Research › Business Verdicts
Artul.ai Research LibraryStudy No. 93Business VerdictsUpdated 2026-08-28

Answering the Slowdown Before It Arrives: A Study of Slowing-Demand Calls

By Artul.ai Research Group · n = 15,797 earnings calls · First published 2026-08-28
Abstract

This study examines 15,797 earnings calls—9.56% of a 165,182-call corpus spanning 1990 to 2026—where demand was read as slowing. Management behavior shifts noticeably on these calls: stress language runs 3.53 versus a 2.43 baseline (a +1.10 gap), while confidence drops 1.09 points and promotion language falls 0.73. Guidance actions diverge sharply from the wider sample: 39.74% of these calls lowered guidance against 11.56% overall, while only 4.13% raised it versus 21.05%. Signature patterns include 'Results Worse Than Direction' (1.46x lift) and 'The Question Left Hanging' (1.41x). Post-call returns are modestly weaker: a median of -0.0920 versus -0.0716, with 38.51% of calls beating versus 39.47% in the base. The paper reports associations, not causal claims.

Key findings
  • Slowing-demand calls make up 9.56% of the 165,182-call corpus, with a 95% interval of 9.42% to 9.71%.
  • Stress language is elevated by 1.10 points (3.53 vs 2.43) while confidence is depressed by 1.09 points (6.12 vs 7.21) on these calls.
  • Guidance is lowered on 39.74% of slowing-demand calls versus 11.56% overall, and withdrawn on 7.67% versus 2.66%.
  • Post-call median returns are -0.0920 versus -0.0716 in the base sample, with 38.51% beating versus 39.47% baseline.

1Introduction

Earnings calls are where demand turns get narrated first, and 'demand is slowing' is among the most consequential things a management team can say or imply. These moments concentrate analyst attention, reshape guidance, and often reset a stock's story. Yet slowdown language is not uniform: some calls pair it with concrete guidance cuts, others with reassurance and vague outlooks. Understanding how language, guidance behavior, and tone shift when demand is read as slowing gives followers of calls a structured way to notice what typically accompanies that narrative. Using 15,797 calls from a 165,182-call corpus covering 1990 through 2026, this study profiles the tone, guidance actions, recurring discourse patterns, and post-call return distribution of calls where demand was read as slowing.

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 slowing (n = 15,797; 9.6% of the reference set, 95% Wilson interval 9.4%–9.7%). 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

Tone shifts are pronounced: stress rises 1.10 points above baseline (3.53 vs 2.43), confidence falls 1.09 (6.12 vs 7.21), promotion drops 0.73, and candor actually rises 0.25—management sounds more stressed but not more evasive. Guidance tells the sharper story: 39.74% of these calls lower guidance versus 11.56% overall, and 7.67% withdraw it versus 2.66%. Discourse patterns over-index on 'Results Worse Than Direction' (1.46x), 'The Question Left Hanging' (1.41x), and 'Underused Fixed Costs' (1.41x); 'Volume About to Step Up' appears at only 0.34x the base rate. Returns are modestly weaker—a median of -0.0920 versus -0.0716—with a wide interquartile range from -0.2948 to 0.1345.

Table 1. Mean behavioral scores (0–9 scale), study group versus baseline
MeterStudy groupBaselineΔ
Candor7.116.86+0.25
Evasion2.872.70+0.18
Specificity7.417.56-0.15
Stress3.532.43+1.10
Promotion4.325.05-0.73
Confidence6.127.21-1.09
Table 2. Guidance actions, study group versus baseline
ActionStudy groupBaseline
Raised4.1%21.1%
Maintained30.8%48.8%
Lowered39.7%11.6%
Withdrawn7.7%2.7%
Table 3. Co-occurring battery signals ranked by lift (group prevalence ÷ baseline prevalence)
SignalLiftIn groupBaseline
Results Worse Than Direction1.46×74.9%51.1%
The Question Left Hanging1.41×67.7%48.0%
Underused Fixed Costs1.41×58.7%41.6%
Scale-Dependent Advantage Claims1.40×15.5%11.1%
When the CFO Dominates1.32×18.6%14.1%
Volume About to Step Up0.34×9.7%28.5%
Deferred Revenue Growing0.45×4.0%8.9%
Pricing Recovering0.48×10.3%21.5%
Skeptic Reassured0.51×34.0%66.4%
Early Products Growing Fast0.60×23.0%38.5%
201518.43%
201610.94%
20175.80%
20185.91%
201911.25%
202011.60%
20212.83%
202212.47%
202313.80%
20249.50%
20259.13%
Figure 1. Share of all analyzed calls matching the study definition, by year.
Table 4. Twelve-month excess total returns versus SPY (descriptive history, not a signal)
StatisticStudy groupReturns sample
Median excess return-9.2%-7.2%
Interquartile range-29.5% to +13.4%
Share beating SPY38.5% (95% CI 36%–41%)39.5%
Observations2,08022,449
Table 5. Most recent calls matching the study definition
TickerQuarterCall dateCall grade
MTHQ2 20252025-07-25C
SAMQ2 20252025-07-24D
FPHQ2 20252025-07-24D
HZOQ3 20252025-07-24D
JAKKQ2 20252025-07-24D
OBKQ2 20252025-07-24A
ANIOYQ2 20252025-07-24D
CYHQ2 20252025-07-24D

4Discussion

A careful reader should conclude that slowing-demand calls carry a consistent tonal signature: more stress, less confidence, more candor, and far more guidance cuts and withdrawals. The discourse patterns suggest these calls disproportionately feature results that understate prior direction and questions left unresolved. What should not be concluded is that any of this predicts returns. The median post-call return difference is small, the spread across calls is enormous, and the beat rate (38.51% vs 39.47%) differs only slightly. These are descriptive associations measured across decades of calls, not trading signals.

5Limitations

All fields here are AI-read and inherently noisy; tone and pattern labels are probabilistic readings of speech, not ground truth. The returns sample covers 22,449 calls and skews toward liquid names, so the return comparison may not generalize. Our own forward tests falsified directional prediction, so no edge should be inferred. Additionally, LLMs partially remember famous stocks' histories, which can contaminate any backtest built on these annotations. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.

Cite this study Artul.ai Research Group (2026). “Answering the Slowdown Before It Arrives: A Study of Slowing-Demand Calls.” Artul.ai Earnings-Call Research Library, Study No. 93. https://artul.ai/research/when-demand-is-slowing-earnings-calls

Related studies

Steady As She Goes: What Happens When Guidance HoldsMargins Are Expanding, and So Is the Confidence: ReaCapex Up, Stress Down: Reading 64,183 Calls Where SpGuidance Is a Vibe: Calls Where Demand Reads as AcceBacklog Is Growing, and So Is the Confidence: EarninMargins Are Fine, Thanks for Asking: What Contractio
Not investment advice. Artul.ai publishes AI-generated earnings-call quality grades and expected-volatility estimates — never buy or sell recommendations. We tested over 1,600 predictive hypotheses against 165,000 transcripts; the honest result, including what failed, is documented in our methodology.