A.4.4 The Research Workflow

Basic Prerequisites: 03 · Screening Metrics (the screen that produces candidates) and Financial Statement Analysis.

---

1. Intuition & Practical Objective

The screen gives you a list of names; this page is what you do with them. It is the human half of the process - the part no screen automates - and it is where most of the value is created, because the numbers nominate but the business decides. The workflow has four stages, run in strict order:

1. Screen → a shortlist (mechanical, from 03).
2. Read the filings → the 10-K (annual) and the most recent 10-Qs (quarterly). The 10-K is the audited, complete picture; the 10-Qs are the freshest and reveal the trend between annual reports.
3. Understand the business → what it sells, to whom, how it makes money, who the competitors are, why the customer chooses it, and what could break it.
4. Check the red flags and the ownership → the governance and quality-of-earnings signals that tell you whether the numbers can be trusted and whether management is on your side.

The discipline: a screen survivor is a hypothesis, not a holding. The reading exists to falsify the hypothesis as cheaply as possible. You are looking for reasons to say no.

---

2. Mathematical Ground Truth & Derivations

The reading stage is qualitative, but the signals it checks have exact arithmetic. Four core tests:

Cash conversion - does profit become cash? The accrual identity is $\text{NI}=\text{CFO}+\text{Accruals}$, so the ratio

$$\frac{\text{CFO}}{\text{NI}}$$

should be comfortably above 1 over time. CFO < NI persistently means earnings are accrual-heavy (recognised before cash arrives) - the Sloan accruals warning. A screen threshold of $\text{CFO}/\text{NI}<0.8$ is a red flag.

Days Sales Outstanding (DSO) - are customers paying on time? With receivables $AR$ and revenue $S$ over $n$ days:

$$\text{DSO}=\frac{AR}{S/n}.$$

A jump of 10+ days year-over-year (receivables growing much faster than revenue) signals channel-stuffing, weak collections, or a customer in trouble. Formally, flag when $\text{recv growth}>\text{revenue growth}+\epsilon$.

Inventory build. Flag when inventory grows much faster than revenue $\left(\frac{\Delta \text{Inv}}{\text{Inv}}>\frac{\Delta S}{S}+\epsilon \right)$ - goods piling up ahead of a demand slowdown.

Ownership and insider alignment. A screen reason to buy - not a number to filter on mechanically, but a signal to weight heavily:

$$\text{insider ownership}\%\text{high and rising},\text{net insider buys}>0.$$

When managers buy their own stock with their own money, they are putting their capital where their confidence is; when they sell consistently, they are telling you something too. Combine with the F-score’s equity-issuance signal: net share issuance dilutes you and often signals management distrust of its own price.

---

3. Computational Implementation - the filing red-flag scanner

Runs on the standard library only. It takes the fields you actually pull from a 10-K/10-Q review and flags the accounting and governance warnings, then tallies the positive signals - the mechanical companion to the reading.

The verdict writes itself: four red flags against one green. The receivables-vs-revenue gap and the 16-day DSO jump together say the growth is being booked, not collected; the restatement says the history cannot be trusted; and the related-party transactions say check whether the company is transacting with itself. The insider buying is genuinely good news - but one green flag does not survive four red ones. This is the reading stage doing its job: a screen candidate rejected on evidence the screen never saw.

---

4. Failure Modes & First-Principles Breakdowns

1. Reading the summary instead of the filing. Press releases and sell-side summaries are marketing; the 10-K’s audited footnotes, risk factors, and MD&A are where the truth lives. Time spent on the summary is time not spent on the footnotes.
2. Trusting a restated history. Once a company restates, every prior-year number - and therefore every historical metric, CAGR, and trend - is suspect until independently reconstructed. Restatement is the loudest red flag in this whole folder.
3. Anchoring on the screen. Having selected a name, the analyst unconsciously looks for confirmation. The reading stage must be run as falsification: actively hunt for the reason this is a value trap.
4. Ignoring the qualitative moat. The numbers cannot tell you whether the competitive advantage is durable. That is Fisher’s “scuttlebutt” - talk to customers, suppliers, ex-employees; read the competitors’ filings. A cheap stock in a structurally declining industry is not value.
5. Confusing the time budget. A deep read is expensive; run it only on screened survivors, and time-box it. The screen’s whole purpose is to make the reading affordable (03).

---

5. References

• Graham, Benjamin: The Intelligent Investor (2003 annotated ed.)
• Fisher, Philip A.: Common Stocks and Uncommon Profits (Wiley reissue)
• Palepu, Krishna & Healy, Paul: Business Analysis and Valuation: Using Financial Statements (Cengage)
• Schilit, Perler & Engelhart: Financial Shenanigans (McGraw-Hill, 4th ed. 2020)
• O’Glove, Thornton L.: Quality of Earnings (Free Press, 1987)
• SEC EDGAR

---

6. Connected Graph Bridges

• Back: 03 · Screening Metrics · 02 · Graham Criteria · Index Hub
• Continue: 05 · Failure Modes & Practice · 06 · Advanced Extensions
• Depth: Accounting Quality & Red Flags · Financial Statement Analysis
• Data/source: Data Sources & Corporate Data (EDGAR, XBRL, point-in-time)