The Question
A biotech startup burning cash on R&D and a mature utility paying steady dividends shouldn't be valued the same way. Yet most quantitative models apply uniform metrics—P/E ratios, PEG, or price-to-book—across the entire market, forcing dissimilar businesses into the same analytical framework.
Peter Lynch popularized PEG (Price/Earnings-to-Growth) for evaluating growth stocks, not for screening everything. Applying growth-stock valuation to cyclical industrials or value metrics to early-stage tech distorts signals and drowns alpha in classification noise.
The challenge is particularly acute for stocks outside major indices. Small and mid-cap companies often lack analyst coverage—60-70% of CSI 1000 constituents have sparse or no consensus forecasts. Traditional quant strategies that depend on analyst estimates simply exclude these names, ignoring a massive segment of the market where information inefficiencies are greatest.
Can you build a systematic framework that classifies companies by growth characteristics, applies appropriate valuation models to each class, and fills forecast gaps for under-covered stocks—all at scale?
The Approach
We construct a three-class growth taxonomy using SAM industry data and Financial Notes segment-level metrics:
1. Stable Growth Stocks
Industries with mature business models, steady participant counts, and predictable long-term growth. Think utilities, consumer staples, established industrials. For these, we use PE / (G - 2YCAGR) where G is near-term growth and 2YCAGR is the two-year compound growth rate. This penalizes companies where current growth is just mean-reversion from a prior downturn, favoring sustained growers.
2. Fast Growth Stocks
Industries in expansion phases with increasing participants, high revenue volatility, and unclear competitive equilibria. Technology, biotech, emerging consumer categories. We use PE / (G - YoY) which focuses on acceleration—is growth speeding up or slowing down? Year-over-year comparisons capture momentum better than multi-year averages for companies in flux.
3. Cyclical Stocks
Capital-intensive industries (materials, commodities, heavy industrials) where profitability swings violently with macro cycles and earnings-based metrics become unreliable. We use PB / (Sales Growth - 2YCAGR)—book value anchors the denominator during earnings volatility, while sales growth measures business momentum independent of margin compression/expansion.
The classification is dynamic—industries move between categories as their characteristics evolve. We use SAM's granular product taxonomy (12-layer hierarchy) to classify at the business-segment level, not just the company level. A diversified conglomerate might have stable growth segments (mature products) and fast growth segments (new product lines) weighted by actual revenue contribution from Financial Notes.
Filling the forecast gap: For stocks without analyst coverage, we train models on companies with coverage, using SAM product similarity and historical financial patterns as features. If a small-cap solar equipment manufacturer lacks forecasts, we synthesize estimates from covered peers making similar products (SAM product nodes) with similar historical ROE/margin trajectories (Financial Notes data).
This isn't naive peer averaging—it's a machine learning approach that learns which product-level and financial characteristics predict analyst consensus, then applies that mapping to uncover stocks. The model updates quarterly as new earnings are reported.
The Finding
Applying lifecycle-appropriate valuation metrics rather than uniform screening delivered 21.89% annualized returns in long-only portfolios and 19.40% annualized returns in long-short strategies with a Sharpe ratio of 2.57. More importantly, the Calmar ratio (return/max drawdown) reached 5.72—indicating exceptional risk-adjusted performance with controlled downside.
Before filling forecast gaps, the strategy already outperformed: 17.84% annualized returns, 18.09% long-short alpha. But extending the framework to under-covered stocks using synthetic forecasts added nearly 4 percentage points of annual alpha.
The alpha came from three sources:
-
Classification precision: Applying PEG to cyclical stocks or P/B to growth stocks creates noise that obscures real value signals. Lifecycle-based classification removes this noise. The strategy's IC (information coefficient) across all classes exceeded traditional single-metric approaches by 40-60%.
-
Under-covered alpha: Stocks without analyst forecasts are informationally inefficient—less institutional attention, more behavioral mispricing. But most quantitative models can't touch them due to missing data. By synthesizing forecasts from product-level comparables (SAM) and financial patterns, we accessed this less-crowded segment with confidence. The forecast-filled universe generated higher IC (3.69% for growth predictions) than the analyst-covered universe alone.
-
Dynamic adaptation: Industries don't stay in one lifecycle stage forever. Fast-growth sectors mature; cyclical sectors enter structural growth phases. The strategy re-classifies quarterly, avoiding the stale classifications that plague static sector models. Stocks migrating into the growth category after being mis-classified as value often delivered the strongest returns.
Segment-level granularity mattered: Using company-level classifications (based on plurality segment) produced weaker results than using SAM's segment-level revenue data to build composite scores. A diversified industrial with 60% mature products and 40% growth products shouldn't be forced into one bucket—weighting both classifications by actual revenue better reflects economic reality.
Try It Yourself
This strategy requires integrating segment-level revenue data (SAM), multi-year financial metrics (Financial Notes), and a forecast synthesis model for under-covered stocks. The infrastructure investment is non-trivial, but the payoff is access to less-efficient market segments with lifecycle-appropriate valuation frameworks.
Practical implementation considerations:
Stock universe expansion: Most institutional quant strategies focus on liquid large-caps with full analyst coverage. Extending into CSI 500 and CSI 1000 with synthetic forecasts unlocks capacity in less-crowded names—but requires careful liquidity management.
Quarterly rebalancing: Lifecycle classifications and synthetic forecasts update with earnings releases. Monthly rebalancing adds turnover without significantly improving performance; quarterly strikes the right balance.
Sector rotation overlay: Combine with macro or sentiment indicators to tilt toward growth vs. value vs. cyclical exposure dynamically. The lifecycle framework provides relative attractiveness within each class; macro views determine which class to overweight.
Risk model integration: Ensure your risk model accounts for segment-level diversification. A conglomerate classified 50/50 growth/value shouldn't be treated as pure growth for exposure calculations.
Ready to implement lifecycle-based valuation screening? Book a call to discuss data integration, forecast synthesis models, and backtest infrastructure for multi-class strategies.