Founder Knows Best: High-Promotion Founder-Led Earnings Calls
We identify 18,659 earnings calls, or 11.3% of a 165,182-call corpus spanning 1990 to 2026, where the model answered YES to Founder-Led Companies and the promotion score reached 6/9 or higher. These calls look distinctive on every behavioral dimension: promotion averages 6.63 versus 5.05 overall, confidence 7.73 versus 7.21, but candor 6.42 versus 6.86. Guidance was raised on 27.1% of these calls versus 21.1% in the base, and lowered on just 7.1% versus 11.6%. Post-call returns, however, show no edge: the median 2,002-call return is -10.6% versus -7.2% in the base sample, while beat rates are nearly identical at 39.6% versus 39.5%. Scale-dependent narrative phrases are the most over-indexed language, and the pattern peaked at 15.32% of calls in 2021 before easing to 9.85% in 2025.
- Founder-led, high-promotion calls score 6.63 on promotion versus 5.05 for the corpus, a +1.58 gap, and 7.73 on confidence versus 7.21.
- The same calls score lower on candor (6.42 versus 6.86) and on specificity (7.36 versus 7.56), pairing heightened promotion with reduced measured openness.
- Guidance was raised on 27.1% of these calls versus 21.1% overall and lowered on 7.1% versus 11.6%, but the median subsequent return of -10.6% trails the base median of -7.2%.
- Beat rates are statistically indistinguishable (39.6% versus 39.5%), and the most over-indexed phrase, Scale-Dependent Advantage Claims, appears 2.39 times more often than in the base corpus.
1Introduction
Earnings-call watchers have long suspected that founder-led companies speak differently from professionally managed ones: founders own the story, so they may promote it harder. If that shows up in measurable language, it matters for anyone screening calls by tone, because a confident founder voice could be mistaken for a stronger signal than it is. This study isolates that voice. Across 165,182 calls from 1990 through 2026, we flag the 18,659 calls where the model answered YES to Founder-Led Companies and the promotion score reached 6/9 or higher, then compare their behavioral profiles, guidance actions, characteristic phrases, yearly prevalence, and subsequent returns against the full 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 "Founder-Led Companies" AND promotion scored 6/9 or higher (n = 18,659; 11.3% of the reference set, 95% Wilson interval 11.1%–11.4%). 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 behavioral gap is one-sided: promotion runs 6.63 versus 5.05 and confidence 7.73 versus 7.21, while candor (6.42 versus 6.86) and specificity (7.36 versus 7.56) sit below baseline, and stress is essentially unchanged at 2.42 versus 2.43. Guidance skews optimistic, with raises on 27.1% of calls versus 21.1% and withdrawals at just 1.6% versus 2.7%. Language lifts concentrate in scale narratives: Scale-Dependent Advantage Claims at 2.39x and A Tiny Fraction of the Market at 2.04x, while The Finished-Story Tell under-indexes at 0.72x. The pattern peaked at 15.32% of calls in 2021, fading to 9.85% in 2025. Returns offer no compensation: the median of -10.6% trails the base's -7.2%, and the 39.6% beat rate matches 39.5%.
| Meter | Study group | Baseline | Δ |
|---|---|---|---|
| Candor | 6.42 | 6.86 | -0.44 |
| Evasion | 2.92 | 2.70 | +0.22 |
| Specificity | 7.36 | 7.56 | -0.19 |
| Stress | 2.42 | 2.43 | -0.01 |
| Promotion | 6.63 | 5.05 | +1.58 |
| Confidence | 7.73 | 7.21 | +0.52 |
| Action | Study group | Baseline |
|---|---|---|
| Raised | 27.1% | 21.1% |
| Maintained | 41.3% | 48.8% |
| Lowered | 7.1% | 11.6% |
| Withdrawn | 1.6% | 2.7% |
| Signal | Lift | In group | Baseline |
|---|---|---|---|
| Scale-Dependent Advantage Claims | 2.39× | 26.4% | 11.1% |
| A Tiny Fraction of the Market | 2.04× | 61.2% | 30.0% |
| Deferred Revenue Growing | 1.82× | 16.1% | 8.9% |
| Early Products Growing Fast | 1.66× | 64.0% | 38.5% |
| Volume About to Step Up | 1.60× | 45.6% | 28.5% |
| When the CFO Dominates | 0.51× | 7.2% | 14.1% |
| The Finished-Story Tell | 0.72× | 3.2% | 4.4% |
| Statistic | Study group | Returns sample |
|---|---|---|
| Median excess return | -10.6% | -7.2% |
| Interquartile range | -36.0% to +17.0% | — |
| Share beating SPY | 39.6% (95% CI 37%–42%) | 39.5% |
| Observations | 2,002 | 22,449 |
| Ticker | Quarter | Call date | Call grade |
|---|---|---|---|
| ASPS | Q2 2025 | 2025-07-25 | D |
| LABFF | Q2 2025 | 2025-07-24 | B+ |
| MC | Q2 2025 | 2025-07-24 | B |
| LADR | Q2 2025 | 2025-07-24 | B |
| ESRT | Q2 2025 | 2025-07-24 | C+ |
| BDN | Q2 2025 | 2025-07-24 | D |
| LOB | Q2 2025 | 2025-07-24 | B+ |
| MBLY | Q2 2025 | 2025-07-24 | B |
4Discussion
A careful reader should conclude that founder-led, high-promotion calls are measurably more promotional and more confident, and that this tone co-occurs with more guidance raises and scale-flavored language. They should not conclude that the tone causes outcomes, that these calls are better or worse investments, or that the near-identical beat rates prove the tone is meaningless; the returns sample covers only 2,002 calls with measured outcomes, and composition effects could explain the lower median. The honest summary is descriptive: this is a distinctive communication style whose measured presence carries no directional information in this data.
5Limitations
All fields are AI-read and noisy; a YES on founder-led status and a 6/9 promotion score are model judgments, not ground truth. The returns sample of 22,449 calls is skewed toward liquid names, so medians and beat rates may not generalize. Our own forward tests falsified directional prediction: nothing here should be read as a trading edge. Finally, LLMs partially remember famous stocks' histories, contaminating any backtest that reuses historical calls, and phrase-level lifts can reflect sector composition rather than company behavior. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.