How it works
A research system, not a prompt
Every Princeton Analytica report passes through the same eleven-stage pipeline: adaptive scoping, planning, current research with a registered source list, deterministic computation, analysis, independent fact-checking, writing, quality review and professional rendering.
Step one
The Report Diagnostic
Instead of forcing you to pick a rigid product, the diagnostic works like a consultation with a senior analyst. It asks one question at a time, adapts to your answers, infers what it can, and stops when the scope is clear.
What it establishes
- The decision you are making and the questions the report must answer
- Industry, company, role and who will read the report
- Entities, geography, time period and market scope
- Research depth, quantitative depth, forecasting, scenarios, benchmarking
- Source standard, charts and tables, uploads, report length, turnaround, recommendations
How it feels
- Conversational questions with examples and helper text
- Cards, multiple choice and free text where each fits
- Conditional follow-ups only when relevant
- A progress indicator and the ability to go back
- Typically 6–12 questions; never a wall-of-form
What you get at the end
- A recommended analysis and template
- A plain-language scope: what we research, questions answered, organizations, analysis, visuals, validation, deliverables
- Estimated depth and delivery
- An exact fixed price — before you create an account or pay
Steps two to eleven
The report pipeline
- 01
Intake
Your diagnostic answers become a structured research brief: the decision question, evidence-answerable research questions, canonical entity names, geography, time window and out-of-scope items. Uploaded documents are parsed privately.
- 02
Research planning
An analyst model writes the research plan: parallel workstreams by entity and topic, specific search queries, preferred source types, date filters, the quantitative tasks to compute, and an outline mapping sections to questions.
- 03
Research
Workstreams run in parallel against Amazon-operated web search. Every result is registered as a source with URL, publisher, publication date, retrieval time, type and authority score. Web content is treated as untrusted data.
- 04
Data extraction
Findings are normalized into an evidence registry: entity, metric, value, unit, date, quote, confidence and source reference. Conflicts between sources are identified, not hidden.
- 05
Quantitative analysis
Growth rates, CAGR, shares, rankings, averages, comparisons, projections and scenario math are computed in code from the evidence — never estimated by a language model — and reproduced in the appendix.
- 06
Analysis
An analytical model identifies patterns, trends, anomalies, risks, opportunities, implications and gaps, referencing evidence and computed metrics by id and labeling each finding as verified fact, derived, analytical conclusion, estimate or forecast.
- 07
Fact checking
A separate review verifies every number, date, name and claim against the evidence registry. Unsupported findings are removed; imprecise ones are corrected.
- 08
Report writing
Sections are written from the validated analysis in the voice of a strategy firm: conclusions first, selective bullets, callouts, KPI blocks, charts and tables by reference, citations only to registered sources.
- 09
Quality review
A critic pass scores factual grounding, citation completeness, analysis depth, numerical accuracy, clarity, usefulness, consistency, executive quality and formatting — and sends the draft back for revision when it falls short.
- 10
Rendering and delivery
The report is typeset into a professional PDF, stored privately, and made available in your dashboard with a short-lived download link. You receive an email when it is ready.
Source hierarchy
Which sources we trust, in order
- Government sources
- Regulatory sources
- Company filings
- Company websites and documents
- Official databases
- Universities and research organizations
- Established industry sources
- High-quality publications
- Other corroborating sources
Important claims are not accepted on one source when independent verification is reasonable. Where sources conflict, the report says so and explains which view is better supported and why.
Labels you will see
Fact, derived, conclusion, estimate, forecast
- Verified fact
- Directly supported by a registered source and cited.
- Derived calculation
- Computed in code from cited inputs; formula and inputs listed in the appendix.
- Analytical conclusion
- Our interpretation of the evidence — labeled as such.
- Estimate
- Assumption-based; assumptions stated.
- Forecast
- Projection with explicit growth assumptions and scenarios — illustrative, not a prediction.
- Recommendation
- Prioritized, specific, tied to findings, with what would change it.
Ready to turn a question into a decision?
Describe what you need to understand. The diagnostic scopes the research and quotes a fixed price in minutes.