Evaluating Stocks Like Products: The Scoring Engine
In a previous post I described the framework: a quantitative lens for timing and a qualitative lens for conviction. What I didn't show was the machinery underneath. Here it is.
Every stock gets two scores, each on a one-to-five scale. The quantitative score captures the financial picture. The product score captures the product picture. They combine through a geometric mean:
Total = √(Quantitative × Product)
Neither side can compensate for the other. A cheap stock with no moat fails. A great business at a terrible price fails. Both must hold. The rest of this post walks through each side and the decisions behind the numbers.
The quantitative engine
Four dimensions, each scored one to five, combined as a weighted average.
- Valuation (30%): PEG ratio (price-to-earnings divided by growth rate) and analyst upside. Are you overpaying relative to growth?
- Quality (30%): ROIC (return on invested capital, how much profit per dollar invested), gross margin, and a net cash bonus. Does the business generate real returns?
- Growth (20%): three-year revenue CAGR (compound annual growth rate) and forward EPS (earnings per share) growth. Is the company actually expanding?
- Risk (20%): beta (how much a stock moves relative to the market) and debt-to-equity. How leveraged and volatile is it?
The thresholds:
| Metric | 5 | 4 | 3 | 2 | 1 |
|---|---|---|---|---|---|
| PEG ratio | < 1.0 | 1.0 - 1.5 | 1.5 - 2.5 | 2.5 - 3.5 | > 3.5 |
| Analyst upside | > 30% | 15 - 30% | 10 - 15% | 0 - 10% | < 0% |
| ROIC | > 25% | 15 - 25% | 10 - 15% | 5 - 10% | < 5% |
| Gross margin | > 50% | 40 - 50% | 35 - 40% | 30 - 35% | < 30% |
| Revenue CAGR (3yr) | > 20% | 10 - 20% | 5 - 10% | 0 - 5% | declining |
| EPS forward growth | > 25% | 15 - 25% | 10 - 15% | 0 - 10% | declining |
| Beta | 0.7 - 1.0 | 1.0 - 1.2 | 1.2 - 1.5 | 1.5 - 2.0 | > 2.0 |
| Debt-to-equity | < 0.3 | 0.3 - 0.7 | 0.7 - 1.0 | 1.0 - 1.5 | > 1.5 |
A few decisions worth explaining.
- PEG over forward P/E. A company growing twenty percent at 25x earnings gets a PEG of 1.25, which scores a four. Forward P/E alone would penalize it for looking expensive.
- Debt-to-equity moved from Quality to Risk. It used to live alongside ROIC, but that double-counted leverage. I replaced it with a simple binary: if the company has more cash than debt, Quality gets a half-point bonus.
- Gross margin over free cash flow margin. Gross margin is independent of capital structure. FCF margin made asset-light businesses look artificially better.
The product side
Two dimensions, equal weight.
- Problem (50%): How severe and durable is the issue this company solves? Is it must-have or nice-to-have? What happens if nobody solves it? Is regulation mandating a solution?
- Defensibility (50%): Why can someone with more resources not just take the market? I lean on Hamilton Helmer's 7 Powers — scale economies, network effects, counter-positioning, switching costs, brand, cornered resource, and process power — and add a product lens for software: how painful is it to migrate away, does the user live inside the application all day, does the product get smarter with more data, is there a free alternative that solves eighty percent of the problem.
The thresholds:
| Score | Problem | Defensibility |
|---|---|---|
| 5 | Critical infrastructure. Regulation mandates it. Severe consequences if unsolved | 3+ strong Powers. No viable alternative exists. Moat widening |
| 4 | Serious problem affecting critical operations. Clear must-have. Regulation favors it | 2 strong Powers. Inferior alternatives. High switching costs |
| 3 | Real problem but customers can live with partial solutions. Important-to-have | 1 strong + 1 moderate Power. Competent alternatives but clear differentiation |
| 2 | Nice-to-have. Low urgency. Customers can postpone the solution | 1 weak Power. Multiple similar alternatives. Weak differentiation |
| 1 | Marginal or perceived problem. No urgency or regulation pushing it | No identifiable Powers. Commoditized market. Pure price competition |
This side cannot be automated. It requires reading earnings calls, SEC filings, and analyst reports. It is the slowest part of the process, roughly thirty to sixty minutes per company, and the part where conviction actually forms.
How they combine
The geometric mean penalizes imbalance. A stock scoring five on one side and one on the other gets a 2.24, not a 3.0. It does not enter the portfolio.
| Score | Classification | Position size |
|---|---|---|
| 4.5 or above | Exceptional | 15 - 25% |
| 3.5 to 4.4 | Strong | 8 - 15% |
| 3.0 to 3.4 | Acceptable | 3 - 8% |
| Below 3.0 | Does not enter | N/A |
Position sizes are not arbitrary. Base weight comes from the score total, adjusted by beta, with a cap at twenty-five percent. If two positions have weekly correlation above 0.7 over a three-year window, only the higher-scored one stays. Maximum thirty-five percent in any single theme, minimum five positions.
Before any of this runs, an ethical filter removes fossil fuels, weapons, tobacco, gambling, and private prisons. Binary gate, no scoring needed.
The framework in action
Two real examples from my evaluations.
MSCI
| Score | Detail | |
|---|---|---|
| Valuation | 4 / 5 | Revenue $3.1B, up 9.75% |
| Quality | 5 / 5 | 96% recurring revenue, strong margins |
| Growth | 4 / 5 | 11th straight year of double-digit EPS growth |
| Risk | 3 / 5 | High debt levels, fees tied to AUM |
| Quantitative | 4.10 |
Product — 5.00. $16.5 trillion in assets are benchmarked to MSCI indexes. ETFs literally cannot be constructed without an index to track, and pension funds are required by ERISA, MiFID II, and Solvency II to measure performance against standardized benchmarks. Problem: 5/5. Pension funds with trillions of dollars do not switch benchmarks — doing so means re-reporting decades of performance against a new yardstick and explaining to the board why. Hence 95% retention. The 57 years of historical index data are irreplicable, and every new ETF launched against MSCI reinforces it as the global standard. Defensibility: 5/5.
Total: 4.53 (Exceptional). Decent numbers, extraordinary moat. MSCI has switching costs, a cornered resource, network effects, brand, and scale all reinforcing each other. The qualitative side reveals what the financials alone cannot capture.
YOU (Clear Secure)
| Score | Detail | |
|---|---|---|
| Valuation | 5 / 5 | Analyst upside +32% |
| Quality | 5 / 5 | ROE 76%, gross margin 86%, net cash |
| Growth | 4 / 5 | Revenue up 17% YoY, $343M FCF |
| Risk | 4 / 5 | Beta 1.10, retention declining |
| Quantitative | 4.60 |
Product — 3.00. Skipping the airport security line is a real annoyance, but people have flown without CLEAR for decades. TSA PreCheck solves roughly 70% of the same pain at $78 for five years versus $209 per year, and United just cut the free CLEAR benefit for elite flyers — a quiet signal that even the airlines no longer see it as indispensable. Problem: 3/5. CLEAR has built genuine infrastructure — 38 million biometric profiles, physical eGates in 37 airports — that takes years to replicate. But that protects the supplier, not the user. A traveler cancels the $209 subscription online, walks back to the TSA PreCheck line, and nothing breaks. Retention is already declining to 86%. Defensibility: 3/5.
Total: 3.71 (Strong). Excellent financials hiding a fragile thesis. CLEAR has scale and a cornered resource on the infrastructure side, but no switching costs on the user side — and a moat that only works for the supplier is half a moat. A full tier below what the numbers alone would suggest.
What the framework favors
This is not a neutral screener. By design, it rewards businesses with high returns on capital, strong moats, and durable problems. That means it naturally gravitates toward quality compounders: companies that reinvest at high rates behind defensible advantages. It also has a structural bias toward asset-light models, since gross margin and ROIC tend to be higher when there is less physical infrastructure. Vertical SaaS companies score well through the PM lens.
What it structurally misses: turnarounds (weak moat today but improving), early-stage growth without an established moat (exceptional revenue growth but pre-profit and no defensibility), and deep value plays. These are intentional gaps, not oversights. I want conviction, and these are the businesses where I cannot build it.
What I don't know yet
This is version one with real money behind it. I am monitoring quantitative-only rankings alongside the combined score to see whether the product side actually adds alpha or just adds work. The weights might be wrong. Some thresholds are probably too generous, others too strict. I will find out as I run this through more market cycles.
What I do know is that the process itself has been valuable. Forcing myself to articulate why a business matters, what its moat actually is, whether the problem it solves is real and durable. That is where my conviction comes from. Not from the final number, but from understanding what the number represents.