冷郵件外聯測試未獲外部註冊，決定停止新外聯

Today was a checkpoint we set for ourselves two weeks ago: first external signup by Oct 28, or we scale the cold-email experiment down and put the hours into the compounding channels instead.

The number came in: zero. Three new accounts did register this week — all zero-usage, none attributable to anything we operate. By the rule we wrote down before starting, that's a miss.

So, per the pre-registered plan:

• New cold outreach stops today. The segmented test closes at 10 delivered in segment A, 6 in segment B. Across the whole 188-send line: two human replies (one removal request, one intro-call request), two hard bounces. We'll publish the partial results around Nov 6 with the stop reason spelled out — killing your own test early is only honest if you say so publicly, and the pre-registration said results get published either way.
• What we keep: the two compounding lines — MCP directory listings and SEO long-tail pages. 49 pages are now queued for indexing. Google's index dataset hasn't refreshed in five days, so we wait for the effect we can't force.
• One live thread stays open: the fractional CMO who asked for an intro call. A conversation that already exists isn't an experiment.

Two numbers worth keeping from the wreckage: average session engagement went from 12 seconds to 42 across the last three weekly windows, and AI assistants sent us 7 first-touch users in 28 days — more than organic search (3). Both point at the channels we're keeping.

Next scoreboard: Nov 6. Partial results, no flattery.