MLI — Member Lifecycle Intelligence
Track every member's lifecycle state, surface members at risk, and steer your community towards retention. The MLI dashboard turns raw activity into a single health number per member.
Before you start
A few quick checks before opening the MLI dashboard.
- You have the mli.view capability on this server
- The MLI module is enabled in Modules
- At least one MLI compute cycle has run (the bot processes data on an interval)
- A channel exists where MLI can post alerts and digests, if you plan to enable them
What you'll see
The MLI dashboard is split into four tabs across the top.
Overview tab
Landing screen — the snapshot you check first thing in the morning.
Health ring + KPIs
- The big ring shows the average health score across all tracked members (0–100). Green above 60, amber 30–60, red below 30.
- Four KPI cards: Total members, Critical, New members (last 7 days) and Average risk.
- Tones on the KPIs change colour based on thresholds — Critical turns red the moment any member is in the critical state.
Score bars
State distribution
Top risks
Recent state changes
Members tab
The full member table. The most useful view for day-to-day moderation work.
Filters and sort
All states, Critical, or any specific lifecycle state.
Priority score, Health, Risk, Engagement or Last active.
Columns
- User — avatar plus display name. Hover the name to see the user id.
- State — colour-coded chip showing the lifecycle state.
- Health / Engagement / Risk — score cells (0–100). Risk is inverted.
- Trend — small arrow icon for whether health is rising, falling or flat.
- Active risks — danger chip with the count of active risk families.
- Last active — formatted date.
Pagination
Risks tab
Risk-first view. Useful when you want to triage problems instead of people.
Severity tiers
Risk family distribution
Active risk details table
- User — who is affected.
- Risk family — coloured chip identifying the category.
- Severity — coloured bar plus percentage.
- Confidence — how sure the model is.
- Explanation — short, plain-English description, truncated to two lines (full text on hover).
- Detected — date of first detection.
Settings tab
The longest tab. Save with the button in the top right — many fields are validated as you type.
Features
- Enable MLI — master switch for the entire module.
- Risk alerts — post alerts when a member crosses the risk threshold.
- Interventions — let MLI suggest or perform automated nudges.
- Social graph — track how members interact with each other for the Social score.
- Cohort tracking — bucket members into cohorts and track them over time.
- Digest reports — periodic summaries posted to the alert channel.
- Quiet hours — suppress alerts during a configurable window.
Channels
Where MLI posts member-risk alerts and digests. Text or announcement channels accepted.
Verbose log of every state change and computation cycle.
Sensitive PII-bearing alerts go here, separate from the public alert feed.
Alert settings
EMBED, CARD or TEXT. Embed is most readable; Text is plain and minimal.
Hard ceiling that prevents alert flooding. 0 disables alerts entirely.
Example: 20
Only shown when Digest reports is on. 10–1440.
Example: 1440
HH:MM strings. Only shown when Quiet hours is on.
Example: 22:00 → 08:00
Thresholds
| Setting | Type | Default | Description |
|---|---|---|---|
| Health alert threshold | Number (0–100) | 30 | Alert when health drops below this value. |
| Risk alert threshold | Percent (0–100) | 60% | Alert when any risk severity exceeds this value. |
| Inactivity days | Days (1–3650) | 14 | After this many idle days, members start moving towards Inactive. |
| Dormancy days | Days (1–3650) | 30 | After this many days they are considered Dormant. Must be ≥ Inactivity. |
| Lost days | Days (1–3650) | 90 | After this many days they are considered Lost. Must be ≥ Dormancy. |
Computation
How often MLI re-runs the lifecycle calculation. 10–1440.
Example: 60
How long raw signal data is kept. 7–365.
Example: 90
Pick at least one of the rolling-window options (1h, 24h, 7d, 30d, 90d). At most six. The chip cluster shows the count.
Definitions
The vocabulary MLI uses across every tab.
- Health
- A 0–100 composite score combining engagement, social and inverse-risk signals.
- Engagement
- How active a member is — messages, voice minutes, reactions and command usage.
- Social
- How connected the member is to other members based on co-conversation, mentions and replies.
- Risk
- Likelihood of negative outcomes — disengagement, leaving the server, moderation events.
- Risk family
- A category of risk such as engagement-decay, social-isolation or moderation-flag.
- Lifecycle state
- NEW, ACTIVE, INACTIVE, DORMANT, LOST or CRITICAL — the bucket a member currently sits in.
Common tasks
Step-by-step recipes for the work most operators do most often.
Turn MLI on with sensible defaults
Get the dashboard populated with data as quickly as possible.
Open DashboardInsightsMLISettings.
- Toggle Enable MLI on.
- Pick an Alert channel in the Channels card.
- Leave the thresholds at defaults (14 / 30 / 90 days). Click Save.
- Wait one compute cycle (default 60 minutes) and refresh the dashboard — the health ring and member table populate.
Triage a member who just turned Critical
The most common reactive workflow.
- Open the Members tab.
- Set Filter by state to Critical and Sort by to Priority.
- Open the top member's row — note their lifecycle state, last active date and risk count.
- Switch to the Risks tab and find the same member. Read the explanation column for context.
- Reach out, mute their alerts (if appropriate) or escalate via your staff channel.
Quiet alerts overnight
Use Quiet hours to keep your staff Slack quiet outside business hours.
- Open Settings → Features and toggle Quiet hours on.
- In the Alert settings card set Quiet hours start to 22:00 and end to 08:00.
- Click Save. Alerts inside the window are queued and posted at the next available time outside the window.