The 90 days have one job: to find out cheaply whether The wedge is right before the sending engine absorbs the budget. The plan front-loads the tests that cost founder time (interviews, API checks, an MX scan, supplier calls) and starts the expensive work, the engine, only after the 6 November gate. By 10 January 2027 Julian should know whether agencies use the desk, whether they pay, and what his mailbox cohort says about Google, Microsoft and private SMTP.
This plan assumes Julian plus two engineers (one on mail protocols and the scheduler, one full-stack on the desk), with a part-time deliverability contractor from week 6. That is below the 3–4 engineers Build costs and team budgets for an MVP sender, which is why the sender is only started here, not finished.
Week by week
| Wk | Dates | Sell and validate | Build | Compliance and infra | Gate |
|---|---|---|---|---|---|
| 1 | 12–18 Oct | Target list of 150 operator agencies; book interviews; script | Data model with external_ref and scopes (MVP spec); Smartlead webhook ingest | MX scan of 50,000 EU company domains; two CASA quotes; Microsoft partner account, Entra app, publisher verification; ask two law firms for fixed fees | K1–K13 signed off |
| 2 | 19–25 Oct | Interviews 1–8 | Test reply-send through Smartlead, Instantly, EmailBison APIs; classifier v0 on a labelled sample | Calls with InboxKit, Zapmail, Mailforge: tenant ownership and reseller status in writing | K3, 23 Oct |
| 3 | 26 Oct–1 Nov | Interviews 9–18; EUR landing page and LinkedIn ad test | Approval queue, hard opt-out, global objection list | Commission legal opinions; publish country ruleset v0 | — |
| 4 | 2–8 Nov | Interviews 19–30; LOIs | Instantly and EmailBison connectors | Order 60 cohort mailboxes and domains; authentication | K1, K2, 6 Nov |
| 5 | 9–15 Nov | First 3 design partners onboarded | Client portal, client-viewer role; engine starts: Mailbox interface, IMAP/SMTP adapter | Cohort ramp begins on real sends | — |
| 6 | 16–22 Nov | 6 partners live | Meeting ledger, weekly client report; scheduler, idempotent send | Deliverability contractor starts | — |
| 7 | 23–29 Nov | 10 partners live; Microsoft checker launched | Graph adapter; ingestion, bounce and auto-reply parsing | — | — |
| 8 | 30 Nov–6 Dec | First pilots end; founding-price conversations | Mailbox health v0: rolling bounce, adaptive caps | Sign partner inbox API terms | — |
| 9 | 7–13 Dec | Partner reviews; usage audit | Country rule engine v1, Art. 14 injection; Google OAuth path in dev | Submit Google brand and scope verification; book CASA lab | K4, 11 Dec |
| 10 | 14–20 Dec | First invoices at €149 | Own outbound runs on the engine over cohort mailboxes | Legal opinions due | K13, 18 Dec |
| 11 | 21–27 Dec | Microsoft reconnect campaign | Hardening; benchmark method written | — | — |
| 12 | 28 Dec–3 Jan | Support partners through SMTP AUTH default-off | Light week | Watch Microsoft Message Center | — |
| 13 | 4–10 Jan | 90-day review with partners | Q1 plan: first client migration, K6 A/B | Cohort day-60 readout | K7, 8 Jan |
Weeks 1–4: validate before building the expensive part
Interviews. Thirty structured calls with operator agencies (5+ clients, 50k+ sends a month) in the UK, France, the Nordics, Benelux and DACH, found through the public lists in Go-to-market plan. Contact each one through a channel that is lawful for their country: email to UK companies or French agencies, LinkedIn for Danish, German or Austrian ones (EU/EEA country matrix). The script asks for:
- the three biggest operational problems, unprompted, and then ranked against a list (per-client fees, reply handling, client reporting, deliverability, support);
- clients, mailboxes, sends and current spend, to check the 10-client arithmetic;
- who handles replies, how fast, and how meetings are proven to clients;
- reaction to a per-held-meeting fee at €5, €10 and €15, and to the flat €199 desk (Pricing strategy);
- whether a client, DPO or lawyer has asked "is this allowed?" in the past year (EU-native compliant outbound);
- the split of Google, Microsoft and SMTP in their sending infrastructure.
Gate on 6 November: K1 (≥10 of 30 rank the pains, ≥8 paid LOIs) and K2 (compliance premium) in Kill criteria. If K1 fails, the engine does not start in week 5.
API feasibility (K3, 23 October). Whether a third party can post a reply into an existing sequencer thread is not documented anywhere in this hub. Test it on live accounts in week 2. If it fails on two of three, the desk drafts and approves and the operator sends from the sequencer.
MX scan. Classify 50,000 EU company domains by mail provider. It costs a day and answers two open questions: the Microsoft 365 share among European recipients, and whether GMX/WEB.DE, T-Online and IONOS matter as destinations (Open questions, ESP matching).
Supplier calls. Ask InboxKit, Zapmail and Mailforge, in writing, who owns the tenants and whether they hold Google or Microsoft reseller status. Maildoso already says new Google Workspace allotment is "not currently available to purchase" (Infrastructure strategy: build or partner).
Weeks 5–9: design partners and the desk
Onboard partners in waves of three or four, each with at least three client workspaces. Every partner gets a shared Slack channel and a named engineer (Trust as positioning). Pilots run 30 days free, then convert at the €149 founding price, so the first invoices go out in week 10.
Measure from day one: the share of each partner's inbound replies handled in the desk rather than the sequencer's own inbox, time from "interested" reply to first response, opt-outs suppressed and their latency, meetings booked versus held, and disputes. The K4 gate on 11 December needs 10 partners live with ≥60% of replies through the desk.
Weeks 5–13: start the engine, do not finish it
Engine work begins only if K1 passes, and in the order Sending architecture argues: the Mailbox interface with IMAP/SMTP and Graph adapters; the per-mailbox scheduler with idempotent sends; ingestion with a reconciliation poll and RFC 3464 bounce parsing; then Mailbox health engine v0 and the country rule engine. Its first traffic is Julian's own outbound to opt-out markets, which dogfoods the rule engine and feeds the cohort (Go-to-market plan).
Target for the end of week 13: the engine sends and ingests reliably on the cohort mailboxes. Migrating a partner's first client is a February 2027 task, followed by the K6 A/B test against the incumbent, due by 30 April.
OAuth and CASA timing
| Step | When | Why then |
|---|---|---|
| Microsoft partner account, Entra multitenant app, publisher verification | Week 1 | Free, and users cannot consent to unverified multitenant apps asking for mail scopes |
| Two CASA quotes | Week 1 | Prices range from about $540 to $6,000 a year by lab and level |
| Google brand and restricted-scope submission | Week 9 | Needs a working Google OAuth path to demo; app passwords cover Google until then (OAuth verification and the end of basic auth) |
| CASA assessment | January–March 2027 | Budget 6–10 weeks of calendar time including remediation (OAuth verification and the end of basic auth); due by 30 April (K8) |
The desk itself needs no Google verification, because it reads replies through the sequencers' APIs, not through Gmail.
Deliverability tests
The 60-mailbox cohort from Open questions: 20 Google, 20 Microsoft and 20 private-SMTP mailboxes, two suppliers each, ramped on real, compliant B2B sends at 15–25 a day with no warmup pool. Track suspensions, hard bounces, deferrals and human replies per mailbox. Day-60 readout on 8 January (K7). Day 90 in February. The cohort costs well under €1,000 for the quarter at $3–4.50 a mailbox (Go-to-market plan budget).
Exchange Online Basic SMTP AUTH becomes disabled by default for existing tenants at the end of December 2026. Any partner whose Microsoft mailboxes still use it will see permanent 550 5.7.30 failures. Weeks 11–12 are reserved for helping them reconnect over OAuth, and the free checker launched in week 7 should have found them first (Microsoft rules: Outlook.com and Microsoft 365 limits).
What day 90 must show
| Question | Evidence on 10 January 2027 | Criterion |
|---|---|---|
| Do agencies have the pain? | 30 interviews, ranked pains, ≥8 LOIs | K1 |
| Will they pay for Europe? | Compliance willingness to pay | K2 |
| Can the overlay work? | Reply-send tested on three sequencers | K3 |
| Do they use it? | 10 partners live, ≥60% of replies in the desk | K4 |
| Will they pay? | First invoices sent; on track for 6 paying by 29 Jan | K5 |
| Is the legal posture containable? | Two written opinions | K13 |
| Which infrastructure survives? | Cohort day-60 suspension and bounce data | K7 |
| Is Google on schedule? | Verification submitted, CASA lab booked | K8 |
What this means for an entrant
- Do not start the engine before the interview gate. Five weeks of founder time are cheaper than five months of two engineers (Kill criteria).
- Invoice before Christmas. A design partner who has never paid is a friend, not a customer (Pricing strategy).
- Use the calendar the incumbents hand you. The Microsoft deadline is a ready-made migration campaign (Go-to-market plan).
- Start Microsoft verification in week one and Google in week nine. The first is free and fast; the second needs a working app and a lab slot (OAuth verification and the end of basic auth).
- Make your own outbound the engine's first customer. It is the cheapest deliverability test and the most honest demo (MVP spec).