What the first scan reads, and how far back
The first run of any Embers agent mines existing public history rather than starting from today. Later runs collect only what is new since the last scan.
Adding an agent does not start a stopwatch. The first run mines the public activity that already exists, so a queue can be useful on day one rather than in a fortnight.
The first scan
When you connect a profile or add a watched profile, the first scan walks back through recent public posts instead of anchoring to the moment you signed up. That window is an account level setting, and by default it is 365 days, roughly the last twelve months. It can be configured between 1 and 365 days.
That default is also the ceiling. Even when a low volume profile returns very little inside the requested window, Embers will not pull a profile’s entire history: the fallback is bounded to the same 365 days.
The other two agents behave slightly differently on their first run:
| Agent | First run |
|---|---|
| Post engagement | Walks the account’s post window, 365 days by default |
| Watchlist | Same window as post engagement, so a new watched profile does not start empty |
| Comment engagement | Reads roughly the last month of comments the profile wrote |
| Keyword tracking | Any keyword that has never been scanned forces a month long backfill |
The onboarding preview
Before the first full scan, onboarding runs a small preview so you see real leads during signup. It lists a shortlist of your recent public posts, then collects engagement only on the handful with the most public reactions and comments, and enriches a small number of the people it finds.
The preview deliberately does not record a scan cursor. That means the first full scan still walks the entire window afterwards, and nothing the preview surfaced is lost or double counted.
Every scan after the first
Once a scan has completed, Embers records where it got to. From then on each run asks only for activity newer than that point, plus a small overlap so nothing published near the boundary is missed. Anything already stored is recognised and updated rather than duplicated.
Keyword tracking narrows the same way. Once a keyword has been scanned, the search window tightens to match the gap since the last run, and the number of posts pulled per keyword scales with that gap. A daily cadence pulls a small window, a long gap pulls a wider one.
This is why the numbers in your scan history drop after the first entry. The first run reports a large number of posts and engagements, and healthy later runs report small numbers or sometimes zero.
Why the first run can come back empty
An empty first scan is a real answer, not a failure, and Embers names the reason rather than showing a blank queue:
- No public posts were found in the lookback window.
- Public posts were found, but nobody engaged with them.
- People engaged, but nobody cleared your ICP.
- People matched your ICP but lacked enough public detail to show factually.
The first two are content problems: there was nothing public to read. The third usually means the ICP is too narrow, or the audience genuinely is not your buyer. Each case comes with a recovery action, such as retrying the scan, trying a narrower source, or watching another profile.
Related
Frequently asked questions
- Why did my first scan find far more people than the second one?
- The first scan mines existing public history. Every scan after it only collects what is new since the previous run, so a smaller number is the system working correctly.
- Can I make an agent re-read older activity?
- Not from the app. The window is an account setting. What you can do is trigger a scan manually, which collects anything new since the last run.
Was this article helpful?
What went wrong?
Thank you. We read every answer.
Still stuck?
Send us the article you were following and what happened. We answer from the same inbox that writes these pages.