CGE · Commercial Governance Engine
Eight questions about the pipeline, answered with figures that can be audited.
CGE reads the commercial pipeline exactly as it lives in the CRM and returns a recurring verdict. Every pillar closes with an owned action and the quantified consequence of not taking it. No figure is written by a language model.
Why it exists
It's not an engineering exercise. It's more than thirty years in technology and business, applied to governing the commercial pipeline.
I've spent more than thirty years building technology businesses across Latin America, a good part of that time directing and auditing commercial pipelines at B2B companies. The traditional pipeline management discipline rests on three pillars: size, shape, and quality. They're useful, but they're not enough: they never explain whether the rep has something to close, whether the data can be trusted, whether the pipeline is moving on time, or whether the commercial inventory is healthy.
CGE keeps those three pillars, which today live inside the eighth, the one that synthesizes everything, and surrounds them with five more questions the traditional discipline never answered. That model of eight is what follows below.
The figure is always calculated by the deterministic engine. But the rules that decide what counts as healthy, what's at risk, and what demands action now are the same judgment I applied auditing real pipelines for more than thirty years, coded so artificial intelligence agents apply it on every run, not reinvented by a model each time.
Felipe Labbé · Founding partner of Ladibu
A forecast doesn't fall apart for lack of data. It falls apart because nobody trusts it.
The pipeline exists, it's in the CRM, and it's reported every month. What doesn't exist is a reading leadership can audit: coverage is calculated differently depending on who presents it, dead deals keep adding up, and an atypically large deal inflates the quarter without anyone flagging it.
The result is a meeting spent negotiating the number instead of deciding what to do about it.
The eight pillars
These aren't eight modules. They're the eight questions the commercial committee has to be able to answer.
- 01Is it enough?
Governable pipeline coverage against the period's outstanding quota, measured against the healthy level the client declared, not an industry average.
- 02Where does it reach?
Weight, attainment, and gap for each segment, using the names the client uses. The segment is never inferred: it comes from their own CRM field.
- 03Is it ready to close?
Coverage and maturity by rep, only when quota exists at that level and the data supports it. Someone on a ramp is read against a different bar.
- 04Is it reliable?
Deals outside the client's typical range go into quarantine for human review, plus opportunities stalled past the threshold and overdue close dates.
- 05Is it on time?
Cycle measured against the client's own threshold, derived from their won deals. A long declared cycle doesn't read as a bottleneck.
- 06Is the inventory healthy?
Commercial levers ranked by severity: conversion, concentration, maturity, stagnation. No single score, because an average hides the lever that matters.
- 07Can the data be governed?
Not the percentage of complete records as a goal, but the concrete list of high-value opportunities that today can't be assigned, segmented, or measured.
- 08Do they sound together?
Harmony across six dimensions: size, shape, and quality, the three classic pillars of pipeline management, plus velocity, balance, and discipline, the three that model never measured. The verdict doesn't come from averaging the six, it comes from whichever one breaks the rhythm. If coverage depends on two deals, the verdict says so.
Every pillar closes with an owned action and the cost of not taking it.
A dashboard informs. A verdict obligates. That's why every pillar produces five components, and the last one, the "or else," always hangs from a figure the engine calculated, never from an estimate.
Example action · Pillar 4, quarantined deals
Validate or confirm in writing the deals quarantined for size.
The flagged opportunities, by name, account, and amount.
The responsible area. The bottleneck is named by area or stage, never by person.
This week.
Period coverage drops from 2.8x to 1.6x if it's not validated.
Figures run in monospace because the engine calculated them. Commentary runs in serif. You can tell them apart at a glance.
The difference
A dashboard informs. A verdict obligates.
How it's built
Two layers separated by design: one calculates, the other explains. The one that explains cannot make up a figure.
This separation is the reason the verdict can be taken to senior leadership.
Layer 1
The engine
Deterministic: the same information always produces the same number, and every number has a visible formula. Calculation parameters live per client, isolated by their own key.
Layer 2
The narrative
Turns the result into a readable verdict and answers questions about that run. It has no permission to produce its own figures: if the engine didn't calculate it, it doesn't appear in the text.
The rules that don't get negotiated
A verdict is only useful if the sales team cannot take it apart, and doesn't want to.
- The threshold belongs to the client, not the average
The cycle, the typical deal range, and the healthy coverage level come from the client's own operation or what they declared. There's no global bar.
- No naming individuals
The bottleneck is named by area or by stage. A verdict that accuses individuals turns into internal politics and stops being used by the second month.
- What's missing is declared, not filled in
When a parameter isn't calibrated, the verdict says so instead of presenting a default value as if it were a finding about the client.
- Not punitive
Someone starting on a ramp isn't measured against a seasoned rep's bar. The verdict distinguishes someone starting out from someone slipping.
It launches in weeks, not quarters.
- 01Configuration
One session to capture currency, fiscal year, pipeline stages with the exact CRM names, segments, thresholds, and quotas. Whatever isn't available starts at a value flagged as not calibrated.
- 02Private validation
A first run nobody else sees, to review what tripped the guardrails and whether the calibration reads the business reality correctly.
- 03Official run
The parameters that validation surfaced get adjusted, and the first deliverable verdict is produced.
- 04Recurring operation
Every run is just loading the updated pipeline, on the cadence of the sales committee. Each client runs isolated.
CGE exists because the same question kept repeating from one client to the next.
It's not a product we set out to sell. It's the instrument we built so we would never again argue over a forecast without evidence, and it's made available to organizations where that problem is the one getting in the way.
Next step
Let's look at your pipeline with your business's own calibration.
A configuration session and a private validation run. Before anything is shown to anyone, you see what tripped the guardrails and how well it reads your reality.
Write to us