Guides

Scoping a monitor

Describe the product the way you would to a person who might buy it, not the way you would to an investor. The description drives the keyword and subreddit suggestions and is part of every scoring decision, so vagueness there costs you everywhere.

Name the problem, not the category. "Scheduling software" matches everything; "double-booking when two people manage the same calendar" matches the conversations where someone is actually annoyed enough to look.

Choosing a score floor

Start high — around 70 — and lower it only if the feed is empty for several days. Starting low teaches you that the product is noisy; starting high teaches you what a good lead looks like, and lowering it from there is a decision you make with evidence.

The floor on alerts should usually sit above the floor on the feed. A lead worth reading when you sit down is not necessarily worth an interruption.

Tuning from real results

Change one thing at a time and give it a few days. The usual first correction is an exclude term: a word that keeps dragging in a neighbouring topic. The second is dropping a subreddit that produces volume without intent.

Resist adding subreddits to increase volume. More sources with the same terms mostly produces more of the leads you were already dismissing.

Working the feed

Judge on the breakdown, not the total. A 72 built from high purchase intent and low freshness is a different lead from a 72 built from a fresh post with vague intent — the first may still be worth a reply, the second usually is not.

Dismiss generously. Dismissed leads stay searchable, so setting one aside is reversible and keeps the feed honest about what is left to do.