Next steps
What the Next steps panel on the dashboard Overview ranks, the seven signals behind it, when it refreshes, who sees it, and how company MRR, renewal dates, and blocked deals move a post up.
Dashboard -> Overview, under the metric cards, shows Next steps: the five feedback posts with the most business impact, the revenue and accounts behind each one, and a short written recommendation on what to tackle first.
It is a daily snapshot, not a live ranking. Nothing is applied for you - no status change, no merge, no post edit happens because of this panel.
What the panel shows
A summary card with one paragraph across the whole top five, then one row per post:
| Column | Meaning |
|---|---|
| # | Rank, 1-5 |
| Post | The post title. Clicking the row opens that post |
| ARR | Yearly revenue represented by the companies that voted on it (MRR x 12) |
| Votes | Total votes on the post |
| Top account | The voting company with the highest MRR |
- A bolt before a title marks a post with at least one voting company flagged as blocking an active deal.
- The line under a title (starting with ↳) is the "why" sentence for that post.
- "updated ..." in the header is when the snapshot was computed, not when you opened the page.
Every member with dashboard access sees the same panel. End users never see it on the portal.
When it refreshes
A background job runs once a day at 03:00 UTC. Each run recomputes the entire workspace and replaces the previous snapshot, so there is exactly one snapshot at any time - no history.
A new workspace, or one whose plan does not include the feature, keeps the empty state: "Your first recommendation will appear tomorrow." There is no "recompute now" button.
What gets ranked
A post is eligible only when all of these hold:
- It is not a draft, not spam, and not merged into another post.
- Its status does not map to the Done public bucket. Shipped work cannot resurface as a recommendation.
The five highest scores make the panel. Ties break on vote count, then on the newest post.
What the score is made of
Ranking is a weighted blend of seven signals. All of them are computed in SQL from your own data; the AI only writes the sentences afterwards.
| Signal | Weight | How it is measured |
|---|---|---|
| ARR | 30% | Sum of MRR x 12 for the companies that voted, archived companies excluded |
| Blocked deals | 15% | Voting companies flagged as blocking an active deal, capped at 3 |
| Votes | 15% | Total votes on the post |
| Distinct companies | 10% | Number of different companies among the voters |
| Recency | 10% | Days since the last vote (or the publish date), decaying on a 30-day half-life |
| Sentiment | 10% | Negative 1.0, neutral 0.4, positive 0.2; a post with no sentiment counts as neutral |
| Renewal soon | 10% | Any voting company renews within the next 60 days |
ARR, votes, and distinct companies are each normalised against the strongest post in the same workspace, so ranks compare posts inside one workspace. They are not comparable across workspaces, and a workspace with no revenue data still ranks by votes, recency, sentiment, and deals.
Where the business signals come from
ARR, renewals, and blocked deals come from Dashboard -> Audience -> Companies. Open a company and use Edit:
- MRR (USD) - monthly revenue. The panel multiplies it by 12 for ARR.
- Renewal date - any date within the next 60 days counts as "renewal soon".
- Blocking an active deal - boosts every post that company voted on.
Three constraints worth knowing:
- Revenue only counts for votes cast by end users linked to a company. The company is recorded on each vote when it is cast, so linking a company later does not backfill the votes that customer already made.
- Archived companies are excluded from scoring. Restore one to bring its weight back.
- Fields you edit by hand are marked as manually set, so a JWT SSO login or an import does not overwrite them later. See JWT SSO for the
companyclaim shape.
The AI text
Two kinds of LLM output ride on top of the ranking:
- One "why" sentence per post in the top five (220 characters max), citing the numbers.
- One 2-3 sentence recommendation across the five, shown in the summary card.
Each call counts against your workspace's AI usage - up to six operations per workspace per day. If the model fails, the ranking still renders: a missing sentence leaves the row without a "why" line, and a missing recommendation hides the summary card. The numbers never depend on the model.
Plan availability
Next steps is an advanced AI feature: it ships on Startup and Scale, not on Growth. On Growth the panel stays in its empty state. If a workspace moves off Advanced AI, the last snapshot stays visible but stops updating.
Limits
- Latest snapshot only - no history of previous rankings.
- One currency - MRR is treated as USD.
- No manual override - you cannot pin or exclude a post; fix the data instead (votes, company MRR, status).
- Dashboard-only - there is no MCP tool for it.
- It does not merge, change status, or edit post content.
Related: AI Insights for sentiment trends, and Triage for the daily queue.
AI Insights & Sentiment
How Feedjolt classifies each post as positive, neutral, or negative and charts the sentiment trend. Covers when it runs, visibility, plans, and limits.
Inbox
The Feedjolt Inbox is your review queue for feedback drafts captured from Slack and email. Triage, edit, approve or dismiss external intake before it becomes a post.