Product Maturity, MVP Cut Line & Pricing

Status: Living · Date: 2026-09-08 · Scope: where Retrieva is on the product lifecycle, what the MVP is, and how it's priced. Companion: DORA ICT-TPRM Domain Model.

Product type

B2B SaaS — regulated compliance — open-core with a self-host enterprise tier. Retrieva sells to regulated EU financial entities (banks, insurers) solving a legally mandatory problem (DORA, in force Jan 2025). This is high-ACV / low-volume / long-cycle B2B with security & procurement gatekeepers — not B2C volume, not a marketplace. The self-host option (RTV-44) makes it specifically deploy-anywhere open-core B2B, which suits a buyer that often refuses multi-tenant SaaS.

Where we are (honest assessment)

Lifecycle stage	Retrieva status
1. Idea / Discovery	Done — problem is regulator-validated (ESA 2024 dry run: 6.5% passed all Register checks; DORA mandatory). Strongest asset.
2. PoC	Substantially done — RAG + multimodal ingestion + LLMOps prove "AI over compliance documents" is feasible.
3. Prototype	Partial — landing/brand + working chat UI exist; the DORA workflows (triggers, Register, concentration graph) aren't prototyped as UX yet.
4. MVP	Not reached — the assess-an-arrangement engine is RTV-28→50, all still Todo.

The mismatch to correct: infrastructure is at ~stage-8 maturity (prod-grade GitOps/Kargo/ observability, live on retrieva.online), but the product core is pre-MVP. Classic strong-engineer trap — over-invested in platform, under-invested in the domain. The RTV backlog is the road to MVP and is correctly sequenced; the fix is to build the domain core next.

The maturity ladder mapped to the backlog

Stage	Scope	RTV stories
MVP — "will a compliance officer assess an arrangement and trust the output?"	Postgres+Drizzle foundation → arrangement graph + evidence + Register → control library → assessment engine (cited, human-approved). One trigger only (🟡 existing provider).	RTV-45→48, 36, 37, 38, 29/39, 30, 41, 42, 43
MLP — delight/retention	WHOOP redesign (done) + the concentration graph as the "wow"	RTV-50 → RTV-28, partial RTV-34
V1.0 GA	Full lifecycle (🟢 new + 🔴 change), AI-provider module, billing/support/docs	RTV-31, 32, 33
V2.0 / Enterprise	Group governance (multi-entity), self-host	RTV-35, RTV-44

The MVP cut line is the decision that matters most now. Everything below the MVP row is the smallest build that validates demand — resist pulling V1/V2 features (change engine, group, self-host) into it. MVP-scope issues carry the mvp label on board #2.

What the MVP deliberately excludes (and why)

• 🟢 new-provider + 🔴 change triggers (RTV-31/32) — MVP proves the engine on the existing estate (🟡) first; the other triggers reuse the same engine and come at V1.0.
• AI-provider module (RTV-33) — a differentiator, not needed to validate the core loop.
• Graph-intake automation (RTV-34) — MVP can seed the graph manually; automate once the loop is proven.
• Group governance / self-host (RTV-35/44) — enterprise expansion (V2), schema-ready but deferred.

Pricing / deal size

High-ACV per-legal-entity annual subscription, tiered by # arrangements, # entities, critical-function coverage, and self-host — not freemium/volume:

Tier	Target
Paid pilot (PoV)	€5k–€15k for a 1–3 month proof against their real vendors
Mid-market	~€12k–€30k / entity / year
Enterprise / group (multi-entity + self-host)	€60k–€150k+ / year

Near-term target is not MRR — it's the PoV row: 1–3 pilot accounts. As a solo project, aim at MVP + 5–20 design partners / 1–3 paid pilots, not user counts. Benchmarks (B2B SaaS): MVP = 5–20 design partners, $500–$5k MRR; PoV = 1–3 pilot enterprise accounts, $0–$5k paid pilots.

RNCP reality check

Retrieva is a certification project first. For the diploma (BC02 concevoir/développer, BC04 optimiser), the stage that scores is a demonstrable MVP + defensible architecture + evidence — not MRR. Ship a working, defensible MVP of the assessment engine; that outweighs any revenue number for the defense. The YC/commercial angle is real but secondary to shipping the MVP.