• TAM ≈ €4.2B — EU + UK D2C subscription commerce revenue-ops spend, derived from ~280k subscription-commerce companies × ~€15k/yr blended spend on CRM/retention tooling and allocated headcount [HYPOTHESIS, Channel A]. This is a top-down estimate built from Dealroom and Statista e-commerce segmentation; it has not been triangulated against bottom-up operator interviews.
• SAM ≈ €420M — the €5–50M ARR band specifically, estimated at ~28k EU D2C subscription companies meeting that revenue threshold [HYPOTHESIS, Channel A]. This is the band where a dedicated Head of Subscription exists (so there is a buyer) but the team is still small enough to feel pain (so there is urgency).
• SOM ≈ €8M ARR by Year 3 — 400 customers × €20k blended ACV, assuming a dunning-wedge entry compresses the sales cycle to under 60 days [HYPOTHESIS, Channel A].

What would have to be true for SOM to land:

• Dunning-wedge trial-to-paid conversion ≥ 35% [HYPOTHESIS] — defensible only if the recovered-MRR delta is observable within 30 days of activation.
• Expansion from dunning-only to multi-lever NRR agent ≥ 60% within 12 months of initial contract [HYPOTHESIS] — without this, ACV stays at €24k and the €20k blended assumption breaks.
• Recoverable churn share ≥ 30% of total churn [HYPOTHESIS, Channel C falsifier] — if it's lower, the ROI per customer compresses below the threshold where the performance kicker is meaningful, and renewal becomes a price negotiation rather than a no-brainer.

Judge's note on sizing rigor: All three sizing numbers carry [HYPOTHESIS]. The TAM is directionally fine for an investor conversation but is not a number S3 should design product capacity against. The load-bearing number for S3-S6 is SOM Year 1, which under the dunning-wedge framing should be re-estimated bottom-up after the 15-interview block in §9 — not inherited from this document.

ICP. Head of Subscription, CRM Director, or Subscription Operations Lead at a French, German, or Benelux D2C brand in wine, food-box, beauty, or specialty grocery, doing €8–25M ARR, running Chargebee or Recurly as the billing system of record, with a team of 2–3 ops people. Critically, the ICP must be autonomy-tolerant — i.e., willing to grant the agent write-access to billing APIs after a shadow-mode trial. The autonomy-intolerant variant of this same persona is a different buyer who purchases a recommender tool at a fraction of the ACV and should be excluded from SOM math until proven otherwise.

Persona — "Camille, Head of Subscription at a €15M ARR wine brand." Reports to the COO, owns NRR as a board-level KPI, and has 2 ops people running Chargebee dashboards manually. She is not technical (won't read an API doc) but she is operationally fluent — she knows her current 3-email dunning sequence is embarrassing and she's been meaning to rebuild it for 18 months. She will sign a €2k/mo contract on the strength of one demo if it shows one number moving on her actual data.

Jobs-to-be-done:

1. Recover involuntary churn from failed payments without manually exporting CSVs every Monday morning or asking the ops team to call €30/mo subscribers (which never happens).
2. Personalize win-back and save offers beyond the generic "we miss you, here's 20% off" that every brand sends and that subscribers have learned to ignore.
3. Have a defensible NRR story for the monthly board deck — a narrative of "here's what we tried, here's what worked, here's what's next" that doesn't reduce to "we ran the same campaign as last month."

Voice (Channel C operator quote, kept verbatim because it's the cleanest articulation of the wedge):

"Honestly, our dunning is embarrassing — it's three emails and a retry on day seven, and I know we're losing good subscribers who just had a card issue. I keep saying I'll fix it but it's never the most urgent thing. Last month we recovered maybe 40% of failed payments; I have no idea if that's good or terrible. If something could just… handle that automatically and tell me what it recovered, I'd pay for that tomorrow."

1. Involuntary churn from failed payments. Pain: 0.8–1.5% MRR/month bleeding through generic retry schedules [CLAIMED, Channel C]. Status quo: a 3-email sequence configured in Chargebee in 2021, untouched since, plus a weekly CSV export to a spreadsheet that someone is "supposed to call through" but rarely does. Cost: on a €10M ARR account, a realistic 0.4% MRR/month recovery gap translates to ~€40k/year in recoverable revenue left on the table [HYPOTHESIS] — roughly 1.5× the SaaS fee.

2. Generic, non-segmented win-back campaigns. Pain: the same "we miss you, 20% off" sent to a 6-month subscriber who churned over price and a 36-month subscriber who churned over product variety. Status quo: a quarterly batch campaign in Klaviyo or Mailchimp, A/B tested on subject line only. Cost: typical win-back conversion at 1–3% [HYPOTHESIS] against a theoretical 8–12% ceiling for personalized outreach — opportunity cost is the entire reactivation funnel.

3. No human contact with at-risk subscribers below €50/mo LTV. Pain: it is economically irrational to have a human call a €30/mo subscriber, even though the LTV of a saved subscriber is €600+. Status quo: nothing. The churn happens silently. Cost: the entire long-tail of high-LTV-but-low-MRR subscribers churns without intervention, which is exactly the bucket where conversational save logic would have highest ROI.

4. Pricing changes are board-meeting events, not continuous experiments. Pain: a €2 price increase is debated for 3 months and rolled out once a year, with no segmentation. Status quo: pricing decisions made in PowerPoint, executed in a single Chargebee config change. Cost: the contribution-margin upside from segment-specific price testing is structurally inaccessible to a 2-person ops team.

5. The Head of Subscription has no defensible narrative in board meetings. Pain: NRR lands at 94% again, the CEO asks why, the answer is "we ran the same playbook." Status quo: monthly cohort reports built manually in Looker or a spreadsheet. Cost: career risk for the persona — this is the soft underbelly that makes the €2k/mo budget approval frictionless.

Direct competitors (named, with positioning + pricing band):

• Churnkey (US, focused on cancel-flow optimization, ~$500–2k/mo). Positioning: "smarter cancel flows + dunning." Not autonomous — recommends and executes within narrow guardrails. [VERIFIED — public pricing & positioning]
• Stunning / Baremetrics Recover (US, dunning-focused, $100–500/mo). Positioning: better retry logic + email sequencing. Rules-based, not agentic. [VERIFIED]
• Retention.com / Bounce (US, retention messaging tools, variable pricing). Positioning: identity resolution + messaging — adjacent rather than head-on. [VERIFIED]
• Vitally / ChurnZero (US, B2B SaaS customer success platforms, $1k–10k/mo). Positioning: wrong ICP (B2B SaaS) but cited by operators as the reference category. [VERIFIED]

Indirect competitors (the duct-tape stack people actually use): Chargebee/Recurly native dunning rules + Klaviyo/Mailchimp for win-backs + a Looker dashboard for cohort reporting + a Google Sheet for the failed-payment call list + a 2-person ops team as the orchestration layer. This stack costs ~€200–400k/year fully loaded [HYPOTHESIS] and is what Subscript is actually displacing.

Adjacent / pivot-in risk:

• Chargebee — probability of shipping an "AI agent mode" within 18 months: high (60-70%) [HYPOTHESIS]. They have the data, the integrations, and a clear strategic motive. Job posts already signal data-platform investment [VERIFIED, Channel A].
• Recurly — probability of same within 24 months: medium-high (50%) [HYPOTHESIS]. Slower roadmap but identical structural position.
• Stripe Billing — probability of an agentic retention layer within 24 months: medium (40%) [HYPOTHESIS]. Stripe has the broadest distribution but has historically left vertical retention logic to partners.
• Klaviyo — probability of pivoting into agentic NRR: low-medium (25%) [HYPOTHESIS]. They own the messaging layer but not the billing write-access, which is the load-bearing capability.

Case vide diagnosis. The white-space Subscript targets is the seam between billing infrastructure (Chargebee/Recurly own this) and messaging infrastructure (Klaviyo/Mailchimp own this). No incumbent currently owns the decision layer that sits between them — the layer that decides which subscriber gets which intervention at which moment. Today that layer is a human ops team with a spreadsheet. Subscript's bet is that this seam is a defensible category before the platforms above and below it close in. Judge's call (siding with Sequoia over Lenny on this specific point): the competitive window is 12–18 months, not 24. Lenny says "don't spend cycles worrying about incumbents yet" — that is correct for prioritization but wrong for architecture. S3 must design for the case where Chargebee ships a competing feature in month 14, which means the moat has to be in playbook quality and named case studies, not in the integration itself.

Martin's asset stack, decomposed:

1. 8 years operating a €40M ARR subscription wine business (LPB) [CLAIMED — needs verification via LinkedIn, LPB press, or reference call]. If true, this places him in the top ~50 EU D2C subscription operators by revenue scale.
2. Authored the churn-save and dunning playbooks circulating in EU D2C operator communities [CLAIMED — verifiable by asking 3 operators independently whether they recognize his name and his playbooks]. This is the load-bearing claim. If two of three operators in the first interview block confirm unprompted, this is verified.
3. Personal network of ~12 Heads of Subscription at EU D2C brands [CLAIMED, Channel C]. This is the first-5-customers channel and the only GTM that matters in the first 90 days.
4. Founder-PM-builder combo [CLAIMED] — operator depth + agent-building skill. The "builder" half is the least verified claim in the entire deck. S2 cannot confirm Martin has shipped production-grade agentic software; this is the single biggest team risk and should be probed before S5 commits to a build plan.

What competitors cannot replicate in 12 months: the playbooks (tacit knowledge, takes years to develop) and the network (relationships, not buyable). What they can replicate: the integration with Chargebee (90 days of engineering) and the LLM orchestration layer (commoditizing fast). This is why the wedge must be playbook-first and case-study-first, not technology-first. The defensibility lives in the named customer logos and the documented NRR deltas, not in the code.

Judge's verdict on founder asset: rare and credible, with one verified-needed flag on (4) above. Channels A and C both treat this as the strongest part of the pitch; Channel B implicitly agrees by making it central to the "compress trust-building from years to quarters" thesis. The wedge is real. Confidence on founder asset: 7.5/10, gated on a reference call confirming the builder claim.

Recommended pricing: €2,000/month base + performance kicker on measured recovered MRR.

• Base: €2k/mo for the dunning-only wedge [HYPOTHESIS — needs willingness-to-pay validation in the 15 interviews]. This is the price Channel C's operator quote ("I'd pay for that tomorrow") implicitly anchors. It is also low enough to be a no-quote, no-procurement decision for a Head of Subscription with budget authority — critical for a 60-day sales cycle.
• Performance kicker: 15% of recovered MRR above an agreed baseline, capped at 2× base fee. This aligns incentives to the metric Subscript wants to be measured on (recovered MRR, observable in 30 days) rather than NRR (12-month lagging, gameable).
• Expansion: as the agent earns write-access to additional levers (win-back, upsell, pricing), tier upward to €4k–8k/mo. Full-stack autonomous-NRR ACV target: €48–96k.

Revenue mechanic: hybrid SaaS retainer + outcome kicker. Channel A proposed this; Channel B is silent on pricing; Channel C anchors the €2k/mo base. Judge sides with the hybrid model over pure SaaS because (a) the outcome kicker is the only credible way to justify the "autonomous NRR ownership" positioning at premium ACV, and (b) it forces Subscript to instrument outcome measurement from day 1, which compounds into the case-study moat.

Year-1 ARR range: €240k–€480k — 10–20 customers at blended €2–2.5k/mo [HYPOTHESIS, derived from Channel A's 12-customer pilot scenario]. Load-bearing assumption: that Martin's network converts at ≥30% from warm intro to paid pilot within 60 days. If conversion is 15%, Year-1 ARR is €120–240k and the business is a consultancy with software, not a scalable SaaS. This conversion rate is the single most important number to measure in the first 90 days.

Phase 1 (Days 0–30) — Network activation + falsification interviews.

• Action: Martin runs 15 calls with his LPB-network Heads of Subscription. Each call serves a dual purpose — falsify the autonomy-tolerance hypothesis (Channel C Test A1) AND prospect for the first 3 design-partner LOIs.
• Channel: warm intros only. Zero cold outreach, zero paid acquisition, zero content marketing.
• DoD: 15 calls completed; 3 signed shadow-mode LOIs at €0 for 30 days; documented decomposition of recoverable vs. unrecoverable churn from at least 5 of the 15 accounts (Channel C Test A2).

Phase 2 (Days 31–60) — Shadow-mode pilots + recoverable-MRR measurement.

• Action: deploy the dunning-only agent in shadow mode (recommends, does not execute) on the 3 LOI accounts. Measure the delta between agent recommendations and customer's actual retry logic. Convert at least 1 of 3 to write-access by day 60.
• Channel: direct delivery to design partners; no broader GTM motion yet.
• DoD: 3 shadow-mode deployments live; documented recovered-MRR delta on at least 2 accounts; 1 write-access conversion with a signed €2k/mo contract.

Phase 3 (Days 61–90) — First public case study + Chargebee/Recurly partner motion.

• Action: publish the first named case study with the write-access customer ("How [Brand X] recovered €Y in failed payments in 30 days with Subscript"). Simultaneously begin partner conversations with Chargebee and Recurly partner ecosystem teams; propose a co-hosted "EU D2C dunning benchmarks" webinar for Q+1.
• Channel: case study distributed via LinkedIn (Martin's network), operator Slack communities (D2C founders, RevOps), and the partner ecosystems.
• DoD: 1 named case study published; webinar scheduled with 50+ registered operators; 5 additional pilot conversations in flight.

Judge's call on GTM (siding with Lenny): do not touch paid acquisition, content SEO, or outbound SDR motion before 10 paying customers exist. Channel A's "PLG dunning wedge" framing is correct directionally but underestimates how much of the early sales motion is founder-network-dependent. The case-study flywheel is the actual GTM moat; everything before the first 3 published case studies is preparation, not scaling.

1. The 110% NRR benchmark is unverified. Sequoia flags it as a sales-cycle risk; Lenny calls it potentially fabricated. Judge sides with Lenny: treat this as a thesis-killer until sourced. Falsification test: in the 15 buyer interviews, ask each operator (a) what their current NRR is, (b) what NRR they believe is achievable in their category, and (c) what fraction of their churn they consider recoverable. If the modal answers are NRR ceiling 102–105% and recoverable-churn-share <30%, the value-prop math needs rebuilding before S3 architecture work begins. (S1 Falsifier Test A1+A2.)

2. Autonomy-tolerance may be a small slice of the SAM. If 70% of buyers will only purchase a recommender tool, the addressable buyer universe is a fraction of the €420M SAM. Mitigation: segment the 15 interviews on a single question — "would you grant write-access to your billing system to a software agent after a 30-day successful shadow pilot?" If <40% say yes, the autonomous positioning is premature and the wedge becomes a recommender tool at lower ACV. (S1 Falsifier Test A1.)

3. Chargebee/Recurly ship competing agentic features within 18 months. Sequoia calls this a 24-month threat; a16z calls it the #1 category-killer. Judge sides with a16z's urgency: 12–18 months, not 24. Mitigation: the moat is named case studies + playbook depth, not technology. Subscript must publish 3+ named case studies with quantified NRR/recovery deltas within 12 months to create switching cost before incumbents ship.

4. Founder builder-claim is unverified. Martin's operator depth is credible; his ability to ship production-grade agentic software is [CLAIMED] and unverified. Mitigation: a 2-week technical spike (S5 pre-work) to validate that the dunning-only wedge can be built and deployed against Chargebee's API by the founder + 1 contractor in <90 days. If this spike fails, the team needs a technical co-founder before S5 commits.

5. Recoverable-churn share may be structurally below 30%. All three channels flag this as the deepest value-prop risk. If churn is mostly product dissatisfaction or financial pressure rather than friction or timing, no intervention agent closes the gap. Falsification: in 5 of the 15 interviews, request access to anonymized churn-reason data and decompose by category. (S1 Falsifier Test A2.)

1. Run 15 buyer interviews with autonomy-tolerance + recoverable-churn questions embedded — Days 0–21. DoD: documented segmentation of (a) autonomy-tolerant vs. recommender-only buyers, (b) reported NRR ceiling per operator, (c) estimated recoverable-churn share per operator. This is the single load-bearing test — every downstream decision (ICP, pricing, SOM, architecture) depends on its output.

2. Source or replace the 110% NRR benchmark — Days 0–14, parallel to (1). DoD: either a verified citation for 110% NRR in EU D2C subscription, OR a re-derived realistic ceiling (e.g., 103% with a 5-point spread to current ~95%) that becomes the new value-prop math. No customer conversation uses "110% NRR" until this is closed.

3. Sign 3 shadow-mode design-partner LOIs at €0 for 30 days — Days 14–45. DoD: 3 signed LOIs from Martin's warm network, with billing-system access confirmed and a recoverable-MRR measurement protocol agreed in writing.

4. Technical feasibility spike on the dunning-only wedge — Days 14–35, parallel to (3). DoD: a working prototype that connects to a Chargebee sandbox, ingests failed-payment events, generates personalized retry + messaging logic, and emits a recommendation log. Validates the founder builder-claim and produces the artifact for shadow-mode deployment in Phase 2.

5. Map Chargebee/Recurly agentic-roadmap signals — Days 30–60. DoD: a documented assessment of incumbent job posts, changelogs, recent acquisitions, and partner-program signals; an explicit 12-month-window thesis with quarterly checkpoints. Feeds S3's competitive-positioning section and S5's architectural decisions on lock-in vs. portability.