How CMO Rulebook works

1. What the engine is — and is not

CMO Rulebook is a deterministic rules-based system. Given the same inputs, it produces the same output every time. This is deliberate.

It does not use a language model to generate text. It does not hallucinate, interpolate, or produce probabilistic output. Every diagnostic card, boardroom narrative, and financial figure is the result of explicit logic applied to your stated KPI, signal direction, severity, business model, and company stage.

This matters commercially: a board-ready output needs to be citable. You need to be able to say "the analysis is based on these inputs, this logic, and this methodology" — not "the AI said so."

2. Inputs the engine uses

The diagnostic engine reads the following inputs to contextualise its output. Every input is optional beyond KPI and signal direction — but each one materially changes the output.

3. The boardroom readout engine

The boardroom readout is the primary output of the CMO module. It generates four fields: situation, judgement, priority, and caution — in the register of a senior marketing leader presenting to a board or CFO, not in the register of a data analyst reporting numbers.

3a. Single-KPI readouts

Each KPI in the registry has a dedicated readout object containing situation, judgement, priority, and caution language. The text is pre-written for each KPI and then modified by three runtime parameters:

• Severity word injection — slight / moderate / critical maps to adjective and adverb variants (e.g. "a slight movement" vs "a significant movement")
• Stage tone modifier — each of five company stages appends a specific urgency qualifier to the priority field (e.g. early-stage adds "validating the signal matters as much as acting on it")
• Business model label — the model type (SaaS, eCommerce, etc.) is inserted into the narrative to contextualise the recommendation

3b. Multi-KPI pattern detection

When more than one KPI is selected, the engine runs a pattern classifier before generating the readout. It maps each KPI to a funnel stage (top-of-funnel, acquisition efficiency, conversion, revenue, retention, margin, or brand/structural) and classifies the combination into one of seven named patterns:

Pattern-matched readouts produce a different narrative than single-KPI readouts — they identify the cluster as the primary commercial signal, not just the individual metrics.

3c. Advanced context modifiers

When advanced context inputs are provided, the engine post-processes the base readout:

• Signal age appends a persistence note to the priority field (e.g. "4+ months — tactical interventions alone will not resolve it")
• Competitive context (gaining) appends a competitive positioning note to the caution field
• Measurement confidence (uncertain or low) adds data-quality warnings to the output

4. The financial exposure model

When monthly revenue is provided, the engine computes a financial exposure estimate based on severity-weighted impact multipliers. These multipliers are stated assumptions — not predictions.

The exposure figure appears in the output with the multiplier disclosed. It is intended to frame the commercial stakes of the signal in board terms — not to predict actual financial loss. Where additional financial inputs are provided (margin %, CAC, LTV, volume), the engine computes derived metrics: payback period, LTV:CAC ratio, contribution margin, and engagement KPI weight at 30% default.

5. The budget simulator — Hill saturation curves

The budget simulator models channel spend efficiency using Hill (Michaelis-Menten) saturation curves. This is the same mathematical framework used in Marketing Mix Modelling (MMM) and academic media economics literature.

5a. The Hill curve formula

Marginal ROI at a given spend level is modelled as:

Where:

• revRoi0 — initial marginal revenue per £1 at near-zero spend. Represents the best-case return before saturation.
• satK — the monthly spend level (£) at which marginal ROI has halved. Derived from the typical budget threshold where diminishing returns become clearly visible for each channel.

5b. Default parameter basis

Each channel's default parameters are derived from published channel benchmarks and typical mid-market economics. They are starting points, not your actuals. The engine makes this explicit in-product with a mandatory disclaimer banner and requires users to override both ROI and incrementality before treating output as a commitment.

5c. Incrementality adjustment

The simulator adjusts attributed revenue by an incrementality percentage — the share of reported revenue that represents genuinely new demand versus cannibalised or organic. This produces a marginal contribution figure rather than a raw attributed ROAS figure. Default incrementality values are conservative estimates based on published holdout test findings; users should replace them with their own test results.

5d. Lag months

Each channel has a lag parameter representing the months between spend and revenue contribution. Paid media has zero lag (immediate attribution); SEO and brand have three to five month lags (compounding channel effects). The simulator accounts for lag in its horizon calculations.

6. The Data Roadmap — KPI tier progression

Each KPI in the registry is assigned a tier (1 through 5) representing the measurement infrastructure required to track it reliably. Tiers are based on the complexity and tooling prerequisites needed to measure each metric at the accuracy required for strategic decision-making.

Each KPI also carries an unlocks relationship — reaching a given KPI unlocks the prerequisites for higher-tier metrics. For example, tracking Traffic (Tier 1) unlocks Conversion Rate; tracking Conversion Rate unlocks CAC. This progression is surfaced in the Data Roadmap view as a dependency graph.

7. Challenge question tiers

For each KPI, the engine maintains three levels of diagnostic challenge questions. These progress from measurement validity (Level 0) through benchmarking (Level 1) to causal inference (Level 2).

• Level 0 — Measurement validity: Is the metric calculated consistently? Are there tracking changes, attribution window shifts, or data quality issues that could explain the move?
• Level 1 — Benchmarking: What benchmark are you measuring against? Is the current figure materially worse than historical trend, industry peer data, or investor expectations?
• Level 2 — Causal inference: Have you isolated the root cause? Can you separate demand effects from execution effects, or audience effects from creative effects?

The engine tracks which level a user has reached per KPI per session and advances automatically. This prevents repetition across sessions and progressively raises the analytical rigour of the diagnostic.

8. Limitations — what the engine does not do

• No live data: CMO Rulebook does not connect to GA4, ad platforms, CRMs, or any data source. All inputs are entered manually. The diagnostic is only as good as the signal you provide.
• No prediction: The simulator models efficiency curves at stated assumptions — it does not predict future revenue. Output is directional and scenario-based.
• No AI generation: Output text is pre-written and logic-selected — not generated by a language model. This makes it consistent and citable but not personalised beyond the input parameters.
• No multi-user or team state: The current version is single-session. Session sharing is available via URL, but there is no team workspace or shared state.
• Simulator defaults require override: All channel ROI, incrementality, and spend values are modelled defaults. They must be replaced with your own data before any output is treated as a commitment.