Research › Signal Combinations
Artul.ai Research LibraryStudy No. 76Signal CombinationsUpdated 2026-08-28

Trust the Raise: When Skeptics Get Reassured and Guidance Goes Up

By Artul.ai Research Group · n = 31,112 earnings calls · First published 2026-08-28
Abstract

This study isolates earnings calls where Artul.ai's model answered YES to "Skeptic Reassured" and guidance was read as raised: 31,112 calls from a 165,182-call corpus spanning 1990-2026, or 18.8% of calls (95% CI 18.6%-19.0%). These calls differ sharply in tone: stress sits at 1.72 versus a 2.43 baseline (a -0.71 gap), evasion runs 2.39 versus 2.70, and confidence runs 7.85 versus 7.21. By construction, 100% raised guidance versus 21.1% in the base. On outcomes, 42.1% of the 6,378 return-sample calls beat, versus 39.5% baseline, with a median return of -5.3% versus -7.2%. "Deferred Revenue Growing" shows the strongest over-representation at 1.59x.

Key findings
  • The pattern covers 31,112 calls, 18.8% of the 165,182-call corpus, with a 95% confidence interval of 18.6% to 19.0%.
  • Tone deltas are pronounced: stress is 1.72 versus 2.43 baseline, evasion 2.39 versus 2.70, and confidence 7.85 versus 7.21.
  • "Deferred Revenue Growing" appears 1.59x more often than baseline, while "Scale-Dependent Advantage Claims" appears at just 0.08x.
  • In the 6,378-call return sample, 42.1% beat versus a 39.5% base rate, and the median return is -5.26% versus -7.16% baseline.

1Introduction

Analysts often treat a raised guidance number as the headline and move on. But the tone surrounding that raise—whether the skeptics in the room came away satisfied—may tell a different, subtler story about how management presented the news. Calls where hard questions end with the questioner reassured, and where guidance is simultaneously raised, form a distinct population in the earnings-call record. Using Artul.ai's model-read fields across 165,182 calls from 1990 to 2026, this study profiles those 31,112 calls: their language patterns, the qualitative flags that cluster with them, how the population shifts year to year, and how their post-call returns compare with the broader 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 "Skeptic Reassured" AND guidance was read as raised (n = 31,112; 18.8% of the reference set, 95% Wilson interval 18.6%–19.0%). 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 tone profile reads as unusually calm and specific: stress is 1.72 against a 2.43 baseline, evasion 2.39 versus 2.70, while specificity (7.89 vs 7.56), candor (6.98 vs 6.86), and confidence (7.85 vs 7.21) all run higher. Flag lift is asymmetric. "Deferred Revenue Growing" runs at 1.59x baseline and "Guidance Worth Underwriting" at 1.35x, while defensive flags collapse: "Scale-Dependent Advantage Claims" at 0.08x and "Results Worse Than Direction" at 0.41x. The share of calls peaked at 28.8% in 2021 before easing to 15.19% in 2025. In the return sample, 42.1% beat versus 39.5% baseline, with a median of -5.26% versus -7.16%.

Table 1. Mean behavioral scores (0–9 scale), study group versus baseline
MeterStudy groupBaselineΔ
Candor6.986.86+0.12
Evasion2.392.70-0.31
Specificity7.897.56+0.34
Stress1.722.43-0.71
Promotion5.285.05+0.23
Confidence7.857.21+0.63
Table 2. Guidance actions, study group versus baseline
ActionStudy groupBaseline
Raised100.0%21.1%
Maintained0.0%48.8%
Lowered0.0%11.6%
Withdrawn0.0%2.7%
Table 3. Co-occurring battery signals ranked by lift (group prevalence ÷ baseline prevalence)
SignalLiftIn groupBaseline
Deferred Revenue Growing1.59×14.1%8.9%
Guidance Worth Underwriting1.35×96.6%71.5%
Early Products Growing Fast1.26×48.6%38.5%
Pricing Recovering1.25×27.0%21.5%
Scale-Dependent Advantage Claims0.08×0.8%11.1%
Results Worse Than Direction0.41×20.7%51.1%
The Question Left Hanging0.46×21.9%48.0%
The Hidden Segment0.66×13.9%21.1%
Underused Fixed Costs0.66×27.6%41.6%
20156.83%
201616.03%
201719.54%
201819.79%
201915.23%
202015.12%
202128.80%
202219.78%
202318.50%
202419.46%
202515.19%
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-5.3%-7.2%
Interquartile range-23.8% to +14.1%
Share beating SPY42.1% (95% CI 41%–43%)39.5%
Observations6,37822,449
Table 5. Most recent calls matching the study definition
TickerQuarterCall dateCall grade
SBFGQ2 20252025-07-25A
HCAQ2 20252025-07-25C
NWGQ2 20252025-07-25B+
BFHQ2 20252025-07-25B
FFICQ2 20252025-07-25B+
OMFQ2 20252025-07-25A
GBCIQ2 20252025-07-25A
MOG.AQ3 20252025-07-25B+

4Discussion

A careful reader should conclude only that this combination of model-read tone and raised guidance coincides with calmer language, fewer defensive flags, and a modestly better beat rate—descriptive differences, not causes. The beat-rate gap of 2.6 points is small and the median return remains negative, so nothing here suggests an actionable edge. The flag lifts describe which framings co-occur with reassured skeptics, not which framings produce good outcomes. Year-to-year swings likely reflect market regime as much as behavior.

5Limitations

The underlying fields—Skeptic Reassured, guidance direction, and the qualitative flags—are AI-read labels and carry real annotation noise. The returns sample covers 22,449 calls and is skewed toward liquid names, limiting generalizability. Artul.ai's own forward tests have falsified directional prediction on this data, so no trading implication should be drawn. Finally, LLMs partially remember famous stocks' histories, which contaminates any backtest-style comparison of returns across flagged groups. 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). “Trust the Raise: When Skeptics Get Reassured and Guidance Goes Up.” Artul.ai Earnings-Call Research Library, Study No. 76. https://artul.ai/research/raised-guidance-that-even-skeptics-buy

Related studies

Small Pond, Big Fish: Calls Claiming Fast Growth in The Founder Is Present: Candor and Confidence on EarLoud Backlogs and Busy Phones: Calls Where the ModelFounder Knows Best: High-Promotion Founder-Led EarniThe Quiet Hand on the Microphone: CFO Dominance MeetDodging the Question, Formally: The Anatomy of an Un
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.