Forecast growth channel response saturation

Learn organization-specific channel saturation from resolved aggregate incrementality estimates: fit a likelihood-weighted Bayesian grid of Hill response curves, reserve the newest periods for honest validation against a linear baseline, expose posterior boundary misspecification and evidence failures, and return contribution and marginal-return distributions for proposed spend levels.

Sequential Bayesian & bandits Forecast CTOs Tech CEOs Investors

What it's for

Turns noisy channel ROAS into a validated saturation curve with visible uncertainty, marginal return and an explicit refusal when the curve cannot beat a simple later-period 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
half_saturation_grid_size integer ≥ 5, ≤ 101 Your calibration Optional
hill_shape_grid array of number ≥ 2 items Evidence Optional
max_detail_rows integer ≥ 1, ≤ 500 Numerical control Optional
maximum_boundary_posterior_mass number ≥ 0, ≤ 1 Your calibration Optional
minimum_validation_improvement number ≥ -1, ≤ 1 Your calibration Optional
minimum_verified_fraction number ≥ 0, ≤ 1 Your calibration Optional
resolved_incrementality_observations array of objects (7 fields) ≥ 8 items Evidence Yes
seed integer ≥ 0, ≤ 2147483647 Numerical control Optional
simulations integer ≥ 100, ≤ 100000 Numerical control Optional
spend_plans array of objects (3 fields) Evidence Yes
validation_period_count integer ≥ 2, ≤ 100 Your calibration Optional

Each resolved_incrementality_observations record

Field Type Required
channel_id string (non-empty) Yes
evidence_verified boolean Yes
id string (non-empty) Yes
incremental_contribution number Yes
incremental_contribution_standard_error number (> 0) Yes
incremental_spend number (≥ 0) Yes
period_index integer (≥ 0) Yes
Example input 
{
  "resolved_incrementality_observations": [
    {
      "channel_id": "events",
      "evidence_verified": true,
      "id": "events-response-0",
      "incremental_contribution": 0,
      "incremental_contribution_standard_error": 1,
      "incremental_spend": 0,
      "period_index": 0
    },
    {
      "channel_id": "events",
      "evidence_verified": true,
      "id": "events-response-1",
      "incremental_contribution": 11.428571428571429,
      "incremental_contribution_standard_error": 1,
      "incremental_spend": 5,
      "period_index": 1
    },
    {
      "channel_id": "events",
      "evidence_verified": true,
      "id": "events-response-2",
      "incremental_contribution": 20,
      "incremental_contribution_standard_error": 1,
      "incremental_spend": 10,
      "period_index": 2
    },
    {
      "channel_id": "events",
      "evidence_verified": true,
      "id": "events-response-3",
      "incremental_contribution": 26.666666666666668,
      "incremental_contribution_standard_error": 1,
      "incremental_spend": 15,
      "period_index": 3
    },
    {
      "channel_id": "events",
      "evidence_verified": true,
      "id": "events-response-4",
      "incremental_contribution": 32,
      "incremental_contribution_standard_error": 1,

Truncated for display — the full payload is 234 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.

Example output 
{
  "channel_diagnostics": [
    {
      "channel_id": "events",
      "failed_gates": [],
      "posterior_boundary_mass": 0,
      "posterior_mean_half_saturation_spend": 31.0009,
      "posterior_mean_hill_shape": 0.9915,
      "posterior_mean_maximum_incremental_contribution": 80.5559,
      "training_periods": 10,
      "validation_improvement": 0.9929,
      "validation_linear_baseline_rmse": 24.4212,
      "validation_model_rmse": 0.1743,
      "validation_periods": 2,
      "verified_fraction": 1
    },
    {
      "channel_id": "paid-search",
      "failed_gates": [],
      "posterior_boundary_mass": 0,
      "posterior_mean_half_saturation_spend": 18.8556,
      "posterior_mean_hill_shape": 0.9997,
      "posterior_mean_maximum_incremental_contribution": 98.022,
      "training_periods": 10,
      "validation_improvement": 0.9812,
      "validation_linear_baseline_rmse": 39.2213,
      "validation_model_rmse": 0.7387,
      "validation_periods": 2,
      "verified_fraction": 1
    }
  ],
  "configuration": {
    "half_saturation_grid_size": 21,
    "hill_shape_grid": [
      0.5,
      0.75,
      1,
      1.5,
      2
    ],
    "maximum_boundary_posterior_mass": 0.5,
    "minimum_validation_improvement": 0,
    "minimum_verified_fraction": 0.95,
    "seed": 5,

Truncated for display — the full payload is 87 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. 1 Order each channel's resolved aggregate experiment estimates by information time, reserve the newest periods, and require multiple positive spend levels with reported incremental-contribution standard errors.
  2. 2 Fit a weighted Hill saturation surface over half-saturation and shape grids, profile nonnegative maximum response and overdispersion, and normalize likelihood weights into a parameter posterior without pretending the grid is universal.
  3. 3 Compare posterior-mean predictions with a through-origin linear baseline on untouched later periods, gate boundary mass and evidence, then simulate parameter uncertainty for every proposed spend and report contribution quantiles, positive-response probability and marginal contribution per spend.

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.
  • Resolved observations are causally identified aggregate incremental effects with comparable contribution/spend units; channel mechanism and measurement epoch are stable; Hill saturation is adequate inside support; reported standard errors capture design uncertainty; later periods are untouched.
  • Posterior probability and adaptive selection are model-conditional; they are not substitutes for randomized propensities or guaranteed safety.
  • The forecast is conditional on the local experimental support and chosen curve family. Later-period underperformance or boundary mass means recalibrate or expand the model; do not extrapolate it as universal causal ROAS or use it for individual targeting.

Minimum evidence

  • resolved_incrementality_observations: at least 8 rows/items
  • spend_plans: 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

  • version-matched aggregate experiment estimates transformed into one non-overlapping channel-period response panel with untouched later validation periods and common currency/horizon
  • channel and measurement epoch, aggregate experimental eligibility, spend/contribution units, causal estimate validity, standard errors, chronological holdout, Hill shape and half-saturation support, validation/boundary/evidence gates, simulation precision and recalibration trigger

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": "learn organizationspecific channel saturation from resolved" }
  → finds "forecast_growth_channel_response_saturation"

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

gitrevio_capability_run
  { "capability_id": "forecast_growth_channel_response_saturation", "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

Audit growth incrementality experiment integrity

Audit aggregate randomized growth experiments before anyone trusts channel incrementality: enforce unique experimental units, nondegenerate logged propensities, both arms, control-spend discipline, baseline balance, spillover and evidence gates; then estimate propensity-weighted baseline-adjusted contribution, cluster-unit bootstrap uncertainty and incremental return on spend.

Causal inference & experiment design

Optimize multi period growth budget saturation

Allocate aggregate growth capital across channels and periods on coherent common scenarios while preserving channel-specific Hill saturation and carryover state: search discrete spend schedules, propagate contribution and unrestricted cash, and maximize expected net incremental value minus CVaR shortfall subject to total/period budgets, liquidity and contribution-probability gates, with exact certification or disclosed deterministic beam search.

Constrained optimization

Audit causal claim negative controls

Gate a causal effect claim using prespecified negative outcome/exposure controls, Benjamini-Hochberg multiplicity control, and an omnibus chi-square falsification test.

Causal inference & experiment design

Audit cluster randomization integrity

Audit cluster-randomized experiments for practical baseline imbalance and differential outcome observation, with cluster-size-weighted standardized differences and assignment permutation diagnostics.

Causal inference & experiment design

Audit joint outcome network integrity

Audit whether a company-specific Bayesian joint-outcome network is fit for reliance by validating point-in-time lineage, DAG and CPT completeness, effective support, protected-attribute exclusions, and strictly out-of-time outcome calibration against a baseline.

Sequential Bayesian & bandits

Audit staggered rollout identification

Audit staggered team-by-team adoption with not-yet-treated controls and require every simultaneous pre-period interval to fit inside a governed equivalence margin.

Causal inference & experiment design

See every tool in Causal evidence & experiments →

Ready to See Your Engineering work clearly?

Request a free demo