Every Call Hides Something: A Study of 'The Hidden Segment' Answers
This study profiles 34,917 earnings calls out of 165,182 in the Artul.ai corpus (21.14%) where the model answered YES to the battery item 'The Hidden Segment', covering calls from 1990 to 2026. Compared with the base population, these calls show lower confidence (6.93 vs 7.21), higher stress (2.86 vs 2.43), and higher evasion (2.86 vs 2.70). Guidance is lowered on 17.02% of these calls versus 11.56% in the base. Among 4,248 calls with next-quarter returns, the median return is -9.35% versus -7.16% for the base, and 38.16% beat versus 39.47% in the base.
- Hidden-segment calls make up 21.14% of the corpus (34,917 of 165,182 calls).
- These calls show lower confidence (6.93 vs 7.21) and higher stress (2.86 vs 2.43) than the base population.
- Guidance was lowered on 17.02% of hidden-segment calls versus 11.56% of base calls.
- Median next-quarter return on hidden-segment calls was -9.35% versus -7.16% for the base sample.
1Introduction
Earnings calls often pivot around a business the company barely describes: a small segment, an early product line, a market the executives call a tiny fraction of the total. Listeners know these asides can carry outsized weight, and analysts routinely press on them. Whether a model's YES answer to 'The Hidden Segment' actually flags a distinctive communication pattern is an open empirical question. This study profiles 34,917 such calls drawn from 165,182 earnings calls spanning 1990 to 2026, comparing their language profiles, guidance behavior, topical lifts, and subsequent return distributions against the base 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 "The Hidden Segment" (n = 34,917; 21.1% of the reference set, 95% Wilson interval 20.9%–21.3%). 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
Hidden-segment calls skew toward more evasive (2.86 vs 2.70) and more stressed (2.86 vs 2.43) language, with confidence running 0.29 points lower (6.93 vs 7.21) and specificity 0.09 lower. The strongest topical lifts are 'Results Worse Than Direction' (1.56x) and 'Early Products Growing Fast' (1.55x), followed by 'Scale-Dependent Advantage Claims' (1.48x). Guidance lowering is more common here (17.02% vs 11.56% base), while raising is less common (15.42% vs 21.05%). The trend is not stable: the share fell from 24.83% in 2015-2016 to 15.83% in 2021, then recovered to 22.59% by 2023 before easing to 19.66% in 2025.
| Meter | Study group | Baseline | Δ |
|---|---|---|---|
| Candor | 6.84 | 6.86 | -0.02 |
| Evasion | 2.86 | 2.70 | +0.16 |
| Specificity | 7.46 | 7.56 | -0.09 |
| Stress | 2.86 | 2.43 | +0.43 |
| Promotion | 5.10 | 5.05 | +0.05 |
| Confidence | 6.93 | 7.21 | -0.29 |
| Action | Study group | Baseline |
|---|---|---|
| Raised | 15.4% | 21.1% |
| Maintained | 49.2% | 48.8% |
| Lowered | 17.0% | 11.6% |
| Withdrawn | 3.0% | 2.7% |
| Signal | Lift | In group | Baseline |
|---|---|---|---|
| Results Worse Than Direction | 1.56× | 79.6% | 51.1% |
| Early Products Growing Fast | 1.55× | 59.6% | 38.5% |
| Scale-Dependent Advantage Claims | 1.48× | 16.4% | 11.1% |
| A Tiny Fraction of the Market | 1.33× | 39.9% | 30.0% |
| Underused Fixed Costs | 1.32× | 54.9% | 41.6% |
| Statistic | Study group | Returns sample |
|---|---|---|
| Median excess return | -9.3% | -7.2% |
| Interquartile range | -29.6% to +11.3% | — |
| Share beating SPY | 38.2% (95% CI 37%–40%) | 39.5% |
| Observations | 4,248 | 22,449 |
| Ticker | Quarter | Call date | Call grade |
|---|---|---|---|
| VRTS | Q2 2025 | 2025-07-25 | C+ |
| FLG | Q2 2025 | 2025-07-25 | B |
| FRST | Q2 2025 | 2025-07-25 | A |
| CHTR | Q2 2025 | 2025-07-25 | C+ |
| DBOEY | Q2 2025 | 2025-07-25 | B+ |
| KNSL | Q2 2025 | 2025-07-25 | C+ |
| BAH | Q1 2026 | 2025-07-25 | C+ |
| LBTSF | Q2 2025 | 2025-07-25 | C+ |
4Discussion
A careful reader should treat these as descriptive correlations, not causes. Hidden-segment calls coincide with more evasion, more stress, weaker guidance posture, and a somewhat worse median next-quarter return (-9.35% vs -7.16%), but the overlap in return distributions is large: the first quartile is -29.62% and the third quartile is +11.35%, so most outcomes span a wide range. The 38.16% beat rate (CI 36.71%-39.63%) sits close to the base 39.47%. Nothing here establishes that the hidden-segment signal causes any of these outcomes, and none of it supports a trading rule.
5Limitations
The fields analyzed are produced by an AI reading transcripts and are noisy; profile deltas of a few tenths of a point may reflect labeling error as much as real differences. The returns sample covers 22,449 calls with outcomes, skewed toward liquid names, so results may not generalize to smaller issuers. Our own forward tests falsified directional prediction from these signals. Additionally, LLMs partially remember famous stocks' histories, contaminating any backtest-style comparison of labeled calls with subsequent returns. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.