How Signals works - Zevari Docs
Signals finds reasons to start a relevant conversation. Choose what to watch, let it collect, then review the posts or people it found.
1. Choose what to watch
Keyword posts: Searches LinkedIn for your keywords and saves posts that meet your filters. For example, ‘looking for a CRM’ can find posts asking for a product recommendation. Your post engagement: Checks your recent LinkedIn posts and collects the people who comment or react. Review their profiles to decide who is worth following up with. Each source belongs to the selected workspace and LinkedIn account.
2. Let Signals collect
Enable a source to let it collect automatically. Zevari manages timing. Keyword searches depend on available search credits; engagement checks need recent posts to read. Keyword Run now queues a check when available; results may take time. Pause stops future collection while keeping saved results and settings.
3. Review the post or person
Collected evidence shows the newest saved posts and interactions first. Read the original content, then consider Signals' interpretation. A keyword post can be saved before its author has a linked lead profile. Engagement can link to a saved person. Opening evidence or a profile sends no message. Decide whether the context is relevant before following up.
Choose useful keywords and understand limits
Use a phrase someone would write in a LinkedIn post, such as ‘looking for a CRM’. Negative keywords exclude unwanted topics, such as ‘recruitment agency’. Each active keyword uses one slot. Slots are shared across workspaces using the same LinkedIn account. Pausing a keyword stops it being searched but keeps the workspace's allocated slots. Search credits are a rolling 24-hour allowance; they return as older searches leave that window.
Signal score, confidence and person fit
A 1 to 5 assessment of how useful and timely the post is as a reason to follow up: 1 is noise, 3 is worth monitoring, and 5 calls for prompt attention. Person fit is assessed separately. How confident the AI is in its interpretation of this post. It is an assessment, not the person's likelihood of replying or buying. ICP fit means how well the person matches your ideal customer profile. A strong signal can come from a person who is not a good customer fit. Engagement records may have no signal assessment. Raising keyword score or confidence thresholds saves fewer posts; it applies to future checks.
What Needs review and Dismiss mean
Opening evidence keeps its review status unchanged. Dismiss moves it out of Needs review into Dismissed; the source and its collection settings stay unchanged. All includes evidence in every review state. Dismissed is where you can find items set aside. If a profile is available, open it to understand the person; if no profile is linked, you can still read the original evidence.
Understand Activity and missing results
Activity shows checks and their outcomes. In progress means results are unfinished. Success means the check completed, even if it saved nothing. For keyword checks, posts reviewed is how many the AI assessed; signals saved is the number that passed the filters. The AI review status describes an automated assessment of the post. No matches means that check saved no signal; inspect your keywords and filters before changing them. Partial or failed results may be incomplete: open the check for the reason and available next step. Engagement can record several comments or reactions from one person, so interaction and people counts can differ.
Understand saved changes and errors
Review changes shows what will change before you save. Back to edit keeps your draft. New settings apply to future checks; checks already started keep their settings. If only some changes save, the editor identifies what saved and what still needs saving. If keywords changed elsewhere, refresh the saved keywords, compare them with your retained draft, then review again. A disconnected account needs reconnection. Exhausted search credits need time to become available again.