Marketing ROI Analysis
Build an honest financial view of marketing performance. This method separates observed outcomes, platform-reported metrics, proxies, and modeled assumptions. It can prepare a proof pack or model specification; it does not alter budgets, tracking, customer records, or financial systems.
Read `references/model-contract.md` for the proof-level ladder, core economic definitions, and mode-specific requirements.
Inputs
Use available spend, funnel volumes, conversion rates, qualified lead or opportunity data, revenue, gross margin, sales-cycle timing, attribution method, and baseline. Ask for only the inputs that materially change the decision. If critical data is unavailable, provide a gap analysis and a model with assumptions visibly labeled rather than a fabricated ROI claim.
Read the relevant local references for attribution, model structure, and client-proof format. Read `references/host-operations.md` only when a user asks to retrieve data, create a spreadsheet, or update a host artifact.
Workflow
Choose the appropriate mode before analysis: a readiness audit when the event path is incomplete, a model when assumptions are necessary, or a proof pack when qualified outcome evidence supports a reviewable claim. Do not label a model as proof.
1. **Frame the decision.** Define the activity, audience, period, investment, success event, and decision the analysis must inform. 2. **Check measurement.** Map the event path from exposure through qualified outcome and revenue. Identify tracking gaps, attribution limits, consent boundaries, and duplicate-count risks. 3. **Build a baseline and counterfactual.** Use pre-period data, comparable cohorts, holdouts, or a clearly limited proxy. Do not treat correlation as incrementality. 4. **Calculate the economics.** Show spend, contribution margin, acquisition cost, conversion rates, payback, and sensitivity ranges using the defined formulas. Include all material costs that the decision treats as investment and reconcile all denominators and time windows. 5. **Classify the result.** Label each conclusion as observed, attributed, modeled, directional, or unknown. Explain the strongest alternative explanation. 6. **Recommend the next decision.** Continue, adjust, pause, improve measurement, or run an incrementality test. State the evidence needed to increase confidence. 7. **Prepare the proof pack.** Include a concise executive readout, model inputs, formulas, limitations, and source list suitable for review.
Output format
```markdown # Marketing ROI Readout ## Decision and bottom line ## Current proof level ## What can be proven today ## Measurement and attribution status ## Funnel and unit economics ## Baseline and counterfactual ## ROI range and sensitivities ## What is proven, modeled, and unknown ## Recommended next action ## Sources and assumptions ```
Quality rules
- Never combine incompatible time windows, cohorts, or revenue definitions. - Platform attribution is evidence, not a causal verdict. - A conclusion may claim only what its evidence maturity supports: measurement readiness supports a plan, modeled inputs support a scenario, qualified outcome evidence supports attribution, and incrementality evidence supports a causal claim. If the data cannot support ROI, return a readiness plan. - Do not present projected revenue or lifetime value as realized return. - Preserve source URLs and data extracts for non-trivial claims. - Propose host writes, exports, or spreadsheet updates separately from the analysis and report them only after verification.