NotebookBloom: 31 installs, 17 opens, 9 successful users — what the funnel changed
I changed my priorities after watching where people actually got stuck.
My earlier analytics showed 31 extension_installed users, 17 extension_open users, and 9 users with at least one successful action. That changed what I chose to build next.
• The biggest issue was not attracting zero users; it was helping users finish.
• AnkiConnect push succeeded 5 times out of 21 starts in that measurement window.
• Citation import and bulk links also failed when prerequisites were unclear.
These are small-sample observations from the earlier measurement window, not proof of product-market fit and not a claim about 1.1.3 performance.
The next useful metric is not another install. It is one more person completing the action they came for.
I am curious how other builders decide when onboarding guidance is more valuable than another feature.
#BuildInPublic #ProductAnalytics #ChromeExtension #NotebookLM
With AnkiConnect succeeding 5 times from 21 starts, I’d be careful about treating this as an onboarding-copy problem.
A preflight check may do more work than a tutorial: confirm Anki is running, AnkiConnect is installed, permissions are correct and a test handshake succeeds before the user begins the push. Then show guidance only for the exact failed check.
The metric I’d watch is recovery-to-success in the same session. If users understand the error but still cannot recover, the integration setup—not the explanation—is probably the product problem.
With a sample this small, I’d instrument the path from “started” to “successful action” rather than add another top-level funnel number: record the intended action, the first unmet prerequisite, and a stable failure reason for each attempt. That should separate unclear onboarding from genuinely broken imports or push delivery, and gives you a better next experiment than install/open rate alone. I’d compare median time-to-first-success and completion rate with and without the guidance, while treating the current counts as directional rather than a product verdict.