Usage Metrics

Three metrics answer the two questions that matter about a voice agent in production: is it connecting to people, and what does each result cost you.

Answer rate (ASR)

Answer-seizure ratio is the share of attempted calls that were actually answered.

asr = connected / attempted

• attempted — every run started in the window
• connected — those that reached an answered state
• asr — the ratio, or null when nothing was attempted

Note

asr is deliberately null rather than 0.0 on an account with no calls yet. Zero would read as “every call went unanswered”, which is a very different and much more alarming statement than “no calls have been placed.”

A low answer rate is almost always a telephony or list-quality problem — wrong numbers, bad caller ID reputation, or dialling outside the hours people pick up — not an agent problem. Check it before tuning prompts.

This is the organization-wide counterpart to the per-campaign connection rate in campaign reports; both use the same definition, so the numbers are comparable.

Cost by outcome

For every outcome your calls actually produced, this reports how many calls, what they cost in total, and the cost per call.

Field	Meaning
disposition	The outcome string, taken from the call’s mapped_call_disposition
calls	How many calls ended with that disposition
total_charged_paise	Total charged across those calls
cost_per_call_paise	Total divided by call count

Why there is no “cost per booking”

There is deliberately no universal success metric here, and this surprises people, so it is worth being explicit.

Decibyl has no fixed, cross-account taxonomy of outcomes. Every workflow defines its own disposition strings — through its End Call tool or its prompt — so one account’s success is appointment_booked, another’s is qualified_lead, another’s is payment_promised. A platform-level “cost per booking” would have to either invent a category that doesn’t match how you actually classify calls, or silently guess which of your dispositions counts as success.

Instead, every disposition your calls produced gets its own row. Read whichever row is your success outcome — that row’s cost_per_call_paise is your true cost per result.

Calls whose disposition was never set appear as unknown. A large unknown row usually means a workflow path that ends without setting a disposition, which is worth fixing — it is the one thing that makes this report hard to read.

Reading it

{
  "answer_seizure_ratio": { "attempted": 420, "connected": 231, "asr": 0.55 },
  "cost_by_outcome": [
    { "disposition": "appointment_booked", "calls": 38, "total_charged_paise": 152000, "cost_per_call_paise": 4000 },
    { "disposition": "not_interested",     "calls": 141, "total_charged_paise": 282000, "cost_per_call_paise": 2000 },
    { "disposition": "unknown",            "calls": 52, "total_charged_paise": 78000,  "cost_per_call_paise": 1500 }
  ]
}

Here the real number to act on is ₹40.00 per appointment booked — not the ₹20.00 average across all answered calls, which flatters the agent by counting cheap early hang-ups as if they were results.

Endpoint

GET /api/v1/organizations/usage/outcomes?days=7

Parameter	Default	Range	Notes
days	7	1–90	Whole days, bounded to IST day boundaries

Both metrics cover the same window, scoped to your currently selected organization. Days are counted as IST calendar days, so a range always starts and ends at local midnight rather than a rolling 24-hour offset from whenever you happened to call the API.

Circuit-breaker trips

Campaign reports include a count of how many times that campaign’s circuit breaker paused it.

The breaker pauses a campaign automatically when calls are failing at a rate that suggests something systemic — a telephony outage, an expired credential, an unreachable number range — rather than letting it burn through a contact list producing nothing. Each pause is recorded on the campaign.

A non-zero trip count on a campaign that otherwise looks finished is worth investigating. It means the campaign spent time stopped, and its results cover less of your contact list than the totals suggest.