Forecast feature adoption revenue
Forecast feature adoption, revenue, and contribution with a grouped discrete-time hazard model trained on reconciled censored cohorts, required to beat a pooled-hazard baseline on later cohorts before posterior and capacity-constrained forecasts are decision-safe.
What it's for
Connects a feature rollout to a validated adoption curve and revenue range—then refuses the forecast when it cannot beat a simple later-cohort baseline.
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.
| Field | Type | Role | Required |
|---|---|---|---|
| current_features | array of objects (9 fields) ≥ 1 item | Evidence | Yes |
| historical_periods | array of objects (8 fields) ≥ 20 items | Evidence | Yes |
| horizon_periods | integer ≥ 1, ≤ 120 | Your calibration | Yes |
| max_detail_rows | integer ≥ 1, ≤ 500 | Numerical control | Optional |
| maximum_iterations | integer ≥ 10, ≤ 1000 | Numerical control | Optional |
| maximum_validation_calibration_error | number ≥ 0, ≤ 0.5 | Your calibration | Optional |
| minimum_log_loss_improvement | number ≥ 0, ≤ 1 | Your calibration | Optional |
| minimum_positive_contribution_probability | number ≥ 0.5, ≤ 0.999 | Your calibration | Optional |
| posterior_draws | integer ≥ 500, ≤ 20000 | Numerical control | Optional |
| ridge_precision | number ≥ 0.000001, ≤ 1000 | Your calibration | Optional |
| seed | integer ≥ 0, ≤ 4294967295 | Numerical control | Optional |
| target_adoption_fraction | number ≥ 0, ≤ 1 | Your calibration | Optional |
| validation_cohort_fraction | number ≥ 0.1, ≤ 0.5 | Your calibration | Optional |
Each current_features
record
| Field | Type | Required |
|---|---|---|
| adopted_accounts | integer (≥ 0, ≤ 10000000) | Yes |
| adoption_capacity_by_period | array of integer (≥ 1 item) | Yes |
| current_age_period | integer (≥ 0, ≤ 10000) | Yes |
| eligible_accounts | integer (≥ 1, ≤ 10000000) | Yes |
| fixed_future_cost | number (≥ 0) | Yes |
| future_enablement_fractions | array of number (≥ 1 item) | Yes |
| id | string (non-empty) | Yes |
| revenue_per_active_adopter_per_period | number (≥ 0) | Yes |
| variable_cost_per_active_adopter_per_period | number (≥ 0) | Yes |
{
"current_features": [
{
"adopted_accounts": 50,
"adoption_capacity_by_period": [
100,
100,
100
],
"current_age_period": 1,
"eligible_accounts": 500,
"fixed_future_cost": 20000,
"future_enablement_fractions": [
0.7,
0.8,
0.9
],
"id": "analytics",
"revenue_per_active_adopter_per_period": 200,
"variable_cost_per_active_adopter_per_period": 30
}
],
"historical_periods": [
{
"adopted_at_start": 0,
"age_period": 0,
"cohort_id": "feature-0",
"cohort_order": 0,
"eligible_accounts": 100,
"enablement_fraction": 0.5,
"id": "0:0",
"new_adopters": 10
},
{
"adopted_at_start": 10,
"age_period": 1,
"cohort_id": "feature-0",
"cohort_order": 0,
"eligible_accounts": 100,
"enablement_fraction": 0.5,
"id": "0:1",
"new_adopters": 10
},
{Truncated for display — the full payload is 228 lines.
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": [
"Historical rows preserve right-censored non-adopters in the period risk set, use one stable adoption definition and cadence, reconcile cohort cumulative adoption, and order cohorts by information available at launch.",
"A grouped logistic hazard with feature age, current saturation, and enablement is transportable from earlier to current features; the later-cohort log-loss and calibration gates test but cannot prove that assumption.",
"Coefficient covariance is Laplace/normal with Pearson overdispersion inflation; predictive simulation includes binomial adoption and hard enablement capacity but omits churn, seasonality, price response, network structure, and unrepresented regime change.",
"Revenue, variable cost, fixed cost, eligibility, and capacity are owner-governed feature economics; adoption or Git activity is not causal revenue, and a failed validation gate is an abstention even when numeric forecasts are returned."
],
"configuration": {
"horizon_periods": 3,
"maximum_validation_calibration_error": 0.05,
"minimum_log_loss_improvement": 0,
"minimum_positive_contribution_probability": 0.8,
"posterior_draws": 1000,
"ridge_precision": 1,
"seed": 47,
"target_adoption_fraction": 0.8,
"validation_cohort_fraction": 0.25
},
"decision": "feature_adoption_revenue_forecast_supported",
"feature_diagnostics": [
{
"adopted_accounts_at_forecast": 50,
"cumulative_contribution_p50": 54885,
"cumulative_revenue_p50": 88100,
"eligible_accounts": 500,
"feature_id": "analytics",
"horizon_adoption_fraction_p10": 0.33,
"horizon_adoption_fraction_p50": 0.384,
"horizon_adoption_fraction_p90": 0.438,
"probability_positive_contribution": 1,
"probability_target_adoption": 0
}
],
"method": "validated_grouped_hazard_feature_adoption_revenue_v1",
"model": {
"coefficient_names": [
"intercept",
"log_feature_age",
"current_adoption_fraction",
"enablement_fraction"
],
"iterations": 6,
"pearson_dispersion": 1,
"posterior_coefficient_medians": {Truncated for display — the full payload is 75 lines.
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 Build cohort-period risk sets retaining non-adopters, reconcile cumulative adoption, and order cohorts by launch information so the newest cohorts form an untouched chronological validation set.
- 2 Fit a ridge-stabilized grouped logistic hazard using feature age, current saturation, and enablement; inflate covariance for Pearson overdispersion and require later-cohort log-loss improvement plus calibration.
- 3 Simulate coefficient and binomial uncertainty through per-feature adoption capacity, then translate persistent active adopters into finance-owned cumulative revenue, variable cost, fixed cost, contribution, and probability gates.
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.
Assumptions & guardrails
- The likelihood or reward model, prior support, action logging, delayed outcomes, and any stationarity assumptions match the deployment process.
- Adoption definition, cadence, eligibility, censoring, cohort order, enablement, and no-churn horizon are stable; the hazard transports to current features and revenue/cost inputs are incremental contribution economics.
- Posterior probability and adaptive selection are model-conditional; they are not substitutes for randomized propensities or guaranteed safety.
- Adoption is not causal revenue and Git activity is not an adoption event; failed later-cohort validation is an abstention, while omitted churn, pricing, seasonality, networks, or regime change require model expansion.
Minimum evidence
- historical_periods: at least 20 rows/items
- current_features: at least 1 rows/items
- horizon_periods: required and organization-defined
How to validate it
Use chronological train/calibration/test windows, compare proper scores and decision value with a simple baseline, and recalibrate only from outcomes resolved after prediction time.
Calibrating it to your org
Same for everyone
The mathematical kernel, validation rules, method version, and JSON output semantics are organization-independent; no tenant-trained coefficients or company benchmark is embedded in the function.
Specific to you
- reconciled cohort-period adoption risk sets preserving right censoring and later-cohort chronological validation split
- adoption and eligibility definitions, cadence, cohort order, enablement semantics, no-churn horizon, adoption capacity, revenue/variable/fixed cost, validation thresholds, target, ridge, draws, and forecast version
Calibration workflow
- 1 Define the management decision, target outcome, aggregate unit, privacy boundary, cadence, and prediction/intervention horizon for this organization.
- 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 Estimate statistical parameters on training history, but obtain costs, utilities, risk tolerance, practical-effect thresholds, capacity, and policy constraints from accountable decision owners.
- 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 Deploy only if the returned decision clears evidence, overlap, calibration, robustness, and guardrail checks; warning, unsupported, schema-gap, and fallback decisions are abstentions.
- 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": "forecast feature adoption revenue and contribution" }
→ finds "forecast_feature_adoption_revenue"
gitrevio_capability_describe
{ "capability_id": "forecast_feature_adoption_revenue" }
→ returns the input schema and agent guidance shown on this page
gitrevio_capability_run
{ "capability_id": "forecast_feature_adoption_revenue", "arguments": { ... } }
→ returns the result shown aboveWorks in Claude Desktop, Claude Code, Cursor, Cline, Continue.dev, Goose and Aider. See the MCP server.
Related tools
Allocate capacity by marginal value
Allocate indivisible aggregate capacity across initiative-specific diminishing marginal-value scenario curves, activation thresholds, hard minimum commitments, unit cost, and portfolio CVaR with discrete next-unit value and explicit solver certainty.
Solve robust multiobjective portfolio
Solve a budgeted dependency-safe portfolio against both scenario-probability ambiguity and every vertex of a bounded stakeholder-preference simplex, using governed utility anchors and returning practically nondominated supported tradeoffs.
Allocate attention budget
Use exact knapsack optimization to allocate limited expert-review time by expected avoided loss.
Allocate capacity nash bargaining
Allocate discrete shared capacity by weighted Nash social welfare over concave team utility curves, with disagreement guarantees and a utilitarian counterfactual.
Calculate engineering runway
Compare three-point roadmap effort with three-point team capacity and expose unfunded commitments.
Calculate hidden work tax
Translate unplanned work, rework, incidents, and coordination into capacity and cost leakage.