Why Your Stock Score Might Be Lying to You

You look up a stock on your valuation tool of choice. It gives you a score out of 5, or a set of coloured bars, or a pentagon graphic with shaded segments. The "Value" dimension is green. The algorithm says it's undervalued.

You buy it. It drops.

This isn't always the algorithm's fault. But sometimes it is — and the reason is surprisingly simple: most stock scoring tools use the same methodology regardless of what kind of company they're analyzing.

A SaaS company and a regional bank and a gold miner are scored by the same formula. The same P/E benchmark. The same debt-to-equity threshold. The same growth expectations.

That is a problem.

Why Sector Context Changes Everything

Let's take three concrete examples of metrics that mean completely different things depending on the business:

P/E Ratio

For a consumer staples company — Colgate, Procter & Gamble — a P/E of 25x is high but defensible given the stability and predictability of earnings. For a SaaS company growing 30%/year, a P/E of 25x might be genuinely cheap. For a cyclical miner at the peak of a commodity cycle, a P/E of 8x might still be expensive because earnings are about to collapse.

A generic "P/E is high = overvalued" score applied to all three is wrong for at least two of them.

Debt-to-Equity

Banks and financial companies are supposed to carry high leverage — it's the nature of the business model. A bank with a debt-to-equity of 8:1 is not financially reckless; it's a bank. Apply the same threshold you'd use for a manufacturing company and you'll flag almost every financial institution as dangerously leveraged — which is nonsense.

Insurance companies, REITs, and utilities all have capital structures that look alarming through the lens of a generic algorithm calibrated on tech and consumer companies.

Free Cash Flow Margin

Early-stage growth companies — think Amazon in 2014, Uber in 2019 — often have low or negative FCF margins because they're investing aggressively in growth. A scoring system that marks them down for "weak free cash flow" is potentially flagging the very reinvestment that makes them valuable. Meanwhile, a mature company with 25% FCF margins but zero growth opportunities might score beautifully while quietly compounding value destruction.

What Simply Wall St's Snowflake Actually Measures

Simply Wall St is the most widely used stock scoring tool among retail investors, and its Snowflake graphic — five axes (Value, Growth, Past Performance, Financial Health, Dividends) — is genuinely intuitive and useful as a quick overview.

But the methodology has a documented weakness: it uses a largely generic algorithm across all sectors.

The Value axis, for example, uses a single DCF based on analyst consensus estimates. This works reasonably well for stable, predictable businesses. For a commodity company whose earnings swing 300% between cycle peaks and troughs, using analyst consensus (which itself has wide uncertainty) to power a DCF produces a confidence interval so wide it's nearly meaningless.

Simply Wall St documents their methodology on GitHub, which is commendable for transparency. But the documentation is separate from the product — in the UI itself, you see a score, not the reasoning. A user who doesn't dig into the methodology will treat a "good Value score" on a cyclical company the same way they'd treat it on a consumer staples company. They shouldn't.

This isn't a criticism unique to Simply Wall St. It's endemic to single-score tools. The moment you reduce a complex, sector-dependent analysis to one number, you lose the nuance that makes valuation meaningful.

The Three Categories of Methodology Errors

When a generic stock scoring algorithm gets it wrong, it usually fails in one of three ways:

1. Using the wrong metric for the sector

Example: Applying P/E to banks. Banks' earnings are heavily influenced by loan loss provisions, which are partly discretionary. A bank can boost earnings by releasing reserves built in bad times, making P/E look attractive precisely when the underlying risk is elevated. Experienced bank analysts use Price-to-Book (P/B) and Return on Equity (ROE) instead. A generic algorithm that weights P/E heavily for banks gives misleading signals.

Example: Using EV/EBITDA for financial companies. EBITDA is almost meaningless for banks and insurers because interest expense is their core cost of business, not a financing cost to be added back. EV/EBITDA is a useful metric for industrial companies; it's the wrong tool for financials.

2. Using the right metric but the wrong benchmark

Example: P/E benchmarked against the broad market for a high-growth software company. If the S&P 500 trades at 22x earnings and a SaaS company growing 30%/year trades at 45x, a generic algorithm might flag it as "overvalued." But compared to other high-growth SaaS companies, 45x might be exactly right or even cheap. The relevant benchmark is the peer group, not the index.

Example: Debt-to-equity for utilities. Utilities carry significant debt because they fund long-lived infrastructure assets that generate stable, regulated returns. A utility's 3:1 D/E is not comparable to a consumer electronics company's 3:1 D/E. The utility's debt is matched by long-duration contracted cash flows; the electronics company's isn't.

3. Anchoring to current earnings when they're abnormal

Example: Cyclicals at the earnings peak. Steel, mining, oil, and chemical companies see earnings swing dramatically through commodity cycles. A steel company that earned $15/share at cycle peak looks cheap at 6x P/E — until the cycle turns and earnings revert to $3/share, at which point the stock was actually trading at 30x normalized earnings when you bought it.

Good valuation practice for cyclicals uses mid-cycle or normalized earnings, not trailing twelve months. Generic algorithms typically use TTM earnings because they're available. This is the value trap mechanism in its purest form.

What to Actually Look For in a Valuation Tool

The goal isn't to find a tool that gives you a better-looking score. It's to find a tool that forces you to think correctly about the stock in front of you.

1. Does it tell you which methods apply to this company type?

A valuation tool should flag when a metric doesn't apply — or at least when to be skeptical of it. Running P/E on a bank, EV/EBITDA on an insurance company, or DCF on a company with unpredictable cash flows should come with a caveat, not a confident score.

2. Does it show you the assumptions behind the number?

A fair value estimate without visible inputs is an opinion dressed as analysis. If a tool tells you a stock's intrinsic value is $85 but won't show you the growth rate, discount rate, and projection period that produced it, you can't evaluate whether you agree with the analysis.

3. Does it answer the reverse question?

"At the current price, what growth rate is the market implying?" is often more useful than "what is the intrinsic value?" The reverse DCF anchors the valuation conversation to a concrete, falsifiable belief: do you think this company will grow at 15%/year or not? Yes or no — now you have something to evaluate.

4. Does it use multiple methods and look for convergence?

A single valuation method is a single hypothesis. Convergence across multiple methods — DCF, PEG, EV/EBITDA, historical average P/E, peer comparison — provides much more conviction. Divergence between methods is equally informative: it tells you there's genuine uncertainty about how to value this company, which should affect your position sizing.

The Right Use of Stock Scores

None of this means visual stock scoring tools are useless. The Snowflake is genuinely useful as a first screen — a way to triage a watchlist quickly and flag candidates for deeper work. If a stock scores poorly across multiple dimensions, that's a signal to look closer (though not necessarily a signal to avoid).

The problem is treating the score as the analysis rather than the starting point for analysis.

A green "Value" bar doesn't mean the stock is cheap. It means the algorithm thinks the stock looks cheap by its particular methodology, applied across all sectors equally. Whether that applies to the specific company you're looking at depends on things the algorithm doesn't know: the sector's appropriate metrics, the company's position in the cycle, the reliability of the earnings forecasts powering the DCF.

That gap — between the algorithm's confidence and the actual complexity of the valuation — is where investors get burned.

Frequently Asked Questions

Is Simply Wall St accurate?

Simply Wall St is a useful first-pass screening tool with an intuitive interface and broad global coverage. Its accuracy varies significantly by company type: it works reasonably well for stable, predictable businesses with consistent earnings. It is less reliable for cyclicals, financials, early-stage growth companies, and businesses in sectors where standard metrics don't apply. The fundamental issue is that its algorithm treats all sectors similarly.

What's wrong with using the same valuation formula for all stocks?

Different industries have different appropriate valuation metrics, peer benchmarks, and earnings characteristics. Applying P/E to a bank is wrong because bank earnings are heavily influenced by discretionary provisions. Using EV/EBITDA for financials is wrong because interest is their core business cost, not an add-back. Using TTM earnings for cyclicals is wrong because they're abnormally high or low relative to mid-cycle. Generic formulas make all of these errors simultaneously.

How should you value a bank vs. a tech company?

Banks: focus on Price-to-Book (P/B), Return on Equity (ROE vs. cost of equity), and Net Interest Margin trajectory. Avoid P/E and EV/EBITDA as primary metrics.

Tech/SaaS companies: focus on EV/Revenue or EV/ARR for early stage, transitioning to EV/FCF and DCF as they mature. P/E is useful but PEG (P/E relative to growth rate) provides more context.

Cyclicals: use normalized or mid-cycle earnings rather than TTM. EV/EBITDA is more stable than P/E through cycles.

What does sector-calibrated valuation look like in practice?

It means flagging which metrics apply and which don't, and adjusting benchmarks to the peer group rather than the broad market. For a regional bank, a "good" P/B is different from a "good" P/B for a tech company. For a SaaS company, a "high" P/E is acceptable at 40x with 30% growth; it isn't at 40x with 5% growth. Sector calibration is about applying the right test, not just applying the standard test.

The Bottom Line

Stock scoring tools that give you a simple output — a pentagon, a bar, a number out of 5 — are useful starting points. They are not reliable endpoints.

The failure mode is treating a generic algorithm's green light as permission to skip the actual valuation work. The bank that looks cheap on P/E, the cyclical that looks cheap on TTM earnings, the SaaS company flagged as "expensive" because its P/E exceeds a market benchmark — all of these are cases where the score obscures more than it reveals.

Good analysis requires knowing which metrics apply to the company you're evaluating, what the relevant peer benchmark is, where earnings are in the cycle, and what growth rate the current price implies. A score doesn't give you any of that. Understanding the methodology does.

The most important question to ask any valuation tool: "What assumptions are you making, and do they apply to this specific company?"

Try EvidInvest — multi-method fair value analysis with sector context and visible assumptions →