The Founder Is in the Room: Language Profiles of Founder-Led Earnings Calls
This study examines 33,092 earnings calls (out of 165,182 from 1990-2026) where the model answered YES to the battery item "Founder-Led Companies." Founder-led calls score higher on promotion (+0.61) and confidence (+0.14) than the base, while scoring lower on candor (-0.22) and specificity (-0.15). They are more likely to claim a tiny fraction of the market (1.6x lift) or a scale-dependent advantage (1.76x lift), and less likely to show the finished-story tell (0.73x). In the returns sample of 3,582 calls, the beat rate is 41.1% versus 39.5% for the base 22,449 calls, with a median next-day return of -0.08 versus -0.07.
- Founder-led calls show higher promotion (+0.61) and confidence (+0.14) but lower candor (-0.22) than the base corpus.
- Scale-dependent advantage claims appear 1.76x more often on founder-led calls; tiny-fraction-of-market claims 1.6x.
- Founder-led calls run the finished-story tell less often than the base (0.73x).
- In the returns sample, founder-led calls beat expectations 41.1% of the time versus 39.5% for the base, with a median next-day return of -0.08.
- The share of founder-led calls fell from 22.69% in 2021 to 18.08% in 2025.
1Introduction
Founders running their own companies are a recurring fascination for earnings-call watchers: the presumption is that founders talk differently, promise more, and level more. With a machine-labeled battery across 165,182 calls spanning 1990-2026, we can check which of those presumptions show up in the language itself. This study isolates the 33,092 calls where the model answered YES to "Founder-Led Companies" and compares their linguistic profile, guidance behavior, characteristic phrases, and returns against the rest of the corpus.
2Data & methodology
The corpus comprises 165,182 earnings-call transcripts published between 1990 and 2026, each scored independently by a large language model on an identical 37-field battery: seven categorical business verdicts, eight 0–9 behavioral meters, and twenty yes/no judgments. The study group is defined as calls where the model answered YES to the battery item "Founder-Led Companies" (n = 33,092; 20.0% of the reference set, 95% Wilson interval 19.8%–20.2%). Baseline figures use all scored calls. Market outcomes join a fixed sample of 22,449 calls with twelve-month total returns in excess of SPY, measured from the first close after each call; this sample skews toward liquid U.S. names and is reported as descriptive history only.
3Results
The language profile is the headline: founder-led calls score +0.61 higher on promotion and +0.14 higher on confidence, but -0.22 lower on candor and -0.15 lower on specificity than the base corpus. Their phrase signature leans toward growth-story framing - scale-dependent advantage claims at 1.76x the base rate and tiny-fraction-of-market claims at 1.6x. The share of founder-led calls peaked at 22.82% in 2022 and slid to 18.08% by 2025. On outcomes, the beat rate is 41.1% versus 39.5% for the base, with a median next-day return of -0.08 versus -0.07 - differences that are modest.
| Meter | Study group | Baseline | Δ |
|---|---|---|---|
| Candor | 6.64 | 6.86 | -0.22 |
| Evasion | 2.79 | 2.70 | +0.09 |
| Specificity | 7.41 | 7.56 | -0.15 |
| Stress | 2.49 | 2.43 | +0.06 |
| Promotion | 5.66 | 5.05 | +0.61 |
| Confidence | 7.36 | 7.21 | +0.14 |
| Action | Study group | Baseline |
|---|---|---|
| Raised | 22.7% | 21.1% |
| Maintained | 42.6% | 48.8% |
| Lowered | 10.6% | 11.6% |
| Withdrawn | 2.4% | 2.7% |
| Signal | Lift | In group | Baseline |
|---|---|---|---|
| Scale-Dependent Advantage Claims | 1.76× | 19.4% | 11.1% |
| A Tiny Fraction of the Market | 1.60× | 48.0% | 30.0% |
| Deferred Revenue Growing | 1.54× | 13.7% | 8.9% |
| Early Products Growing Fast | 1.38× | 53.0% | 38.5% |
| Volume About to Step Up | 1.26× | 35.9% | 28.5% |
| The Finished-Story Tell | 0.73× | 3.2% | 4.4% |
| Statistic | Study group | Returns sample |
|---|---|---|
| Median excess return | -8.5% | -7.2% |
| Interquartile range | -32.6% to +17.4% | — |
| Share beating SPY | 41.1% (95% CI 39%–43%) | 39.5% |
| Observations | 3,582 | 22,449 |
| Ticker | Quarter | Call date | Call grade |
|---|---|---|---|
| KNSL | Q2 2025 | 2025-07-25 | C+ |
| LBTSF | Q2 2025 | 2025-07-25 | C+ |
| ASPS | Q2 2025 | 2025-07-25 | D |
| WZZAF | Q1 2026 | 2025-07-25 | F |
| FINW | Q2 2025 | 2025-07-25 | B |
| MLLGF | Q2 2025 | 2025-07-25 | C+ |
| LABFF | Q2 2025 | 2025-07-24 | B+ |
| SAM | Q2 2025 | 2025-07-24 | D |
4Discussion
A careful reader should conclude that founder-led calls in this corpus carry a distinct verbal signature: more promotional and confident, slightly less candid and specific, and richer in scale-advantage and market-fraction framing. They also run the finished-story tell less often (0.73x). What a reader should not conclude is that founder leadership causes these language patterns, or that the small differences in beat rates or returns make founder-led calls better or worse investments. The labels come from a model, not from verified founder identity, and every comparison here is descriptive.
5Limitations
The founder-led labels are AI-read battery items and inherit the noise of any model judgment; they are not verified against corporate records. The returns sample covers 22,449 calls and is skewed toward liquid names, so the beat-rate and return comparisons may not generalize. Our own forward tests falsified directional prediction from these features. Finally, LLMs partially remember famous stocks' histories, which can contaminate any backtest that relies on model-generated labels or scores. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.