Company filings tell you what management chooses to disclose. Market data and independent sector research tell you what those disclosures meant in practice: how accounting earnings translated into hard cash, how leverage behaved through economic cycles, and whether the underlying industry economics actually support a decade-long hold.
If your Gemini Notebook contains only 10-Ks and annual reports, you are feeding the AI a heavily curated narrative. To make Gemini a true analytical partner, you must provide it with an independent reality check.
This article outlines how to enrich your Gemini Notebook with decade-scale financial time series, concise macroeconomic context, and qualitative management assessments—and crucially, how to format these inputs so Gemini can cross-reference them without hallucinating.
Why Filings Alone Are Not Enough for AI
Annual reports are foundational, but they are inherently biased. Management naturally emphasizes strategic triumphs; accounting choices can legally obscure economic reality; and a single year’s footnotes rarely reveal if a high Return on Equity (ROE) is a permanent feature of a “moat” or a temporary illusion fueled by cheap debt.
To evaluate a five- to ten-year holding period, you need external time-series data that answers specific questions:
- Did reported top-line growth actually show up in free cash flow per share?
- Did ROE stay elevated through down-cycles, or did it collapse?
- Are dividends funded by organic cash generation or by taking on debt?
- How did the market price value these fundamentals over a full economic cycle?
When you place this historical data alongside the filings inside Gemini Notebook, the AI can detect discrepancies between what management said would happen and what actually happened.
Key takeaway: Treat market and sector data as an adversarial second source. Its job is to either corroborate or challenge the pristine narrative found in the annual report.
The 10-Year Quantitative Scoreboard
Do not rely on the LLM to search the live web for historical financial data—it will often hallucinate or pull mismatched figures. Instead, manually build a consistent metric set for every candidate in your notebook.
Gather a 10-year view of the following:
| Metric | Why it matters for long-term holds |
|---|---|
| Price (Adjusted) | Provides context for valuation history, drawdowns, and market sentiment. |
| EPS (Diluted) | Shows earnings power and profitability through full economic cycles. |
| Dividends per Share | Highlights cash returned to shareholders vs. capital retained for growth. |
| Revenue | Indicates top-line durability, market share trends, and cyclicality. |
| ROE / ROIC | Serves as a proxy for capital efficiency and the persistence of the company’s moat. |
| Debt-to-Equity | Tracks the leverage path, management aggressiveness, and refinancing risk. |
| Free Cash Flow (FCF) per Share | Reveals true economic cash generation after necessary reinvestments. |
Pro Tip: Always prioritize per-share metrics when evaluating earnings and cash flow. This prevents you (and the AI) from being fooled by aggregate growth that was actually funded by massive shareholder dilution.
Where to Find the Data
You do not need a Bloomberg terminal to compile this. Use free or low-cost public sources:
- Company IR Sites: Financial factbooks and historical Excel downloads.
- SEC EDGAR / National Registries: The ultimate source of truth when aggregators disagree.
- Macrotrends / Yahoo Finance: Great for pulling quick 10-year historical pricing and basic ratios.
- FRED (Federal Reserve Economic Data): Essential for macro series (rates, industrial production) if your theme is cycle-sensitive.
Always document your source. If two financial websites disagree on Free Cash Flow, reconcile the figure using the cash flow statement from the original 10-K.
Macro, Sector, and Qualitative Context
Numbers show the outcomes; qualitative sources explain how those outcomes were engineered. However, you must avoid overwhelming Gemini’s context window with 100-page industry reports. Synthesis is key.
1. The Environment (Macro & Sector)
Add two to three concise notes per theme covering:
- Sector Demand Structure: What drives volume and pricing over a decade?
- Regulation: Capital rules, environmental standards, antitrust risks.
- Structural Risks: Technological disruption, customer concentration, geopolitical exposure.
A useful sector note answers one question: What fundamental truths must remain intact for these businesses to compound over the next ten years?
2. The Pilots (Management & Strategy)
Include a highly curated selection of articles, interviews, or profiles focusing on:
- Management quality, track record, and incentive design.
- Capital allocation philosophy (M&A patterns vs. organic reinvestment vs. buybacks).
- Major strategic pivots.
When saving an article to the notebook, add a quick three-bullet abstract at the top: what it claims, what evidence it uses, and what you doubt about it. This frames the document for Gemini.
Formatting for Gemini: The “Summary Document”
Raw CSV files and messy HTML dumps confuse LLMs. To get the best analytical output, convert your external research into short, structured Markdown documents and attach them to your theme notebook.
Use these standardized templates:
Template 1: “10-Year Metrics Summary – [TICKER]”
Table: Year Price EPS DPS Revenue ROE D/E FCF/Share - Context Bullets:
- Peak-to-trough cycle observations.
- Notable leverage turning points.
- Dividend cuts/raises or accounting restatements.
- Data sources and retrieval dates.
Template 2: “Sector & Macro Note – [THEME]”
- Demand & Regulation: Two brief paragraphs on structural drivers.
- Risks: Bulleted list of structural threats.
- Scope Constraint: Explicit statement of what is out of scope for this analysis to keep the AI focused.
Example Gemini Prompt (Once data is uploaded): “Using only the uploaded annual filings and the ‘10-Year Metrics Summary’ documents for each ticker, compare how revenue growth translated into FCF per share over the last decade. Flag any company where earnings rose but FCF/share stagnated, or where leverage rose faster than cash generation. Cite specific years from the summary tables. If a figure is missing, state that it is missing—do not invent data.”
The Practical Enrichment Sequence
To systematically enrich your notebook, follow this checklist:
- Extract 10-year series for price, EPS, dividends, revenue, ROE, D/E, and FCF/share for all candidate stocks.
- Reconcile any contested data back to the primary annual filings.
- Draft 2–3 sector/macro notes hyper-focused on industry durability and structural risk.
- Curate 1–2 high-quality management profiles per company, prepending your own brief abstract to each.
- Format everything into the clean “Summary Document” templates outlined above.
- Upload the summaries to your existing Gemini Notebook alongside your thesis notes and filings.
- Run a data-check prompt to ensure Gemini recognizes the files before moving on to complex analysis.
Market context is how you keep fundamental AI analysis honest. By intentionally building a decade-long quantitative scoreboard and framing it with concise qualitative context, you give Gemini the exact inputs it needs to pressure-test a company’s narrative.