StackStats

Local-first analytics is the right architecture for this — subscriber data is exactly the kind of thing writers shouldn't ship to a third-party SaaS. Question on the churn-risk model: is it a heuristic (days-since-open style) or a trained model? We went the local-compute route for the same privacy reason, and the honest tradeoff is you can't retrain on aggregate patterns across users.

The growth sources view is the part I'd use most. Can it attribute new subscribers to a specific post or recommendation, or only at channel level (Substack network vs direct vs social)? Knowing which posts actually turn readers into subscribers would make the best-posting-time suggestion a lot more actionable.

Computing everything locally is a good fit for subscriber data. How does the 180-day growth forecast handle publications with fewer than a few months of history, does it show a confidence range or just a single line?

Computing everything locally is a smart choice for a tool that handles subscriber lists, and a live demo with no sign-up lowers the barrier a lot. Question: how does the 180-day forecast handle growth that comes in spikes, like a viral post or a big recommendation? One unusual week seems like it could skew the trend line.

The superfans view caught my eye. How do you define a reader as upgrade-ready? Is it based on open and click patterns, or on how long they have been subscribed?

Doing the analytics locally with no subscription is a smart call for Substack writers who don't want to hand their list to another service. How is the 180-day forecast calculated? Does it adjust for one-off spikes, like a post going viral, or is it a straight trend line?

Substack has no public API, so how does the one-click sync get the data? A session cookie, or an export? I'd like to know what happens when Substack changes its dashboard, and whether the local copy is encrypted at rest, since it's a full subscriber list on disk.

Substack shows me opens and clicks already. What I can never get out of it is cohorts — of the people who joined in March, how many still open in November. If StackStats does that, you’ve solved the only thing I actually need.

The churn-risk list is the feature I'd act on, but only if I can get it out of the app. Can I export the at-risk readers as a CSV (email + last open date) so I can send them a win-back note from Substack or my own mail tool? Also, for best posting time: is that worked out in each reader's local time zone or in mine? With readers split between the Pacific coast and Europe, those give very different answers.

In growth sources, do referrals from chatgpt.com and perplexity.ai show up as their own source? Older Substack posts get cited in AI answers more than people expect, and when that traffic is lumped into "direct" or "other" it's invisible. Seeing which posts those AI visitors land on, and whether they subscribe at a different rate than social traffic, would be a really useful view.