Kill criteria

Run 30 structured interviews with operator agencies (5+ clients, 50k+ sends a month) by 6 November 2026. Stop the agency wedge if fewer than 10 rank reply handling, per-client fees or client reporting among their top three problems, or fewer than 8 sign a paid design-partner LOI at a founding price. The grievances in Lead-gen agencies and Customer voice come from reviews and vendor case studies; interviews test whether they are common enough to switch for.

At the same interviews, ask for willingness to pay for per-recipient country rules, Art. 14 notices and a cross-client objection list. If fewer than 5 of 30 would pay a premium and fewer than 3 report a client, DPO or lawyer asking "is this allowed?" in the last year, the EU-native compliant outbound layer stays in the product as the default but is dropped from pricing and from the headline. This is a re-scope, not a stop: EU-native compliant outbound scores pain at 3, and Woodpecker's EU base bought it no EU market.

By 23 October 2026, test on live accounts whether a third party can send a reply into an existing thread through the Smartlead, Instantly and EmailBison APIs. Their reply webhooks are documented (Smartlead, Instantly), but a reply-send endpoint is not documented anywhere in this hub. If fewer than two support it, the desk ships as triage, approval, booking and handoff, with the operator replying in the sequencer, and the own engine moves forward by a month.

By 11 December 2026, at least 10 agencies should be live with at least 3 client workspaces each, and at least 60% of their inbound replies should be handled in the desk rather than the sequencer's own unibox, measured over the trailing two weeks. If not, the desk is a nice-to-have. Pivot toward whatever job the adopters hired it for.

By 29 January 2027, at least 6 design partners should pay at least the €149 founding price, after a 30-day free pilot. Fewer than 3 paying ends the desk-first sequence. Also count meeting disputes: if more than 30% of "held" meetings billed under the optional per-meeting plan are disputed by the agency's clients, drop per-meeting billing and price the desk flat (Pricing strategy).

Stop, or raise money only on a different thesis, if MRR is below €10k on 30 June 2027 or below €20k on 31 October 2027, or if runway falls under six months without a funded path. These are my targets, not benchmarks. For calibration, Instantly reported $2.4M ARR nine months after launch, starting from its own agency book.

From February 2027, split comparable lists for at least 3 design partners: one arm on the incumbent, one on your engine, at least 20,000 sends per arm, same copy and mailbox supplier. By 30 April 2027, your arm must reach at least 80% of the incumbent arm's positive replies per 1,000 sends, with hard bounces under the 2% Instantly treats as the ceiling. Sanity check: Danish Lead Co. reports 300–500 emails per positive reply on typical campaigns (vendor case study). If your arm loses, stop migrating sending. Stay an overlay plus client portal, and treat the engine as an R&D line until Mailbox health engine closes the gap. Both arms use the same reply classifier (Reply benchmarks).

The 90-day mailbox cohort from Open questions (20 Google, 20 Microsoft and 20 private-SMTP mailboxes across two suppliers, 15–25 compliant sends a day) reads out at day 60 on 8 January 2027 and at day 90 in February. If more than 20% of a supplier's Google mailboxes are suspended within 90 days, close to EmailBison's unverified 22% claim, drop that supplier and stop offering Google inbox purchase. Steer partners to Microsoft and bring-your-own (Infrastructure strategy: build or partner). This is a re-scope, not a stop.

Google brand verification and restricted-scope review start once the Google OAuth path exists (target December 2026). If Google verification and CASA are not complete by 30 April 2027, or the assigned assessment costs more than €15k a year all-in, stay on app-password IMAP/SMTP for Google and Graph for Microsoft. Listed lab prices run $3,000–6,000 for AL1 at Leviathan and lower at others per Nango (OAuth verification and the end of basic auth). It is a stop condition only in combination with K9's app-password trigger.

Three triggers, watched monthly (Provider crackdowns): (a) Google announces the end of app passwords for IMAP/SMTP with under six months' notice while CASA is not passed: pause Google onboarding, push Microsoft and bring-your-own-OAuth. (b) A suspension wave hits 25% or more of design partners' Google mailboxes within 30 days, the share one vendor estimated for autumn 2025: switch on automatic re-homing and stop selling Google inboxes. (c) Google and Microsoft both act against connected cold-email tools by name, for example revoking restricted-scope apps or shipping Microsoft's promised "adaptive" limits against provisioned tenants: stop the sender and keep only the desk.

Affiliates start in H2 2027 at 35% lifetime, paid from the referral's third paid month (Pricing strategy). By 30 September 2027, affiliate-referred cohorts must show monthly logo churn of 5% or less. Instantly lived with about 10% early on (Distribution). If three consecutive monthly cohorts churn above 10%, close the programme to creators and keep it for operator agencies. Separately, cut any paid channel whose blended CAC payback exceeds nine months of gross margin for two quarters running.

Check quarterly. Reposition if two of these happen: Smartlead makes client workspaces and white-label free and fixes scoping; Instantly ends or ring-fences VIP; lemlist ships recipient-country gating plus an EU data region; Instantly or Smartlead ship client-scoped approval queues free on agency tiers. Two together remove the median agency's reason to switch (Agency operating system, The reply desk). Lean harder on what is slow to copy: the sourced ruleset, the EU entity, support.

Commission two fixed-fee opinions (UWG and data protection) on the platform's own liability in October 2026, deliverable by 18 December 2026 (Platform liability: what the sequencer itself risks). If counsel concludes that an EU-domiciled sender faces Störer liability for customers' sends to German recipients even with default blocking, a consent ledger and AUP enforcement, change the posture before public launch: insurance, terms, and DE/AT blocking that customers cannot override. Stop only if the exposure cannot be insured or contained.