes but doesn't restate. - incomplete#11's attribution problem — 'confusing outcomes with attribution' — is stated as an observation but not analyzed: the post doesn't distinguish cases where attribution is genuinely impossible (overdetermined outcomes) from cases where it is merely hard but tractable with better instrumentation (e.g. A/B tests, holdout groups), which would call for different remedies.
Read next: This corpus's extract-tbm-437-tokens-hours-points-proxies.json works this exact problem for AI-spend ROI, proposing 'the team as the indivisible cost unit' plus an explicit causal model instead of attributing outcomes to individual units of work — a template for resolving Cutler's #11 in a specific domain. - illogicalThe post implicitly treats 'outcome-based measures' as harder to game than output measures (the standard critique it's reacting against), yet its own #4/#9/#10 show outcomes are gameable too — just via narrative/timing manipulation (retrospective rationalization, quarter-end clustering, evangelism-over-efficacy) rather than volume inflation. The post doesn't reconcile this tension: it never explicitly revises the premise that switching to outcomes solves the gaming problem, even though its own examples show it merely relocates the gaming.
Read next: Read alongside Goodhart's Law material in this corpus (extract-8-engineering-metrics-ai-broke.json, extract-shore-ai-money-worth-part1.json) which makes explicit what Cutler leaves implicit: any measure that becomes a target, outcome or output, is subject to gaming — the fix is retiring metrics after they've been learned from, not choosing 'harder to fake' metric categories. - incompleteNo mention of psychological safety or incentive/reward-system design, even though several failure modes (#4 credit-seeking, #9 cherry-picking around promo time, #11 fabricating causal stories) are natural responses to a reward system that punishes honest reporting of ambiguity, setbacks, or unattributable work. The post diagnoses the symptom (behavior) without naming the likely structural cause (what gets rewarded).
Related documents
- The 8 software engineering metrics AI broke
- TBM 437: Tokens, Hours, Points, and Other Curious Proxies
- Presentation: Engineering at AI Speed: Lessons from the First Agentically Accelerated Software Project
- How to Build a Superteam That Keeps Getting Better
- Shifting Platform Development from Projects to Products
- Is Your Strategic Plan Too Ambitious? Or Not Ambitious Enough?
- John Cutler's Post
- #35 Lessons learned from a 2x Unicorn Founder (Uri Levine)
- Recommendation algorithms might be making your entertainment boring, new research suggests
- Measuring AI's Impact on Delivery Speed