| The dashboard is the wrong first build. |
| It looks like progress. Charts, counts, a tab you can show someone in a meeting. Then it turns into one more thing you don't open, sitting next to the notifications feed you already don't read. |
| Same flaw in both. They show you everything and decide nothing. |
| The useful build is boring. |
| Ask for the last seven days of engagement, drop anyone outside your ICP, rank what's left by what the person did and how recently they did it. Then ask for five names and one sentence each on why they're worth a message today. |
| You read five lines before your first call and message the ones you agree with. That's the whole workflow. |
| You will have to tell it what counts. |
| Your first version over-ranks likes. Say that a comment beats a reaction, that a repeat profile view from a director beats both, and that anything older than ten days sinks to the bottom. |
| Four of the five builds we wrote up are filters. A filter with no criteria just moves noise around. |
| If you build a second thing, build the cheap one. |
| A Slack alert on target-account signals. Fifteen minutes and a webhook. Leave reactions out of it for the first week or everyone mutes the channel. |
| The test for any of it. |
| The build should end in a decision. If the output is a screen, you built another feed. |
| — steven |