Product documentation · v0.4.0 beta

Marketing Intelligence Commercial Platform

Revenue evidence. Decisions you can trace.

MICP helps B2B marketers, RevOps teams and agencies reconcile buyer-supplied marketing and CRM records, diagnose funnel uncertainty, and create evidence-linked decision briefs. It does not invent revenue lift or causal attribution.

Sample records are synthetic. The original customer files remain under the buyer's control.

Three skills in one workflow

01 / Reconcile

Revenue evidence

Check campaign spend and CRM opportunity values, separate recurring and total-contract bases, and compare optional buyer-supplied finance bookings.

02 / Diagnose

Funnel and attribution

Examine lead-to-MQL-to-SQL conversion, cohort maturity, campaign costs, first/last-touch evidence and source-data exceptions.

03 / Decide

Executive decision brief

Produce a readable brief with source-linked evidence, unresolved data issues and a cautious next-step plan.

Getting started

What you need.

A compatible agent that can run Python 3.11+ and use buyer-authorized file reading, writing and terminal commands. MICP does not require external AI tool purchases or network access for its packaged calculations.

Prepare an authorized and preferably anonymized export. The package accepts a normalized JSON dataset, or a folder of specified CSV files plus scope.json. Map custom CRM export columns to the documented names before execution.

  1. Start with the downloadable synthetic input to learn the supported structure.
  2. Load the purchased MICP pack in a compatible agent. Allow only the necessary local shell and file permissions.
  3. Ask the agent to follow the bundled shared/references/EXECUTION.md and execute Reconcile, Diagnose and Decide in order.
  4. Review the generated brief and the excluded/ambiguous record IDs before making decisions.

Input specification

Supported data format.

Use ISO dates (YYYY-MM-DD), one declared reporting currency and stable IDs. The pack does not convert foreign currencies. Missing inputs are not interpreted as confirmed zero activity.

FileRequired headers
spend.csvspend_id, date, campaign_id, amount, currency
leads.csvlead_id, campaign_id, created_at
events.csvevent_id, lead_id, stage, occurred_at
deals.csvdeal_id, lead_id, created_at, closed_at, amount, value_basis, currency, stage, origin_channel, source_campaign_id
touches.csvtouch_id, deal_id, campaign_id, occurred_at

Required scope fields: start, end, currency. Event stages: mql or sql. Opportunity stages: open, won, lost. Deal value basis: annual_recurring or total_contract. Optional finance rows use finance_id, deal_id, amount, currency, value_basis.

Deliverables

Files you receive.

  • decision-brief.htmlReadable executive decision brief
  • decision-brief.mdMarkdown decision brief
  • evidence.jsonCalculation details, assumptions and provenance
  • campaigns.csvCampaign spend and funnel costs
  • finance-comparison.csvFinance booking comparisons when supplied
  • attribution-evidence.csvFirst/last campaign relationships and joins
  • matured-cohorts.csvLead-stage maturity and conversion uncertainty
  • exceptions.csvFlagged and excluded record IDs

Reproducible example · synthetic data

An honest sample audit.

The bundled synthetic fixture contains USD 10,000 marketing spend, 4 leads, 3 MQLs, 2 SQLs, USD 30,000 closed-won annual recurring opportunity value, and USD 90,000 open total contract value. Finance records were not supplied in this fixture. The engine flags source exceptions and withholds a budget-reallocation recommendation until they are resolved.

These figures describe a fictional dataset. They are not customer results, recognized revenue, realized collections or independent evidence of marketing lift.

Interpretation

What the tool cannot claim.

Marketing-sourced and marketing-influenced values can overlap and must not be added together. Closed-won CRM opportunity value is not proof of cash collection or recognized revenue.

Finance reconciliation compares the booking rows you supply; it does not authenticate the underlying ledger. Wilson confidence intervals reflect uncertainty in observed conversion proportions, not causal attribution.

MICP does not offer native live CRM/API connections, automated advertising-budget changes, full sales-cycle prediction, guaranteed ROI, customer-specific financial certification or built-in PDF exports. Results depend on the quality of your inputs and the host agent's own data-handling rules.

Distribution and support

Access MICP.

MICP is available on Agensi as a paid skill pack for compatible AI agents. Purchase and installation are handled on Agensi; this page provides independent documentation and a synthetic example, not a separate checkout.

Version 0.4.0 beta. This page describes the currently implemented offline skill pack, not a deployed SaaS product.