Tracked FY21 Q1 → FY24 Q1 · 13 calls in our system
Behavioral baseline · career averages (the “normal” we score every call against)
0.5%
Hedge density baseline (words flagged as hedges / total)
92
Tokens / Q&A turn baseline (median answer length)
5
Recurring boilerplate templates inventoried
3.9 / 10
Career avg Call Weirdness Index
Descriptor sentiment distribution · prepared remarks
Positive 41%
Neutral 47%
Hedged 12%
Coleman is a precise, numbers-first speaker — her tells are verbosity spikes and conditional qualifiers on guidance, not tonal language.
Per-call history · every call we've scored them on
Click any call to open it. The FY24 Q1 call is the highest-deviation call by any ATDX speaker in our coverage window.
Top novel phrases ever used
Descending by scoring significance. Click a phrase to open the call where it appeared.
“different from the way we planned” — explicit plan-deviation acknowledgement
9.6 sigFY24 Q1 →
“no net change overall” — guidance maintenance · implies offsetting moves
9.0 sigFY24 Q1 →
“currently assuming” — conditional time qualifier on assumptions
8.3 sigFY24 Q1 →
“back-half loaded year than we originally expected” — cadence reframe
6.4 sigFY24 Q1 →
“continue to monitor a number of variables” — vague forward-monitoring
5.2 sigFY24 Q1 →
“normalization of ordering patterns” — first used FY23 Q2
3.6 sigFY23 Q2 →
Recurring boilerplate templates
Sentence frames Sarah reuses, with the variable slot tracked over time. The slot is where the signal lives.
“our guidance reflects {FRAMING} from prior”
FY22 an increaseFY23 consistent withFY24 no net change overall ↓
“we {VERB} our full-year guidance”
FY22 will deliverFY23 remain on track forFY24 expect to achieve ↓
Sarah vs. the median Health Care CFO in our corpus
Hedge density (Q&A)1.1× CFO median
Tokens / Q&A turn1.8× CFO median
Novel-phrase rate / call1.4× CFO median
Conditional-qualifier frequency1.9× CFO median
Comparisons are against the median CFO in the Health Care sector across our corpus. Coleman's standout tell is verbosity — when pressed she runs nearly 2× her peers' answer length — alongside an elevated rate of conditional qualifiers on guidance. Each call is still scored against her own baseline.
Why this is per-speaker, not generic sentiment
A generic sentiment model reads “solid financial performance” as positive and moves on. We score it against Sarah's own prior calls — where the same boilerplate carried stronger language. The deviation, not the absolute tone, is what carries information. Every metric on this page is calibrated to this one speaker; a new executive builds their baseline over their first several quarters before we score them with confidence.
Compounding tells · FY21 Coleman vs. FY24 Coleman
FY21 Coleman
- • Median Q&A answer 92 tokens — terse, declarative
- • Guidance verb: “will deliver” / “are raising”
- • Assumptions stated, not qualified
- • Names specific variables by line item
FY24 Coleman
- • Pressed-question answers up to 487 tokens (5.3×)
- • Guidance verb: “expect to achieve”
- • “currently assuming” / “no net change overall”
- • Vague monitoring (“a number of variables”)
The same executive, three years apart, with a measurably different language profile. None of this is visible in a single call read in isolation — it only emerges when each call is scored against the speaker's own longitudinal baseline.
Every executive has a baseline.
Spoken Alpha builds a longitudinal language profile for each of 12,408 speakersat S&P 1500 companies — and scores every new call against the person, not a generic sentiment model.