# Pricing inputs established in selected source packages

Version 1.0. Secondary descriptive analysis of the **17 November 2026** cost-research snapshot. The question is which budget inputs those retained official-source records establish. This secondary analysis applies a retrospective coding framework to an existing price-research roster.

## Inputs and selection

The source of observations is the existing `pricing-source-register.csv` and `pricing-inputs.csv` at `/resources/ai-receptionist-cost-australia/data/`. The original `method.md` and `manifest.json` are also frozen by SHA-256 in `frozen-inputs.json`. No original file is changed. Source dates stay 17 November; no new retrieval, vendor test, quote request or customer data was performed for this secondary analysis. Public extraction can return cached text. The retained register contains fact summaries and source locators. These support reproduction of the counts. Full historical pages would be needed to assess any disclosures beyond those records.

The roster is inherited from the cost guide: Valory AI, Johnni AI, Nexwin, Smith.ai, RingCentral, BotBloke and TransferToAI. It is a purposive existing roster, not a random or complete Australian-provider sample. It omits other services and does not establish Australian availability for every listed option. Five provider source packages have retained readable pricing evidence. Two have access failures; failures do not prove an absent offer or absent disclosure.

There are **12 worksheet rows**, comprising ten evidenced plan/payment rows and two unverified quote placeholders. The evidenced rows contain four Valory tiers, monthly/annual Johnni Pay As You Go alternatives, Nexwin Business, Smith.ai Free/Pro and the RingCentral AI Receptionist add-on. Johnni AI counts once at provider level despite having two payment rows. BotBloke and TransferToAI appear only as unavailable placeholders. The selected offers span different service models and scopes, including an add-on requiring another subscription.

## Units, eligibility and codes

`fact-coded.csv` contains one row for one assessment unit and field:

- `PROVIDER_SOURCE_PACKAGE` counts a provider once per policy field, even when several plans reuse its source. Nine fields across seven providers give 63 coded rows.
- `PLAN_PAYMENT_ROW` counts each selected worksheet row once per plan field. Five fields across 12 rows give 60 coded rows. Only ten rows have readable plan evidence; placeholders remain unassessed.

Never combine these units into one percentage or calculate a completeness score. The default policy denominator is the five readable provider packages. Source access itself uses all seven selected providers. Main-voice time-rounding/minimum fields use four readable providers with at least one evidenced minute-based plan. Smith.ai's selected main-voice plans bill per call, so those two time-policy fields are outside the defined eligibility rule. That exclusion says nothing about other duration limits or transfer charges. Plan-field denominators use ten evidenced plan/payment rows, with the two unavailable placeholders separately shown.

Codes:

| Code | Meaning |
|---|---|
| `ESTABLISHED` | The specified input/basis is established by retained accepted source evidence. It may be a floor, a zero standard setup fee, an allowance or a rate; it is not necessarily a final quote. |
| `UNKNOWN` | Not established in this bounded retained package. Do not interpret as absent from the provider's website or unavailable from sales. |
| `UNAVAILABLE` | The selected pricing source could not be retrieved. Do not recode its workbook defaults as facts. |
| `QUOTE_REQUIRED` | The source identifies separately quoted scope or an unselected required base component; no amount is inferred. It does not mean every buyer must purchase every optional integration. |
| `NOT_APPLICABLE` | The field's narrow time-based main-voice eligibility rule does not apply to selected per-call plans. This is not a favourable or unfavourable outcome. |

`eligibility` makes inclusion explicit: `ALL_SELECTED`, `READABLE_SOURCE`, `READABLE_MINUTE_BASED_SOURCE`, `READABLE_SELECTED_PLAN`, `SOURCE_UNAVAILABLE` or `NOT_APPLICABLE_SELECTED_VOICE_UNIT`. Every row retains provider, stable row ID where relevant, source IDs, rationale and historical snapshot date. Source IDs resolve to the original register and must belong to the same provider.

## Field definitions

Provider fields:

1. `source_access`: at least one retained readable pricing record versus recorded failed access.
2. `currency`: explicit currency identity; a regional URL or dollar sign alone is insufficient under this snapshot's conservative rule.
3. `gst_basis`: explicit inclusion/exclusion of GST. No buyer tax-credit entitlement is assessed.
4. `setup_basis`: an explicit standard setup amount/basis, including a published minimum or explicit no standard setup charge. A floor is established as a floor, not as the final setup quote.
5. `sms_charge_basis`: an established allowance/extra-unit rate or priced flat add-on. Segmentation, fair-use limits and delivery are separate questions and are not implied by this code.
6. `transfer_wait_charge_basis`: sufficiently explicit treatment of both transferred/waiting time to price it. A transfer feature is not a billing policy.
7. `voice_rounding_increment`: explicit time increment for main-voice billing, among eligible minute-based packages. Only published increments count.
8. `voice_minimum_duration`: separately stated minimum billable time, among eligible minute-based packages. A rounding rule alone does not establish a separate minimum. Only published minimums count.
9. `additional_scope_or_base_price`: an explicit per-business quote dependency for selected setup/custom work, a job-software integration or a required separate base subscription. These different dependencies remain labelled, not ranked. Otherwise unknown.

Plan/payment fields are `headline_amount`, `voice_billing_unit`, `included_voice_allowance`, `extra_voice_unit_rate` and `payment_basis`. Establishing an amount does not establish its currency, tax or equivalent service scope. A published starting amount is retained as such. The Nexwin offer retains its original alternate-amount qualification. The Smith.ai Free amount preserves its published zero. The placeholder MONTHLY defaults are ignored. Payment basis describes the source's billing cadence or advertised price period, not a minimum term or cancellation requirement. Johnni's retained annual record J02 establishes annual payment and a monthly-equivalent amount. The annual row's unit, allowance and extra rate are unknown because only J01, explicitly identified as the monthly price record, documents them. The original cost worksheet carries those inputs to the annual row for modelling; the secondary coding requires a retained source cue for that payment option.

The coding decisions are explicit in the CSV and source IDs. Counts use the status code for each row and retain unknowns. One analyst produced this coding. The automated checks and separately stated fixtures verify the package and calculations. The method does not include independent recoding.

## Calculation, files and reproduction

`counts.csv` retains total selected units, eligible denominator, all five code counts, unavailable exclusions and non-applicable exclusions for each field. Counts across all codes reconcile to seven provider packages or 12 worksheet rows. `summary.json` adds the distinct provider/row inventory and descriptive billing-unit/payment distributions. `plan-row-index.csv` identifies ten evidenced rows and two placeholders. `quote-questions.csv` is a blank practical buyer record, not observed quotes or customer data.

`analyse.py` uses Python standard library only. It checks frozen source and method hashes, exact rosters/fields, unique assessment IDs, code/eligibility consistency, provider-specific source links, placeholder exclusions and category totals. It regenerates only files in this new report's data folder and never writes the cost package. `test_analyse.py` supplies separately stated expected counts and deliberate malformed-row fixtures; it does not contact providers.

From a repository checkout:

```bash
python3 -B public/reports/ai-receptionist-pricing-disclosures-2026/data/test_analyse.py
python3 -B public/reports/ai-receptionist-pricing-disclosures-2026/data/analyse.py --check
```

For standalone reproduction, download this report’s files into one directory and the four original input files into a child directory named `cost-data`. From the report directory run `python3 -B test_analyse.py` and `python3 -B analyse.py --check`. The analysis script also accepts `--source-dir /absolute/path/cost-data` for inputs stored elsewhere. Run without `--check` to intentionally regenerate the three analytical outputs, the method hash and the package manifest. The manifest covers both the report inputs and outputs. Hashes verify file integrity. If the original input hashes differ, use the dated files or version a separate analysis with a new observation date.

No significance test, confidence interval, national prevalence estimate, causal comparison, subjective score, provider league table or invoice accuracy claim is warranted. Missing evidence and source-access failures remain visible beside each finding. The report does not measure actual spend, call quality, form/booking functionality, enquiries, revenue or savings. Valory AI produced the source research and this secondary analysis and sells managed AI reception; its commercial interest is disclosed.
