OBA cheat sheet

Outcome-based accountability (OBA) defines success by measurable results, not activities or outputs; every objective must have at least one indicator with a baseline and target.

A logic model links five components in order: inputs, activities, outputs, outcomes, and impact; outcomes describe changes in behavior, knowledge, or conditions, while outputs are countable deliverables.

SMART criteria: Specific, Measurable, Achievable, Relevant, Time-bound; a target lacking any one criterion is not verifiable.

Indicator types: output indicators count deliverables; outcome indicators measure change; efficiency indicators compare outputs to inputs; effectiveness indicators compare outcomes to targets.

A baseline is the pre-intervention value of an indicator; without a baseline, improvement cannot be quantified.

Theory of change differs from a logic model by articulating causal pathways and assumptions between each link, not merely listing components.

Stakeholder analysis maps interest and influence; high-influence, high-interest stakeholders require active engagement in defining outcomes.

Attribution versus contribution: attribution claims sole causation, contribution acknowledges multiple actors; OBA typically uses contribution analysis for complex settings.

A results framework nests impact, outcome, and output levels; each level must trace upward to the next with explicit means of verification.

Risk analysis scores each risk as likelihood times impact; risks above the tolerance threshold require a mitigation plan with an owner and date.

Assumption mapping separates conditions outside the program's control (assumptions) from factors it can manage (risks).

Cost-effectiveness analysis divides total cost by units of outcome achieved; cost-benefit analysis compares monetized benefits to costs and reports net present value.

A Gantt chart sequences activities over time and shows dependencies; a critical path is the longest dependent sequence and sets the minimum project duration.

Sampling for baseline studies: probability sampling supports generalization; purposive sampling supports depth but not statistical inference.

A performance measurement plan specifies for each indicator: definition, data source, collection frequency, responsible party, and quality check.

A work breakdown structure decomposes deliverables into manageable work packages; each package has a single accountable owner.

RACI assigns roles: Responsible performs the work, Accountable approves, Consulted gives input, Informed is notified; exactly one Accountable per task.

Data collection protocols must state instrument, respondent, timing, and storage; pilot-test instruments before full rollout.

Adaptive management cycle: collect data, compare to targets, decide whether to continue, adjust, or stop, then document the decision.

Change control requires a written request, impact assessment on scope, schedule, and budget, and formal approval before implementation.

Capacity building for OBA includes training data collectors, standardizing definitions, and establishing a data flow map from field to report.

Implementation fidelity measures whether activities were delivered as designed; low fidelity weakens any claim that outcomes resulted from the intervention.

Monitoring is continuous and internal; evaluation is periodic and assesses merit, worth, or significance against criteria.

Variance analysis: variance = actual minus target; percentage variance = variance divided by target times 100.

Data quality dimensions: validity, reliability, timeliness, precision, integrity, and completeness.

Formative evaluation improves implementation during delivery; summative evaluation judges results after completion.

Evaluation criteria commonly include relevance, effectiveness, efficiency, impact, and sustainability.

Triangulation combines multiple data sources, methods, or evaluators to strengthen confidence in findings.

Earned value metrics: cost variance = earned value minus actual cost; schedule variance = earned value minus planned value; negative values indicate overrun or delay.

Informed consent requires disclosure of purpose, risks, benefits, voluntariness, and the right to withdraw without penalty.

Data protection principles: lawful purpose, minimization, accuracy, storage limitation, confidentiality, and security safeguards.

Conflict of interest exists when personal, financial, or political interests could bias professional judgment; disclose and recuse where objectivity is compromised.

Evaluation independence: evaluators should not report to managers of the program being evaluated when findings could affect their own interests.

Do no harm requires assessing risks to participants and communities, including stigma, privacy breach, and unequal distribution of benefits.

Audit trail: retain source documents, versions, and decision records so results can be independently verified.

Whistleblowing and reporting obligations: suspected fraud, falsification, or misuse of funds must be escalated through established channels.