Call Operations Diagnostic — Estimation Methodology

Version: v0.1-2026-07 · Prepared: 15 July 2026 · Owner: GoCadence (gocadence.ai) Status: DRAFT — every number this model produces is an ESTIMATE, not a measurement.

0. What this is, and the honesty rules

The Call Operations Diagnostic is a per-network page estimating the monthly economics of missed and abandoned phone calls for a named ANZ healthcare network, built only from public data (site counts, published financials, job ads, tenders, press) plus the labeled benchmark model below. It never uses, or claims to use, the target's internal call data.

Non-negotiable output rules (enforced on every generated page):

1. Every figure is labeled ESTIMATE and shown as a conservative / typical / aggressive band — never a single point number.
2. Every page links to this methodology page, showing every input, parameter, and source.
3. Revenue figures are gross revenue, never profit, and are labeled as such.
4. Segment-appropriate modeling: we do not claim booking-revenue recovery in segments where calls are predominantly service traffic (pathology, most aged-care volume) — those use a cost-deflection model instead, clearly labeled.
5. Parameters resting on vendor-published or unaudited sources are marked [VERIFY] in this document and inherit an "industry-reported, unaudited" footnote on the page.
6. Sanity caps (§C) are always applied and displayed: estimates that would exceed a plausible share of the organisation's published revenue or client intake are capped, and the cap shown.
7. If the target organisation tells us a figure is wrong, we correct or remove the page. The page carries a visible "correct our assumptions" contact link.
8. No vendor recommendations and no affiliate links appear on diagnostic pages (GoCadence is vendor-neutral by design).

A. Inputs (per network)

B. Benchmark parameters (with sources)

All bands are [conservative / typical / aggressive]. Where a parameter has no defensible published ANZ benchmark, it is modelled and flagged; the band is deliberately wide.

B1. Inbound calls per site per day, by segment

Working days: 22/month (Mon–Fri) for site-based segments; 22 / 26 / 30 for extended-hours contact centres (Healius ads specify Mon–Sun 7am–9pm).

B2. Missed / abandoned call rate

B3. Booking/enquiry-intent share of calls

B4. Permanent-loss rate of a missed booking call

Not every missed call is a lost booking — most callers retry. Band: 15% / 25–30% / 35–40%. Upper anchor is the widely repeated (vendor-published, unaudited [VERIFY]) claim that ~85% of callers whose first call goes unanswered don't call back (AgentZap) — we treat that as an extreme ceiling and band far below it.

B5. Value per recovered booking / call, by segment (A$, gross)

B6. After-hours share / uplift

Applied as a multiplier to the revenue-recovery line only (missed after-hours demand a 24/7 AI agent can capture): 1.00 / 1.10–1.15 / 1.25–1.30.

• Basis (all US vendor-published, unaudited [VERIFY] as a class): ~41% of patient calls arrive outside 8am–5pm weekdays, and only ~19% of healthcare call centres operate 24/7 (AgentZap); 30–40% after-hours share per ACC Solutions. Because these are unaudited, we cap the uplift at +25–30% even in the aggressive band and use 1.00 (no uplift) in the conservative band.
• Aged care gets the higher band ceiling (1.30) given the Aged Care Act 2024's requirement for complaints/whistleblower channels reachable by anyone — in practice, by phone, at any hour (Dept. of Health).

B7. AI-recoverable share (full handling, no human)

Band: 30% / 40% / 60%. Anchored to verified ANZ deployments:

• Uniting NSW.ACT "Jeanie" (Webex/Optus, launched 23 Oct 2025): >40% of incoming calls fully handled without human intervention, >60% of routine inquiries resolved at first interaction, 35% wait-time reduction (Cisco/Webex). Honesty note: trade press separately reports Jeanie managing ~14% of total inbound across the whole organisation while migration continues (TechPartner.news) — the >40% figure applies to migrated contact centres. Our conservative band (30%) sits below the migrated-centre figure precisely because rollouts are partial in year one.
• HealthEngine "Helen" (GP receptionist): completed 63% of patient calls end-to-end in pilot (ITBrief, Aug 2025) — this sets the aggressive ceiling (60%).
• Cross-check: Sierra reports >70% resolution at Singtel (non-healthcare) (Sierra).

C. Formula chains

C1. Revenue-recovery model (GP, dental, imaging; aged-care new-client line)

monthly_calls        = sites × calls_per_site_per_day × working_days
missed_calls         = monthly_calls × missed_rate
lost_bookings        = missed_calls × booking_intent_share × permanent_loss_rate
lost_bookings        = min(lost_bookings, intake_cap)          # sanity cap, aged care (C3)
est_recoverable_A$   = lost_bookings × value_per_booking × ai_full_handling_share × after_hours_uplift

Rationale for the final two factors: an AI agent answers previously-missed calls; we only claim the share it fully handles end-to-end (B7), and only then apply any after-hours uplift (B6). Calls the AI answers and hands to a human also recover value — we deliberately exclude that (conservatism), and say so on the page.

C2. Cost-deflection model (pathology; aged-care answered-call line)

handled_calls_mo     = agents × calls_per_agent_per_day × operating_days
offered_calls_mo     = handled_calls_mo / (1 − abandonment_rate)
abandoned_mo         = offered_calls_mo − handled_calls_mo      # reported, not monetised
est_deflection_A$    = handled_calls_mo × ai_full_handling_share × cost_per_call
est_deflection_A$    = min(est_deflection_A$, 0.60 × agents × loaded_agent_cost_pa / 12)   # cap (C3)

Labeled on-page as estimated monthly cost-to-serve deflection, not revenue. Caveat shown: deflection at the margin is less than average cost per call until seats are actually re-planned.

C3. Sanity caps (always applied, always displayed)

1. Cost-deflection cap: savings ≤ 60% of estimated fully-loaded staffing cost (loaded agent cost banded A$75k/85k/95k p.a. [VERIFY — modelled from Matchboard's A$60–80/hr fully-loaded rates]).
2. Aged-care intake cap: permanently-lost new clients/month ≤ a plausible share of the provider's monthly client intake (banded per provider from published client counts).
3. Revenue ratio display: where the target publishes revenue, recoverable-A$ is also shown as a % of monthly revenue; any band exceeding ~4% of revenue is flagged on-page as "aggressive scenario — likely overstates".

C4. Bands

Conservative, typical, and aggressive use the respective parameter from every band column — i.e. bands compound. This makes the conservative figure genuinely floor-like and the aggressive figure an explicit ceiling, and is stated on the page.

D. Known weaknesses (public register)

• Most missed-call and after-hours rates come from US vendor content marketing — no audited ANZ study of practice-level missed-call rates was found. This is the model's weakest layer; all such parameters carry [VERIFY] and conservative discounts.
• Aged-care and imaging calls/site/day are modelled, not published benchmarks.
• The DigiSurf dental case and VoiceStack's "78% industry answer rate" are single-vendor, unaudited claims [VERIFY].
• Per-exam imaging value should be re-derived from the IDX FY25 annual report (exact exams and fee-per-exam figures) before any page ships [VERIFY].
• The model treats every site as average; real networks have heavy-tailed site volumes.

Change log: v0.1-2026-07 — initial model.