New City Mastery — Call Intelligence Dashboard

Avatars: who prospects are

Loading…

Relative size

Each avatar

Table view — avatars

Outcome

Close rate among analyzed calls, and how strong each close was.

Close strength among closed_won calls

Two different bars, deliberately kept apart: payment confirmed is a strict definition of a close (money moved on the call); payment intent is a broader one (the prospect said yes but payment wasn't confirmed on-call).

Table view — outcome distribution

Pains

Distinct calls mentioning each pain category — a call with multiple matching labels counts once.

Table view — pain categories

Objections

Distinct calls raising each objection category.

Table view — objection categories

Goals

Distinct calls expressing each stated goal category.

Table view — goal categories

Marketing vs. what prospects said

Claim by claim

Pains the marketing never names

Pain categories that show up often in calls but that no claim in the marketing copy addresses at all — real problems prospects arrive with that the funnel never mentions.

Table view — pains the marketing never names

Close mechanics

Prices below are as DESCRIBED ON THE CALL, parsed from natural language the model transcribed — not billing records. The list price is taken as the first dollar amount mentioned in that description (the price before any discounting); a description that leads with an already-discounted figure will understate it. Monthly and one-time/unknown prices are shown separately because they are not on the same scale and are not comparable.

Monthly price ranges

One-time / unclear-basis price ranges

Discounting

Median list price

Table view — price ranges

Script adherence

Per-section presence rate and mean fidelity (0–100) across analyzed calls.

Table view — script adherence

Contrast: close rate vs baseline

Read carefully. Outcomes are INFERRED from transcripts, not from payment records — the extraction model reads what was said and predicts an outcome. Outcome rule validated at 7 of 8 on the only labeled calls available. Every row below is association only, never causation — a category appearing more often among closed calls does not mean it caused, drove, or led to the close. Rows marked LOW N have too few calls or too few closes to be statistically reliable; treat them as noise, not signal.

Pain categories

Objection categories

Table view — contrast (pains + objections)

Method & limits

How evidence works. For every extracted field, the model cites a sentence index into the transcript; code then resolves that index back to the actual transcript line before it is trusted. The model cannot fabricate a supporting quote — it can only point at a real line, and that pointer is checked. Across analyzed calls: ….
Outcome rule. … Outcome is a model inference from call content (closed_won / not_closed / no_content), never a CRM or payment record.
Truncated transcripts. Some source transcripts are truncated at ingest; a truncated call can undercount pains, objections, or goals discussed after the cutoff, and can bias outcome detection if the close happened after the cutoff.
Partial corpus. This dashboard reflects … of the full call corpus. Category counts, rates, and the contrast section will all shift as more calls are extracted — nothing here should be treated as final.
Single extraction per call. Each call has exactly one extraction pass. There is no repeated-measure or inter-rater estimate of how much extraction noise exists on any individual call.