Recurring Revenue: Subscriptions, Care Plans, Costs & Time

How the monthly-fee side of the business actually runs — what we sell, what it costs us, and how much of Daniel's week it eats. Operational playbook — derived from BUSINESS-MODEL.md + DECISIONS.md (settled) + BRAND-STRATEGY.md.

Read this as the answer to three questions Daniel keeps asking: what will the monthly subscription look like, what are our costs, and how much time will it take? The honest one-liner that governs everything below: recurring is the FLOOR, not the rate leader. A Care Plan nets R771/h — below the Repaint project rate (R848/h SA). Recurring exists to smooth revenue against project lumpiness and cover roughly half the R20k floor in dry months, not to beat project economics. Pricing/deposits/discounts on the one-time build fees live in Pricing & payments; this doc is only the recurring layer.

1. The recurring menu

Every recurring line we offer. Prices ZAR-first, USD/£ alongside; planning math uses the R16.0/$ buffer. Client-facing language stays outcome-led — the stack names here are internal only.

Voice is deferred and stays deferred until (a) the metering + monitoring substrate exists (Q3) and (b) ≥2 paying voice leads are in hand. It carries an implicit real-time SLA a day-job cannot honour. Note it, quote it only when both gates are met, and move on.

1a. Care Plan — what's IN vs NOT

Care Plan is best-effort, async, NO hard SLA. That honesty is what protects the day job — never promise an uptime number or a response-time guarantee.

Explicitly NOT in any Care Plan: new pages/features, redesigns, new integrations, e-comm/catalogue work, guaranteed response or uptime SLA, emergency/after-hours. All of those are new scope → quote as a build or bill at the change-order rate (R900/h · $110/h). Rule: upsell to the next package, never bend the current one.

1b. WhatsApp Receptionist — what's IN vs NOT

Broadcasts are always a separate, metered line item — never folded into R1,800. Meta marketing messages run ~R1.40/msg; 5,000 msgs/mo ≈ ~R7,500 in Meta fees alone. Price broadcasts at Meta cost + margin, billed on actual volume, quoted per campaign.

2. Costs — what each plan costs us

The whole point of the lean stack: self-built orchestration (Daniel's existing agent-infra pattern) + pay-as-you-go BSP (Twilio, ~$0.005/msg, no monthly). Do not use a bundled platform — WATI alone is ~R1,400/mo and eats the entire retainer before a single Meta fee. 360dialog (€49/mo flat) only wins at high volume.

The two Meta clocks — price against them from day one, not against today's free-window economics:

1. Oct 1, 2026 — Meta meters the free service window. Every competitor's blog price assumes free in-window replies; that ends. All figures above are already the post-October regime — quote on these, or margin quietly evaporates on live clients in Q4.
2. Aug 1, 2026 — Meta's native "Business Agent" starts billing (~4–5c/msg). Meta itself commoditizes the generic bot. Never pitch "we have a WhatsApp bot." Pitch what native can't do: real calendar/PM-system integration, SA payment-provider choice, CRM push, POPIA-aware health handling, bespoke multi-step flows.

Margin discipline: the ≤600-convo tier is the healthy one (64–86%). The busy 2,000-convo tier is ~30%+ and thin — at that volume infra alone approaches the fee before counting Daniel's time. Qualify prospects toward the low-volume end; if a client consistently runs high, move them to the Busy tier or reprice. The build fee (R15k) — not the retainer — is what earns the setup effort back.

3. Time — how much of the week

Budget 1.5–1.75 h/mo per recurring client. No real per-client data exists yet, so this is a deliberately conservative measurement target for clients 1–3 (time-tracking is on from client 1).

Hard KPI ceiling: ≤2 h/mo per client. This is the M6 checkpoint. If measured support exceeds it:

• >2 h/mo → warning; tighten scope, batch requests to a monthly window.
• >2.5 h/mo → the retainer is underpriced/over-scoped → reprice (per the M6 pivot: Care to R1,200 Basic / R1,800 Std) and cap scope harder.

Portfolio reckoner (for "how much of my week"): support time ≈ clients × ~1.75 h/mo.

• 6 recurring clients ≈ ~11 h/mo support (the R20k-floor scenario).
• 9 recurring clients ≈ ~16 h/mo support (the steady m9–12 scenario).

Framing to hold onto: recurring floors at ~R9–14k/mo net by Q4 — about half the R20k floor — before any project closes. It does not clear R20k on its own. It is ballast against dry months, bought with hours that earn less per hour than a build. Don't over-attach recurring at the expense of the project pipeline, which is where the rate actually lives.

4. Substrate — what must EXIST before selling each tier

Reuse Daniel's existing pattern (VPS + systemd + Discord + event-log + SQLite) — the monitoring muscle is genuinely de-risked. Stage the build; don't sell ahead of the substrate.

No hard SLA until Q4 (and only "SLA-lite" then). Best-effort retainers are the interim state by design — that is Daniel's stated constraint, not a gap to apologise for.

Interim overage gap (Proposed — not yet in DECISIONS.md): the R1,800 tier sells with a 600-convo cap + ~R3/convo overage now, but automated usage metering + overage billing is a Q3 substrate milestone. Until Q3, handle overage manually: read the Twilio/Meta dashboard once a month, flag the client at ~80% of cap, and bill overage on the next invoice by hand. Don't advertise hard real-time metering before it exists.

5. Billing & onboarding ops

5a. Onboarding a Care Plan retainer

1. Retainer starts at handover of the build it attaches to (Care Plan is a build convert, not a standalone sale).
2. Set up recurring billing: Paystack (SA) / Stripe (INTL), monthly, month-to-month.
3. Add the site to monitoring once Q2 alerting is live; log the client in clients/.
4. Send the one-paragraph "what your plan covers / what's extra" note (mirrors §1a) so scope is written down before the first request arrives.

5b. Onboarding a WhatsApp Receptionist — set expectations PRE-SIGNATURE

There is no same-day go-live. Say this before the client signs, so the lead time and the number requirement don't surface as post-signature surprises (this belongs in the Sales & intake conversation too).

• Meta Business verification: 2–14 business days (needs business-registration docs, 2FA/PIN) — the long pole.
• A dedicated phone number the client is not already using on personal WhatsApp (common objection — surface it early).
• Template message review — a separate Meta queue on top of verification.
• Display-name approval.
• Confirm the integration path in the audit/scope: Google/Outlook Calendar sync (psychologist MVP) or GoodX (dental, proven writable diary); Healthbridge = gated ~60-day partner process, expect friction.
• Payments: hosted link (Yoco / Paystack / PayFast) + webhook confirmation. Demo honestly as "tap, pay, get confirmed here" — never "in-chat checkout."
• POPIA: implement real opt-in, minimisation, deletion, human-in-loop for health-adjacent flows (it's hygiene, not a selling point — incumbents ship it).

Budget a 1–2 week buffer per client in the sales-to-live timeline. The R15k build fee covers this setup effort, so the retainer can stay honestly month-to-month.

5c. Term, cancellation & continuity

• Month-to-month, no lock-in — this is a settled brand promise (soft risk-reversal; the client owns everything). Do not introduce a minimum term or lock-in; the build fee — not the retainer — earns back setup, so no minimum is needed.
• Cancellation: month-to-month, 30 days' notice (Proposed — not yet in DECISIONS.md; docs say "month-to-month / no lock-in" but set no notice period). On cancel: hand over code/accounts, remove from monitoring/billing, archive in clients/.
• The continuity pitch for non-technical buyers (from BRAND-STRATEGY, order matters): lead with "a person on the other end, month-to-month" — then code/account ownership and written handover as the supporting detail. Code-ownership language reassures only technical buyers; the human-on-retainer line is what converts ICP-A.

5d. Destination hygiene (hard rules)

• Client-site submissions go to the CLIENT's inbox, never to Daniel's Discord (POPIA — berlein especially). The Discord/outbox pipeline is for Daniel's own funnel, not client customer data.
• NO hospitality clients, ever. No recurring line is ever routed to a hospitality lead; the restaurant app is anonymised capability proof only, never a referral seed.

Open decisions (→ DECISIONS.md)

Flagged Proposed above — new operational choices not yet in the settled log; Daniel to ratify or amend:

1. Care Plan tier scope split — Basic = ~2 small content changes/mo; Standard = reasonable-use batched monthly + a monthly health check. (Docs set the prices R900/R1,500 but don't differentiate what each tier includes.)
2. Best-effort response window — state "async, ~2 business days, no guaranteed uptime SLA" as the published habit. (BRAND-STRATEGY says "stated response habit" but names no number.)
3. Interim overage handling (before Q3 metering) — manual monthly dashboard read + ~80%-of-cap heads-up + hand-billed overage.
4. Cancellation notice period — month-to-month with 30 days' notice. (Docs settle "no lock-in" but no notice period.)