The Product Lab is a governed weekly optimisation engine for ecommerce stores. It reuses five original DSA methods, compiles at most three evidence-backed proposals, asks for the operator's decision, and measures approved work after somebody else executes it. Store #1 is NeckReset.
Verified on the VPS on 16 October 2026: the lab-cos account, Product Lab Chief, five DSA-method wrappers, read-only gate, NeckReset profile, weekly cycle, proposal QC, Slack/Notion handoff, watcher, desk and systemd units are installed. Its self-test passes and the first operator-started validation cycle is running. The Lab timers and gate are enabled and active (weekly cycle Monday 06:30 Brussels, with dispatcher, watcher and desk polling) as of 21 October 2026. The business status is live in validation, read-only: validation is not a production launch, and the Lab has no authority to change NeckReset. NeckReset plan and boundaries →
account_spend_limit_reached since 26 Oct (EUR 0.12 in the last 7 days). The creative set has not been tested yet.The sections below describe the Lab as built in October. Where they differ, this block is current. The Lab itself stays read-only.
Queues one weekly cycle per active store, compiles evidence-backed proposals, publishes the reviewed brief and answers the operator in the store's Lab channel. It does not perform an agent's analysis itself.
Diagnose, CRO, Schild and Coach write separate reports. QC is isolated and checks the proposed changes independently before anything is shown to the operator.
The collector reads a store summary, paid-media evidence and named public pages. The staged gate has no store, ad, theme, email or budget write route.
Each QC-passed proposal is recorded in Notion and shown in the store's Lab Slack channel. Reactions and thread replies record the operator's exact decision. Approved does not mean executed.
Every store has its own profile, claim rules and cycle folders. NeckReset is the first profile. The engine contains no hard-coded second store.
The unprivileged runtime, read-only gate, weekly timer, dispatcher, watcher and desk are installed. Their presence proves the build; a completed production acceptance cycle is still required before the Lab is called live.
Capture current store, paid-media and page evidence for one store and one ISO week.
Diagnose, CRO, Schild and Coach each return their own bounded report.
The Chief merges duplicates and prefers cheap, reversible steps that unblock a fair test.
QC filters unsupported proposals. The operator approves, rejects or requests changes in Slack.
A later cycle compares the named before/after metric. Too little data stays too little data.
The NeckReset profile enables only the optimise loop. Product discovery and store building remain off.
The Lab never changes Shopify, Meta, a budget, a theme or an email tool. The operator or a separately governed cockpit capability executes an approved proposal.
A second store or another loop waits for four clean weekly cycles and at least one proposal that was approved, executed and measured.
Installation and self-test are verified. A complete accepted weekly business cycle is still required before the status changes from built to live.