GOALS.2 - Construct evidence of achievement and a jointly sufficient KR set

Type: Method pattern. Status: Stable.

GOALS.2:1 - Problem frame

Use this when people need to tell whether a goal has been achieved, whether progress supports a response, or whether a proposed set of key results would establish its objective. The object is the achievement claim and the evidence used for it. The result is a usable test with its population, conditions and limits.

Use adequate existing subject criteria and observations directly. Designing an elaborate measure is unnecessary when the recipient can already judge the required result competently. A missing professional criterion returns to that practice; goal wording alone cannot supply it.

GOALS.2:2 - Problem

A list of completed activities can leave the intended result missing. An average can hide a failed necessary condition, and a favorable percentage can exclude the difficult unfinished cases. Even true observations can support the wrong conclusion when the subject, time or assistance differs. The goal needs an evidence argument that survives these differences.

GOALS.2:3 - Forces

Simple measures aid attention and comparison; omitted conditions can defeat their meaning. Early indicators support timely intervention but can precede the actual benefit. Joint results can depend on outside participants and delayed effects. Numerical precision helps only when its definition and evidence answer the receiving question.

GOALS.2:4 - Solution

GOALS.2:4.1 - Construct the achievement claim from a good result

Describe a concrete result that would satisfy the goal and a relevant case that would fail it. Identify the subject, eligible population, horizon and protected conditions. For a continuing service, choose the window over which its condition matters. For a finite task, state what closes it and how unfinished work remains visible. For a capability, identify performance under the relevant support and variation.

Separate the final result, a qualified intermediate contribution and an early signal. All can be useful observations. State which decision each supports. A completed design may be the entire result of a design task; in a delivery goal it is an intermediate contribution whose downstream realization still matters. This distinction follows the assigned result rather than a universal ban on outputs.

GOALS.2:4.2 - Challenge the proposed evidence

Ask whether the proposed observation could look successful while the goal remains unmet. Construct a plausible case using the actual work, such as reducing calls by making the service inaccessible, increasing completed tickets by excluding reopened cases, or improving a learning score through unrecorded assistance. Add or change only the observation or criterion needed to distinguish that case.

Use qualified existing data when it is adequate. If the missing distinction can change a decision, obtain the smallest suitable observation, subject judgement or inquiry. OPS.15 supplies a decision-specific operating account. A new collection regime also requires usable access, effort and permitted use. Changing visibility, ranking or personnel use can change incentives and participation; examine those consequences when the proposed arrangement introduces them. Merely reusing adequate existing service data creates no automatic new organizational intervention.

A numerical criterion needs a defined numerator, denominator, inclusion rule, time window and interpretation where those affect the conclusion. A qualitative criterion needs an inspectable result and competent grounds for judgement. Neither forces a universal score. Where decisive evidence remains unavailable, return what is established and the exact undecidable claim.

GOALS.2:4.3 - Test the key results together

For an OKR, test two questions separately: is the objective worth pursuing, and would the complete KR set establish its achievement under the stated assumptions? Work through a case in which every KR passes. Identify any necessary result still missing, any required quality that could fail, and any condition that could make the apparent success worthless. Repair the set at that omission; adding more indicators without a reason increases work.

Explain how the KRs combine. Some are jointly necessary conditions; some express independent contributions; others are alternatives or diagnostic signals. A mandatory privacy, safety or contractual condition retains its force even when other scores are high. Average scores can summarize a defined compensable set, but an average must not decide a claim whose logic is conjunctive. Keep this decision visible when a tool offers averaging by default.

Then test an unfavorable case. Could the objective be achieved by another sound means while a proposed KR fails? If so, decide whether that KR expresses a necessary outcome, an independently wanted result or a chosen means that should remain revisable. The result is a defensible construction under its conditions, not a universal causal guarantee that meeting any selected numbers produces the benefit.

GOALS.2:4.4 - Preserve exceptions, learning and uncertainty

A genuine final exception is handled according to the goal’s actual terms. Define who may decide it and what evidence makes it final. Keep unresolved cases open with the action and response they still require. An explanation, escalation or future answer can be essential accompaniment without completing the original result. If an authorized agreement changes the result or date, preserve the original basis for interpreting what happened.

For inquiry, specify what makes the investigation adequate and which possible conclusions it can support, including a qualified negative or inconclusive answer. For learning, use HCD.11 for the assessment inference and HCD.12 when transfer is part of the claim. Activity counts can describe effort or provision, while the claimed result needs its corresponding evidence.

Return the achievement test with only the definitions, observations and limits its users need. Recognition finds a misleading success. Assurance establishes why the evidence supports the particular claim, with stronger professional qualification where the consequence demands it.

GOALS.2:5 - Archetypal Grounding

A refund objective concerns 100 eligible cases by Friday. Proposed KRs are “FAQ published,” “bot launched” and “contacts down 30%.” All can pass with zero refunds. The team retains the FAQ and bot as possible means and follows each eligible case to paid refund or a valid final disposition under its terms. Open cases retain next action and contact; they remain open in the original cohort.

On Friday, 82 are paid, 8 have justified final dispositions and 10 remain unpaid. The original full result is unmet. If all 100 instead have a clear explanation and a promised answer, the result is still unmet. If the relevant parties agree a valid final settlement in a case, that is handled under the stated terms. A unilateral internal “exception” label establishes less.

For a different task, a researcher must supply a decision-ready comparison of two algorithms. A complete report with qualified measurements and an explicit finding that neither meets the operational requirement can achieve that inquiry goal. Requiring deployment success as its only passing result would suppress the answer the decision needs.

GOALS.2:6 - Bias-Annotation

Available data tend to privilege easy-to-count and easy-to-reach cases. Include relevant exclusions and inconvenient failures before drawing the intended conclusion. Respect the permitted use of personal or sensitive information; a goal does not create access rights.

GOALS.2:7 - Conformance Checklist

Does the evidence concern the right result, population, horizon and support conditions?
Could every KR pass while a necessary part of the objective fails?
Is each apparent exception final under the applicable terms, with unfinished cases still visible?
Do combination rules preserve required conditions and uncertainty?
Can existing evidence answer the question without unnecessary new collection?

GOALS.2:8 - Common Anti-Patterns and How to Avoid Them

Averaging away a required failure: retain the mandatory condition as a condition of achievement and use the aggregate only for its narrower descriptive purpose.

Replacing the cohort: keep the original population available when new easy cases would conceal old unfinished ones.

Proof through the chosen activity: examine the receiving result. Keep an activity as a criterion when that activity is itself the assigned useful contribution, such as a qualified investigation.

GOALS.2:9 - Consequences

A team can distinguish progress, adequate partial results and completion. Better definitions can expose previously hidden work or require a different source of evidence. The resulting test remains revisable when the goal, population or decision use changes, with the earlier basis preserved.

GOALS.2:10 - Architectural Rationale

The achievement argument is the bridge between a desired condition and observation. This pattern develops that goal-specific connection while leaving measurement, causal inference and professional assessment with their full suppliers. Counterexamples test a proposed interpretation economically; they do not replace empirical evidence when an effect or reliability claim needs it.

GOALS.2:11 - SoTA-Echoing

Doerr’s resource guide supplies the demand for meaningful, collectively adequate KRs and illustrates misleading inputs. This method adapts it through OPS.15, HCD and subject acceptance criteria, retaining useful outputs and qualified milestones. Contemporary evidence that more reported effort need not improve measured outcomes reinforces the distinction; it supplies no universal rule against effort measures. Reopen the test when a changed way of working defeats its inferential basis.

GOALS.2:12 - Relations

GOALS.1 supplies the result and kind of reliance. GOALS.3 uses newly exposed contributions and observation burden in feasibility. GOALS.4 receives observations with their actual meaning. GOALS.8 determines any distinct personnel use; an achievement result is an input to that decision, not a complete personnel assessment.