[P1] Refresh monetization assumptions and maintain an authoritative source register

[postigodev]

Context

The current business-plan pricing and COGS assumptions were useful for an academic forecast, but they do not yet represent a production subscription model. Important examples include a flat payment-processing percentage, a voice cost assumption that is not tied to duration, and an annual price that was selected before real retention or provider usage existed.

Pricing claims must be traceable to current authoritative sources and separated from internal hypotheses.

Current verified baseline

Verified on 2026-07-16/17. Re-verify before publishing or using in a financial decision.

Category	Current baseline	Authoritative source
OpenAI model used by repo	gpt-5.4-mini; $0.75 / 1M input tokens and $4.50 / 1M output tokens	https://developers.openai.com/api/docs/models/gpt-5.4-mini
Lower-cost OpenAI comparison	gpt-5-mini; $0.25 / 1M input and $2.00 / 1M output	https://developers.openai.com/api/docs/models/gpt-5-mini
ElevenAgents voice	Charged by connected call duration with 95% discount for silence longer than 10 seconds; LLM cost passed through separately; included minutes and credit conversion depend on plan	https://help.elevenlabs.io/hc/en-us/articles/29298065878929-How-much-does-ElevenAgents-cost
Stripe Payments	2.9% + $0.30 per successful domestic card transaction	https://stripe.com/pricing
Stripe Billing PAYG	0.7% of Billing volume	https://stripe.com/billing/pricing
Stripe Tax Basic	0.5% of volume for supported no-code transactions where registered to collect tax	https://stripe.com/tax/pricing
Vercel Pro	$20/month with $20 included usage credit	https://vercel.com/pricing
Supabase Pro	From $25/month; first project included, additional projects from $10/month	https://supabase.com/pricing
Railway	Hobby $5/month; Pro $20/month; subscription counts toward resource usage	https://docs.railway.com/pricing
Samsung Food+ comparison	$6.99/month or $59.99/year; pricing may vary by location	https://support.samsungfood.com/hc/en-us/articles/32709269852052-What-s-Included-in-Your-Samsung-Food-Subscription

Initial pricing hypothesis to document

• Founding beta: $6.99/month.
• No annual plan until measured retention and at least two complete billing cycles of provider-cost data exist.
• Remote voice must be limited/metered; local guided cooking and timers should be modeled separately.
• Permanent plan price and allowances are hypotheses to validate with real checkout conversion, retention, and contribution margin.

Scope

Source register

• Create a version-controlled monetization source register in repository documentation.
• For every source record: vendor, product, metric, value, unit, currency, geography, effective/published date when available, verification date, authoritative URL, and notes.
• Prefer official vendor pricing/docs/support pages over blogs, aggregators, or secondary summaries.
• Mark promotional, enterprise, grant, annual-contract, included-credit, and tax-exclusive pricing explicitly.
• Preserve prior values and effective dates instead of silently overwriting history.

Financial model refresh

• Replace flat per-user COGS assumptions with event/duration/token/storage/transaction-based drivers.
• Separate registered, active, AI-active, voice-active, paid, trial, and heavy-user cohorts.
• Model Stripe’s percentage and fixed transaction fee separately, plus Billing and optional Tax fees.
• Model voice using connected minutes, discounted silence/provider-billed minutes, LLM pass-through, and plan-specific included credits.
• Separate fixed infrastructure from variable COGS.
• Add conservative/base/heavy usage scenarios and contribution margin by plan.
• Remove or flag the $59.99/year assumption until re-approved.

Documents and product claims

Audit and update where applicable:

• Business plan financial assumptions and source notes.
• Pitch-deck pricing and margin claims.
• README/docs monetization statements.
• Landing/pricing page copy.
• Internal issue descriptions that cite old provider prices.
• Any investor, grant, or accelerator material reused from the prior forecast.

Review cadence

• Re-verify provider and competitor pricing monthly during beta and before every public pricing change.
• Reconcile source-register assumptions with actual provider invoices and [P1] Attribute provider usage and COGS by user, feature, and plan #122.
• Record the approval date and evidence for any annual plan launch or permanent price change.

Acceptance criteria

• Every externally sourced pricing assumption has an official URL and verification date.
• Internal hypotheses are labeled as hypotheses rather than presented as vendor facts.
• The model uses token, duration, transaction, storage, and infrastructure drivers instead of one flat paid-user COGS value.
• Stripe’s fixed fee and Billing fee are represented correctly.
• Voice modeling distinguishes connected duration, discounted silence/provider-billed usage, and LLM pass-through.
• The old annual price is removed or explicitly marked inactive across current product and financial materials.
• Business-plan and pitch claims agree with the source register.
• A monthly source-verification owner and process are documented.
• Pricing decisions reference measured data from [P1] Attribute provider usage and COGS by user, feature, and plan #122 after beta begins.

Related

• Stripe billing: [P0] Build Stripe subscription billing and customer lifecycle #118.
• Entitlements and usage: [P0] Define plan entitlements and durable usage accounting #119.
• Pricing/billing UX: [P1] Ship pricing, checkout, usage, and billing account UX #121.
• COGS attribution: [P1] Attribute provider usage and COGS by user, feature, and plan #122.
• Privacy/terms/payment disclosures: [P0] Publish privacy, terms, AI, and voice disclosures before public beta #99.

Out of scope

• Treating competitor price as proof that Preppie should copy competitor limits or economics.
• Legal or tax advice; obtain appropriate professional review where needed.

[mheilimo]

One addition would make the source register usable as an audit trail rather than only a current-price table: keep each assumption, evidence snapshot and derived value as separate records.

MonetizationAssumption
- assumption_id, metric, cohort and scenario
- state: observed | contracted | forecast | hypothesis
- value, unit, currency, geography, tax_basis and period
- formula_version and driver_refs[]
- evidence_refs[]: uri, locator, retrieved_at, effective_at, digest
- owner, reviewed_at, expires_at, supersedes and limitations

That supports a few acceptance tests which are easy to automate:

• a source change creates a superseding record and never rewrites the value used in an earlier forecast;
• monthly, annual, per-transaction, per-minute and per-token inputs cannot be combined until their units and periods are normalized;
• every derived COGS value can be replayed from a formula version and immutable input snapshots;
• observed invoice data and a forecast never occupy the same field or silently change state;
• missing or expired evidence returns insufficient_evidence, not zero cost;
• a competitor price can inform a comparison scenario, but cannot serve as evidence for Preppie's retention, usage or contribution margin.

I would also pin the exact source excerpt or table row with each URI. Pricing pages change in place, so a URL plus retrieval date alone may not explain a later variance.

I work on DDScore at Playful Pixels Oy. In an adjacent founder materials review, submitted financial claims, current public-source checks and confidence gaps remain visible inside a 12-dimension written analysis. This is not a Chef integration or a view on Preppie's pricing. DDScore does not provide investment advice or replace full diligence.