# We Analyzed 360+ Stocks with Gemini AI to Find Out
When screening for growth stocks, investors are inundated with options: revenue per share, EBITDA, free cash flow, book value, and normalized earnings. But if your goal is beating the market over a 5- or 10-year holding period, which number actually matters most?
We ran a correlation and regression analysis on a universe of over top 360 companies by market cap exported from GuruFocus to determine which growth metrics best predict 5-year and 10-year annualized total returns.
Here is what the data revealed.
# 1. For 10-Year Horizons, Normalized EPS growth is Undisputed King
Over short periods, market sentiment and valuation multiple swings (P/E expansion or contraction) create substantial noise. Over a 5-year horizon, even the best individual growth metrics explain only about 16% of total return variance (r≈0.40).
However, stretch the horizon to **10 years**, and multiple noise washes out—underlying fundamentals take over:
* **EPS without NRI Growth Rate** emerged as the single strongest predictor of 10-year total returns (r=0.61, R2=37.4%).
* Stripping out Non-Recurring Items (NRI) proves crucial: it captures pure operational earnings growth rather than one-off asset sales or accounting windfalls. Over a decade, share price performance tightly tracks bottom-line per-share compounding.
# 2. EBITDA Per Share growth is the Most Reliable Universal Screener
While EPS takes the crown over 10 years, **EBITDA Growth Rate (Per Share)** tied for first place over 5 years (r=0.40) and finished a close second over 10 years (r=0.58).
More importantly, EBITDA solves a major data problem:
* Many financial databases (including GuruFocus) assign a default value of `0.0` when a growth rate cannot be computed due to negative initial-period numbers.
* In our dataset, Free Cash Flow and Owner Earnings were uncomputable for **30% to 55%** of companies due to cyclical dips or heavy capital expenditure cycles.
* Because EBITDA stays positive far more reliably across cycles, it suffered the fewest missing data points, making it the most robust, all-weather metric for automated stock screens.
# 3. Revenue Growth is Overrated on its Own
A popular adage suggests that "growth begins with the top line." While true, revenue expansion alone is a surprisingly weak standalone predictor of long-term returns:
* **10-Year Revenue Growth (Per Share)** registered a correlation of just **r=0.26**, explaining less than **7%** of 10-year return variation.
* “Growth at all costs” without operational leverage or profit conversion rarely translates into market-beating returns. A business growing revenue at 25% that dilutes shares or compresses margins will consistently underperform a steady compounder growing top-line at 10% with rising earnings per share.
# 4. The "Sweet Spot": Combine Bottom-Line Conversion with Top-Line Growth
While revenue growth on its own is weak, it becomes exceptionally powerful when paired with normalized earnings:
* A dual regression model combining **10-Year EPS without NRI** with **10-Year Revenue Per Share** bumped explanatory power from **37.4% up to nearly 44% (R2=0.438)**.
* This combination filters for companies expanding both sales volume and profit margins, which historically produces the highest long-term market winners (such as NVIDIA and Comfort Systems USA in this dataset).
# Key Takeaway for Your Stock Screener
If you are setting up screens on GuruFocus or your preferred research terminal:
1. **Primary Screen:** Filter for companies delivering high annualized **EBITDA per share** and **EPS without NRI** (w/o non-recurring items) growth.
2. **Watch the Data Defaults:** Beware of `0.0` values in cash flow fields—they often signify negative base years rather than flat growth.
3. **Avoid Pure Sales Growth Screens:** Top-line growth without bottom-line EPS translation is one of the easiest ways to land in a value trap or unprofitable high-flier.
[https://share.gemini.google/49faO7zBaHPI](https://share.gemini.google/49faO7zBaHPI)
