Only 36% of Accelerating-Demand Calls Raised Guidance — and Just 3% Lowered It
We studied 58,672 earnings calls from 1990-2026 where demand was read as accelerating, drawn from a corpus of 165,182 calls (35.5% of all calls). These calls show a distinct behavioral profile: confidence 7.76 vs 7.21 on the baseline, promotion 5.55 vs 5.05, and stress 1.97 vs 2.43. Yet the guidance picture is more mixed than the tone suggests: 37.0% raised guidance versus 21.1% in the base, while 3.1% lowered it against 11.6% at baseline. The most over-indexed language pattern was 'Volume About to Step Up' at 1.6x, and 'The Hidden Segment' appeared at only 0.67x. Among 8,095 calls with forward returns, the median was -4.6% versus -7.2% at baseline, and 43.7% beat versus 39.5% — a modest gap, not a signal.
- 35.5% of all 165,182 calls (58,672 calls) were read as showing accelerating demand.
- On these calls, confidence scored 7.76 vs 7.21 at baseline and promotion 5.55 vs 5.05, while stress fell to 1.97 from 2.43.
- 37.0% of accelerating-demand calls raised guidance versus 21.1% of baseline calls, and only 3.1% lowered guidance versus 11.6% at baseline.
- Among 8,095 calls with forward returns, the median return was -4.6% versus -7.2% at baseline, with 43.7% beating versus 39.5% in the base sample.
1Introduction
Earnings calls where management describes demand as accelerating are among the most scrutinized moments in the calendar — analysts listen for whether the tone matches the numbers. These calls carry a distinctive behavioral signature: more confidence, more promotion, less stress. But tone and outcomes are different things, and the gap between how a call sounds and what follows is where careful readers earn their keep. This study examines 58,672 accelerating-demand calls from a 165,182-call corpus spanning 1990-2026, comparing their language profile, guidance actions, recurring phrases, and forward returns against the full baseline.
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 demand was read as accelerating (n = 58,672; 35.5% of the reference set, 95% Wilson interval 35.3%–35.8%). 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 profile is consistent: specificity runs 7.69 vs 7.56 and candor 6.79 vs 6.86, while confidence (7.76 vs 7.21) and promotion (5.55 vs 5.05) show the largest deltas. Guidance skews positive — 37.0% raised versus 21.1% at base, and 3.1% lowered versus 11.6%. The phrase 'Volume About to Step Up' appears 1.6x the base rate (45.5% vs 28.5%), followed by 'Deferred Revenue Growing' at 1.57x. 'The Hidden Segment' is under-represented at 0.67x. In the returns sample (n=8,095), the median forward return was -4.6% vs -7.2% at baseline, and 43.7% beat versus 39.5% — a real but modest difference.
| Meter | Study group | Baseline | Δ |
|---|---|---|---|
| Candor | 6.79 | 6.86 | -0.07 |
| Evasion | 2.58 | 2.70 | -0.12 |
| Specificity | 7.69 | 7.56 | +0.13 |
| Stress | 1.97 | 2.43 | -0.46 |
| Promotion | 5.55 | 5.05 | +0.50 |
| Confidence | 7.76 | 7.21 | +0.55 |
| Action | Study group | Baseline |
|---|---|---|
| Raised | 37.0% | 21.1% |
| Maintained | 45.6% | 48.8% |
| Lowered | 3.1% | 11.6% |
| Withdrawn | 1.3% | 2.7% |
| Signal | Lift | In group | Baseline |
|---|---|---|---|
| Volume About to Step Up | 1.60× | 45.5% | 28.5% |
| Deferred Revenue Growing | 1.57× | 13.9% | 8.9% |
| Early Products Growing Fast | 1.47× | 56.5% | 38.5% |
| A Tiny Fraction of the Market | 1.34× | 40.4% | 30.0% |
| Pricing Recovering | 1.33× | 28.6% | 21.5% |
| The Hidden Segment | 0.67× | 14.1% | 21.1% |
| The Finished-Story Tell | 0.70× | 3.1% | 4.4% |
| Results Worse Than Direction | 0.71× | 36.1% | 51.1% |
| Statistic | Study group | Returns sample |
|---|---|---|
| Median excess return | -4.6% | -7.2% |
| Interquartile range | -24.2% to +15.3% | — |
| Share beating SPY | 43.7% (95% CI 43%–45%) | 39.5% |
| Observations | 8,095 | 22,449 |
| Ticker | Quarter | Call date | Call grade |
|---|---|---|---|
| SBFG | Q2 2025 | 2025-07-25 | A |
| USCB | Q2 2025 | 2025-07-25 | B+ |
| MOG.A | Q3 2025 | 2025-07-25 | B+ |
| AMSF | Q2 2025 | 2025-07-25 | C+ |
| LARK | Q2 2025 | 2025-07-25 | B |
| FLG | Q2 2025 | 2025-07-25 | B |
| FRST | Q2 2025 | 2025-07-25 | A |
| HMDPF | Q2 2025 | 2025-07-25 | B |
4Discussion
A careful reader should conclude that accelerating-demand calls are, on average, accompanied by more confident language, more guidance raises, and somewhat better forward returns than the baseline. They should not conclude that the tone causes the outcomes, that any single call is informative, or that the 43.7% vs 39.5% beat rate offers an exploitable pattern. The differences are descriptive, measured over decades, and small in magnitude. The median forward return remains negative (-4.6%) even in this favorable-sounding subset, which is worth sitting with before drawing any conclusion.
5Limitations
The AI-read fields — candor, confidence, stress, and the phrase categories — are noisy model outputs, not ground truth. The returns sample covers 8,095 calls drawn from a 22,449-call base skewed toward liquid names, so results may not generalize. Our own forward tests falsified directional prediction: nothing here should be read as a trading edge. Additionally, LLMs partially remember famous stocks' histories, which can contaminate any backtest of language-based signals. All figures are descriptive associations within this dataset, not evidence of cause, effect, or future performance. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.