Going Back for Seconds: The Pull-Forward Earnings Call
We ask how often earnings calls feature the claim that existing customers are pulling more from the company than originally said, and what such calls look like. Of 438 calls in a 2015-2024 corpus, 126 (28.8%) answered yes. These calls sound distinctly confident: confidence scores 7.60 versus 7.34, and stress 2.07 versus 2.37. Their language over-indexes on growth framing - 'Early Products Growing Fast' appears 1.51x more often (54.0% versus 35.8%) - while 'Volume About to Step Up' is under-represented at 0.74x. Guidance behavior is modestly different (24.6% raised versus 22.4%), but the returns sample is sobering: the 48 eligible calls show a median return of -18.2% versus -7.3% for the 173-call base, and a 33.3% beat rate versus 38.7%. The pattern is descriptive, not predictive.
- Of 438 calls spanning 2015-2024, 126 - a 28.8% share - answered yes to the hypothesis that existing customers are pulling more from the company than originally said.
- These calls skew sunny rather than evasive: confidence 7.60 versus 7.34 and stress 2.07 versus 2.37, with candor, evasion, and specificity essentially flat.
- 'Early Products Growing Fast' is the top over-indexed phrase at 1.51x (54.0% versus 35.8%), while 'Volume About to Step Up' is under-represented at 0.74x.
- In the returns sample of 48 calls, the median return was -18.2% versus -7.3% for the base, with beat rates of 33.3% versus 38.7%.
1Introduction
Few sentences on an earnings call are as flattering as 'our existing customers are pulling more from us than they originally said.' It flatters the product, suggests momentum, and tends to arrive wrapped in confident language. For anyone who parses calls for signal, the question is what such claims travel with: do they coincide with bolder guidance, different vocabulary, calmer delivery - and how do the stocks behave afterward? The share of such calls rose to 17% in 2022 and 16% in 2023 before fading to 5% in 2024, making pull-forward talk a lens on demand timing. This study examines the 126 calls that answered yes, profiling their tone, guidance behavior, phrase lifts, and returns against the base.
2Data & methodology
The corpus comprises 438 earnings-call transcripts published between 2015 and 2024, 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 that answered YES to the research hypothesis "Existing customers are pulling more from the company than they were originally s" (n = 126; 28.8% of the reference set, 95% Wilson interval 24.7%–33.2%). Baseline figures use the set of calls on which this question was tested. 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 the headline: confidence runs 7.60 versus 7.34 and stress 2.07 versus 2.37, while candor (6.84 versus 6.92), evasion (2.74 versus 2.69), and specificity (7.61 versus 7.64) barely move - these calls are sunnier, not dodgier. Phrase lifts reinforce the framing: 'Early Products Growing Fast' at 1.51x (54.0% versus 35.8%), 'A Tiny Fraction of the Market' at 1.48x (43.7% versus 29.5%), and 'Founder-Led Companies' at 1.40x (29.4% versus 21.0%), with 'Volume About to Step Up' at just 0.74x. Guidance is little changed - 24.6% raised versus 22.4% - though 1.6% withdrew versus 0.9%. The annual share peaked at 17% in 2022 and 16% in 2023, then faded to 5% in 2024. In the returns sample, the median was -18.2% versus -7.3%, with 33.3% beating versus 38.7% for the base.
| Meter | Study group | Baseline | Δ |
|---|---|---|---|
| Candor | 6.84 | 6.92 | -0.08 |
| Evasion | 2.74 | 2.69 | +0.05 |
| Specificity | 7.61 | 7.64 | -0.03 |
| Stress | 2.07 | 2.37 | -0.30 |
| Promotion | 5.44 | 5.09 | +0.35 |
| Confidence | 7.60 | 7.34 | +0.27 |
| Action | Study group | Baseline |
|---|---|---|
| Raised | 24.6% | 22.4% |
| Maintained | 53.2% | 51.6% |
| Lowered | 11.1% | 13.7% |
| Withdrawn | 1.6% | 0.9% |
| Signal | Lift | In group | Baseline |
|---|---|---|---|
| Early Products Growing Fast | 1.51× | 54.0% | 35.8% |
| A Tiny Fraction of the Market | 1.48× | 43.7% | 29.5% |
| Founder-Led Companies | 1.40× | 29.4% | 21.0% |
| Volume About to Step Up | 0.74× | 17.5% | 23.5% |
| Statistic | Study group | Returns sample |
|---|---|---|
| Median excess return | -18.2% | -7.3% |
| Interquartile range | -36.0% to +10.2% | — |
| Share beating SPY | 33.3% (95% CI 22%–47%) | 38.7% |
| Observations | 48 | 173 |
| Ticker | Quarter | Call date | Call grade |
|---|---|---|---|
| SFIX | Q3 2024 | 2024-06-04 | C+ |
| CRGO | Q1 2024 | 2024-05-20 | C+ |
| WRBY | Q1 2024 | 2024-05-09 | A |
| HCKT | Q1 2024 | 2024-05-08 | C |
| PPC | Q1 2024 | 2024-05-03 | A |
| PDS | Q1 2024 | 2024-04-25 | B |
| SNV | Q1 2024 | 2024-04-18 | B |
| DUOT | Q4 2023 | 2024-04-01 | F |
4Discussion
A careful reader should conclude that pull-forward calls are measurably more confident and more saturated with growth vocabulary, and that their returns in this sample ran worse than the base. They should not conclude that the language causes anything, that these calls predict losses, or that avoiding them is an edge - our own forward tests falsified directional prediction. The median gap (-18.2% versus -7.3%) is a descriptive comparison of 48 against 173 calls, and the lift table shows co-occurrence, not mechanism. Read these numbers as a portrait of how such calls sound, not a forecast.
5Limitations
The fields are AI-read and noisy, so phrase flags and tone scores carry labeling error. The returns sample is 22,449 calls skewed to liquid names, so the 48-call subset inherits selection effects. Our own forward tests falsified directional prediction, so nothing here should be traded. Because LLMs partially remember famous stocks' history, contaminating any backtest, the return comparisons are descriptive at best. Year shares also rest on small counts, and the 2025 window is partial. See the full methodology, including the C1 pattern’s forward-test failure and the LLM-memorization finding.
Companion page: every company matching this hypothesis is listed at the question’s own page.