Optimize sample size by decision value

What it's for

Stops teams from buying statistical precision that cannot change an economically governed decision—and shows exactly when a larger experiment pays for itself.

What you give it

Inputs split into evidence read from your connected systems, calibration your team owns, and numerical controls that affect precision but never the result's meaning.

{
  "affected_opportunities": 100000,
  "candidate_sample_sizes": [
    100,
    250,
    500,
    1000,
    2000
  ],
  "control_allocation_fraction": 0.5,
  "harm_effect_threshold": 0,
  "implementation_cost": 10000,
  "maximum_harm_probability": 0.1,
  "minimum_adoption_probability": 0.8,
  "minimum_net_experiment_value": 5000,
  "outcome_standard_deviation": 1,
  "prior_effect_mean": 0.01,
  "prior_effect_standard_deviation": 0.05,
  "sample_cost_per_observation": 10,
  "seed": 2026,
  "simulation_confidence_level": 0.95,
  "simulation_draws": 5000,
  "value_per_effect_unit": 100
}

What you get back

This is the actual output of running the example above — computed by the same function the platform calls, not an illustration.

{
  "assumptions": [
    "The incremental effect is normally distributed under a governed prior, the two-arm estimate is unbiased with known stable outcome standard deviation, and observations satisfy the declared allocation and effective independence assumptions.",
    "Adoption value is linear over affected opportunities, implementation and sampling costs are complete, and the effect/economics transport from the experiment population and horizon to the deployment decision.",
    "The posterior policy requires both probability of positive economic value and a harm-probability guardrail; expected sample information value is relative to that governed no-sample policy, not statistical power or significance.",
    "Monte Carlo common-prior simulation is finite; recommendation requires a positive confidence lower bound, while unmodeled bias, interference, attrition, sequential peeking, implementation failure, and prior misspecification remain outside the estimate."
  ],
  "configuration": {
    "affected_opportunities": 100000,
    "candidate_sample_size_count": 5,
    "control_allocation_fraction": 0.5,
    "harm_effect_threshold": 0,
    "implementation_cost": 10000,
    "maximum_harm_probability": 0.1,
    "minimum_adoption_probability": 0.8,
    "minimum_net_experiment_value": 5000,
    "sample_cost_per_observation": 10,
    "seed": 2026,
    "simulation_confidence_level": 0.95,
    "simulation_draws": 5000,
    "value_per_effect_unit": 100
  },
  "decision": "decision_value_positive_experiment_size_supported",
  "method": "normal_normal_bayesian_expected_sample_information_value_v1",
  "sample_size_diagnostics": [
    {
      "control_observations": 1000,
      "effect_standard_error": 0.04472136,
      "expected_regret_vs_governed_perfect_information": 131976.9691,
      "experiment_cost": 20000,
      "gross_expected_sample_information_value": 115558.4047,
      "net_expected_sample_information_value": 95558.4047,
      "net_value_confidence_lower_bound": 87545.5898,
      "net_value_monte_carlo_standard_error": 4088.246,
      "posterior_effect_standard_deviation": 0.03333333,
      "probability_matching_perfect_information_decision": 0.597,
      "probability_policy_adopts": 0.1868,
      "sample_size": 2000,
      "treated_observations": 1000
    },
    {
      "control_observations": 500,
      "effect_standard_error": 0.06324555,
      "expected_regret_vs_governed_perfect_information": 182438.7832,

How it works

Sequential Bayesian & bandits — Learn while deciding — update beliefs as evidence arrives and choose where the next unit of effort is worth spending.

1. 1 Freeze a transportable normal prior, effective outcome variance, feasible two-arm sample sizes, deployment opportunities, value per effect unit, implementation and observation cost, and posterior economic and harm decision rules.
2. 2 For each sample size, simulate true and observed effects, update the conjugate normal posterior, apply both governed posterior gates, and compare realized deployment utility with the governed no-sample policy.
3. 3 Subtract experiment cost, quantify Monte Carlo error, regret, decision agreement, EVSI and governed EVPI, and recommend only the highest-value candidate whose net-value confidence lower bound and minimum-value gate are positive.

Before you trust it

Every tool in the catalog ships with the conditions under which its answer is meaningful — and the conditions under which it should abstain instead of guessing.

Calibrating it to your org

Calibration workflow

1. 1 Define the management decision, target outcome, aggregate unit, privacy boundary, cadence, and prediction/intervention horizon for this organization.
2. 2 Build a tenant-scoped historical cohort using only information available before each prediction or decision; preserve zero periods, censoring, assignment probabilities, and unresolved outcomes when the method requires them.
3. 3 Estimate statistical parameters on training history, but obtain costs, utilities, risk tolerance, practical-effect thresholds, capacity, and policy constraints from accountable decision owners.
4. 4 Validate on later time windows or held-out aggregate units at the deployment grain, against a simple baseline and the function-specific validation strategy.
5. 5 Deploy only if the returned decision clears evidence, overlap, calibration, robustness, and guardrail checks; warning, unsupported, schema-gap, and fallback decisions are abstentions.
6. 6 Monitor realized outcomes, data drift, coverage, and decision regret; recalibrate at a governed cadence or after a detected regime/definition change, never merely because a stakeholder dislikes the result.

Call it from your AI

You don't wire up 388 tools in your MCP client. The GitRevio MCP server exposes 18 tools, three of which let an agent search the catalog, read a tool's schema, and run it — so the assistant finds this one on its own.

gitrevio_capabilities_search
  { "q": "choose a twoarm experiment sample size" }
  → finds "optimize_sample_size_by_decision_value"

gitrevio_capability_describe
  { "capability_id": "optimize_sample_size_by_decision_value" }
  → returns the input schema and agent guidance shown on this page

gitrevio_capability_run
  { "capability_id": "optimize_sample_size_by_decision_value", "arguments": { ... } }
  → returns the result shown above

Works in Claude Desktop, Claude Code, Cursor, Cline, Continue.dev, Goose and Aider. See the MCP server.

Related tools

See every tool in Finance & unit economics →