Portfolio Analysis Process
Choose the right analysis path, assemble the source packet, and hand off a decision-ready output without duplicating specialist methods.
Executive Summary
Start every portfolio analysis by naming the decision: product quality, marketing allocation, KPI monitoring, experiment reliability, or company intelligence. Each decision has a different source floor, method owner, and output. This page is the router; the specialist guides own the math, diagnostics, and deep execution.
Reviewed May 20, 2026 against the local ROAS Analysis Pipeline, Company Analysis Standard v1.1.0, and portfolio diagnostic template. The ROAS methodology file contains future-dated metadata; this page treats it as a local methodology reference only and does not surface that date as a current factual claim.
1. Decision Router
The first choice is the analysis track. Do not blend tracks just because they share data. Blended questions create blended recommendations and make it impossible to know which method should be trusted.
| Track | Question It Answers | Minimum Source Floor | Output | Specialist Owner |
|---|---|---|---|---|
| Product Review | Is the product improving or degrading after releases? | Version history, isolated cohort data, product change notes. | Product quality read, release impact, open confounds. | ROAS pipeline + product-marketing analysis. |
| Marketing Allocation | Where should we scale, pause, or investigate spend? | All-geo/all-network MMP export, revenue definition, platform fees. | Kill/scale recommendations and budget implications. | ROAS analytics and marketing report. |
| KPI Monitoring | Did a metric move enough to require action? | Stable metric series, baseline window, known release/events calendar. | Alert, watch, or no-action status. | Control chart diagnostics. |
| Experiment Read | Is the test result reliable enough to decide? | Hypothesis, variants, sample size, guardrails, duration. | Scale, stop, extend, or redesign decision. | Statistical significance guide. |
| Company Diagnostic | What do we know, what changed, and what can we do with it? | PROFILE, Affinity notes, diagnostic findings, company data sources. | Forward-looking diagnostic findings and next support actions. | Company Analysis Standard + diagnostic template. |
Default split
| If the user asks... | Route to... | Reason |
|---|---|---|
| "Is the game better?" | Product Review | Needs isolated product signal, not blended channel mix. |
| "Where should we spend?" | Marketing Allocation | Needs full market/channel view and revenue economics. |
| "Is this drop real?" | KPI Monitoring | Needs process-control thinking before diagnosis. |
| "Did the test win?" | Experiment Read | Needs power, confidence, and guardrail checks. |
| "What is the portfolio support plan?" | Company Diagnostic | Needs source-backed synthesis and opportunity framing. |
2. Source Packet Before Analysis
No analysis starts until the source packet is explicit. For company-specific work, read local company docs and check Affinity before touching files or writing recommendations.
| Source Class | Required For | What It Must Answer | Blocker If Missing |
|---|---|---|---|
| Company context | All company-specific analysis. | Company status, GP owner, product, data sources, known confounds, Affinity history. | Analysis cannot start. |
| Data manifest | MMP, ROAS, retention, and KPI work. | Source system, extraction date, revenue type, timezone, platform fee treatment. | No projection or budget recommendation. |
| Version history | Product review and regime attribution. | Release dates, hotfixes, status, expected impact, known bugs. | No version-impact claim. |
| Event calendar | KPI monitoring and diagnostics. | Launches, campaigns, pricing changes, outages, creative refreshes. | No causal explanation. |
| Review owner | External or founder-facing output. | Who approves facts, tone, recommendations, and open questions. | No send/share. |
Security Gate
Raw exports and private company metrics stay inside the authorized company or analysis workspace. Public guide pages can explain methods, but they should not embed company-specific incidents, raw exports, private LP/fund metrics, or CRM notes.
3. Five-Step Operating Process
| Step | Action | Pass Standard | Common Failure |
|---|---|---|---|
| 1. Intake | Name the decision, audience, source owner, and deadline. | One decision sentence under 20 words. | Starting with a file dump. |
| 2. Validate | Check schema, completeness, revenue definition, and obvious data bugs. | Known gaps are documented before analysis. | Projecting from broken or mixed data. |
| 3. Isolate | Choose the cut that matches the question: product, channel, geo, platform, cohort, or time window. | Confounds are either controlled or named. | Blending product and marketing signals. |
| 4. Analyze | Run the specialist method and preserve assumptions. | Output includes method, source, confidence, and limitations. | Producing a number without a decision implication. |
| 5. Document | Write the finding, opportunity, next action, and unresolved questions. | Every insight answers "what can we do with this?" | Leaving insight only in a chat thread. |
4. Output and Handoff Rules
The output format should match the decision and audience. A founder-facing diagnostic, an internal investment dossier, and a raw analytics run should not collapse into the same document.
| Output | Audience | Canonical Role | Next Route |
|---|---|---|---|
| Product Review | Founder, product lead, Transcend advisor. | Explain whether product changes improved quality or monetization. | Product-marketing analysis or ROAS diagnostic. |
| Marketing Report | Founder, UA lead, Transcend advisor. | Translate cohort economics into spend decisions. | ROAS analytics or experimentation system. |
| Diagnostic Findings | Portfolio company team and GP owner. | Capture what changed, why it matters, and what support action follows. | Company folder diagnostic record. |
| Company Analysis | Investment or strategy team. | Maintain verified intelligence and evaluative context. | Investment assessment or strategic memo. |
| Open Questions | Data owner or founder. | State missing data and why it blocks interpretation. | Follow-up request in the same channel. |
5. MECE Boundary With Specialist Guides
| This Page Owns | Specialist Page Owns | Boundary |
|---|---|---|
| Choosing the analysis track. | Product-Marketing Analysis | Product-marketing owns state-machine diagnosis and funnel interpretation. |
| Requiring a source packet. | ROAS Diagnostic | ROAS owns metric decomposition and root-cause workflow. |
| Routing noise vs action questions. | Control Chart Diagnostics | Control charts own monitoring thresholds and special-cause logic. |
| Routing experiment questions. | Statistical Significance | Stats owns power, confidence, sample size, and test interpretation. |
| Document handoff discipline. | Company Analysis Standard | Company analysis owns dossier structure and verified intelligence. |
6. Internal Source References
These local references are the source floor for this router. Do not cite them as external evidence; use them to choose the right internal workflow and artifact owner.
| Reference | Use in This Guide |
|---|---|
skills/skills/ROAS_ANALYSIS_PIPELINE.md | ROAS stages, source manifest, producer/consumer boundaries, and methodology ownership. |
investment-assessment/COMPANY_ANALYSIS_STANDARD.md | MECE split between PROFILE, Company Analysis, and Investment Assessment. |
skills/skills/portfolio-diagnostic/templates/DIAGNOSTIC_FINDINGS_TEMPLATE.md | Forward-looking diagnostic findings and "what can we do with this?" framing. |
guides/product-marketing-analysis/, guides/roas-diagnostic/, guides/control-chart-diagnostics/, guides/statistical-significance/ | Specialist pages that own methods after this router chooses the track. |