Where to find it | Sidebar → Client Health → Churn Risk |
Who can see it | Everyone except standard-access users |
Can you change data here? | Yes — notes, follow-ups, action tracking and Asana tasks |
How fresh is it? | Scores refresh weekly. Notes and triage are live. |
What it's for
The point is to find out a client is unhappy weeks before renewal, not at renewal.
Signal pulls together the signals that are otherwise scattered — contract timelines, invoice behavior, call sentiment, weekly check-ins, performance pacing, sales activity — into one score per client, recalculated every week. Then it answers three questions:
Where is revenue at risk right now? A ranked list, weighted by dollars.
Why is this client at risk? The score broken into parts, a list of specific issues, and a written summary.
Has anyone done anything? Recommended actions with accountability chips tied to the assigned Account Director and Group Director.
That third one is what makes this a workflow rather than a report. You triage actions, write notes, create tasks, and hand off drafts — all from the client row.
What you'll see
The overview table — one row per active client, sorted by gross profit with the higher risk bands pre-filtered:
Churn Risk — the 0–1 score, as a bar
Client Profile / Sentiment / Performance Risk — the three components
MGP, Current MRR, Last Month VABO — from finance
Revenue — the active contract retainer, from Nova
Weeks to End and the human renewal forecast
Status — Acknowledged, Reviewed, Needs Review or Unassigned
Summary cards — Critical and High counts, plus high-risk counts by business tier.
Distribution charts — a risk histogram, a sentiment breakdown, and at-risk clients by tier.
A Churned / Contract Ended table at the bottom, dimmed, for reference.
Expand any row for the client detail panel:
Tab | What's in it |
AI Summary | A written account of the client's state, refreshed weekly |
Top Issues | Specific problems, derived by rule rather than by AI |
Recommended Actions | What to do, with Done / Dismiss / Undo and threaded notes |
Brand Intel | External context — earnings, leadership changes, talking points |
Iris Draft | A narrative to hand off to Iris, with the three-week trajectory |
Sentiment Signals | Budget signal, satisfaction, relationship health, mood, check-in sentiment, NPS |
Performance Signals | Revenue trend, spend efficiency, goal pacing, invoice status |
POC Contacts | Per-contact sentiment from call analysis over recent weeks |
Salesforce | Open opportunities, close rates, recent wins and losses |
Risk Score History | The weekly score plotted, with the threshold bands marked |
Asana Tasks | Live task status, and a button to create one |
Client notes | Your own journal on the account |
Plus a Churn Log tab for clients we've already lost — covered in its own guide.
How to use it
Monday triage (Group Director). Open the page. It's already sorted by gross profit with the risk filter applied, so the most valuable at-risk clients are on top. Scan the Status column for Needs Review on your accounts. Expand each: read the summary and top issues, then mark each recommended action Done or Dismissed. The chip flips to Acknowledged — which is this page's definition of "this risk has an owner and a response."
Pre-call prep (Account Director). Search the client, expand. Read the AI Summary for the state of the account, check Brand Intel for external context, open POC Contacts to see which specific stakeholder is cooling, and skim Salesforce for open opportunities. Generate the Iris Draft if you want a starting point for client-facing follow-up.
Renewal-window sweep. Filter to contracts ending within 60 days, exclude month-to-month, sort by gross profit. Then compare the model's score against the human renewal forecast. Where they disagree — model says Critical, human says 90% likely to renew — that's your meeting agenda. Export the filtered set to CSV.
Turn risk into tasks. From a recommended action, create an Asana task pre-filled with the client context. Add a manual follow-up for anything the rules missed, and leave notes on the action so the next reviewer knows what was tried.
Did the intervention work? Expand the client and open Risk Score History. Did the weekly score bend down after you acted? The action history shows who did what and when.
Sentiment-led save play. Look for clients with high Sentiment Risk but healthy Performance. Open Sentiment Signals to see which component is driving it, then use POC Contacts to find the specific person whose sentiment dropped, and plan the outreach.
What the scores mean
Churn Risk — 0 to 1. Critical from 0.6 · High 0.5–0.6 · Medium 0.3–0.5 · Low below 0.3. By default the table hides Low, so use the filter if you want the full book.
Client Profile Risk — structural risk from the client's shape and history.
Sentiment Risk — how the client feels. Built from call analysis and weekly check-ins.
Performance Risk — how delivery is going.
The Status chip:
Chip | Meaning |
Acknowledged | The assigned AD or GD has triaged every recommended action |
Reviewed | Someone else did |
Needs Review | A priority client whose actions are untouched |
Unassigned | A priority client with no identifiable AD or GD |
A priority client is one with a score of 0.5 or above that's also in the top 25 by score × monthly gross profit. That's why most clients never show a Needs Review chip — it's deliberately focused on the dollar-weighted top of the book.
What you can change
Everything you do here stays in Signal or goes to Asana. Nothing writes back to the model, finance, Nova, Salesforce or Gong.
Triage a recommended action — Done / Dismiss / Undo. This drives the Status chips everyone else sees, so it's the one action with visible consequences beyond your own screen.
Manual follow-ups — for anything the rules missed, with a severity.
Notes — threaded on a specific action, or a general journal on the client.
Asana tasks — creates a real task, with status syncing back.
Regenerate the AI Summary, Iris Draft or Brand Intel — Brand Intel runs a live web search and costs about 50 cents a time, shown on screen, with a daily limit per person.
Column choices — saved to your browser only, not synced across devices.
You cannot change the churn score, the renewal forecast, team assignments, the client tier, or any finance figure. Those all come from upstream.
Good to know
Two revenue numbers, both correct. MGP, MRR and VABO come from finance. The Revenue column is the contract retainer from Nova. They diverge legitimately, and MGP is $0 for new contracts until Finance posts.
The model and the human forecast disagree on purpose. The score is the model's weekly prediction; the renewal forecast is what a person typed in from a separate tracking process. Disagreement is the signal, not a bug — those accounts are the ones worth discussing.
Thresholds are current values, not policy. The bands and the priority-client rule can be retuned.
Freshness is mixed within one page. Scores are fetched fresh; supporting data is cached for a few minutes. Small transient inconsistencies are harmless at a weekly model cadence.
The model itself is built outside Signal by the AI/ML team. Signal displays what it produces without recalculating anything.
AI text is a draft to verify. The summaries, Brand Intel and Iris Draft are all generated. The Iris Draft includes grounding markers checked against real data, with warnings when a claim can't be matched — read those warnings.
"gsheets" as a sentiment source is a legacy label. It means the input came from Client Pulse, which is now entered in Signal. "gong" means call analysis; both can appear together. It's a naming lag, not a data problem.
The POC "decision maker" flag is never set. It's awaiting an upstream data field. The rest of the POC data is real.
Recommended actions may be frozen weekly or regenerated live, depending on a deployment setting. If the list is stable all week with "week of" labels, it's the weekly snapshot; if it shifts as data changes, it's live.
Common questions
Why doesn't the revenue match finance? MGP, MRR and VABO should match. The "Revenue" column is the contract retainer from Nova — a different system. And MGP is $0 for new contracts until Finance posts.
Who can see this page and how do I get access? Anyone above standard access. See Who can see what.
Where does the churn score come from and how often does it change? The AI/ML team's model, recalculated weekly outside Signal. Nothing on this page recomputes or overrides it. New scores appear only when the weekly run lands.
What do Critical / High / Medium / Low mean? Bands on the 0–1 score: 0.6+, 0.5–0.6, 0.3–0.5, below 0.3. Low is hidden by default.
What makes the Status say "Needs Review", and how do I clear it? It appears on priority clients — score 0.5 or higher and in the top 25 by score × monthly gross profit — whose actions haven't been triaged. It clears when every recommended action is marked Done or Dismissed. If you're the assigned owner it becomes Acknowledged; if someone else does it, Reviewed.
The model says Critical but the forecast says 90% renewal. Which is right? Both are legitimate opinions from independent sources. Treat the disagreement as the thing to investigate.
Can I change a score, the Account Director, or the renewal status here? No. Scores come from the model, the forecast is entered upstream, finance figures come from finance, and client and team changes belong in Nova.
Why does the sentiment source say "gsheets"? Legacy naming for Client Pulse data. See "Good to know".
Why did the recommended actions change (or not change) mid-week? Depends on whether the weekly snapshot is switched on in this environment. See "Good to know".
Is the AI Summary written by a person? No — all three AI tabs are generated. Treat them as drafts to verify.
Related
Portfolio Review — the same clients, one screen, for the weekly sweep
Churn Log — the clients we've already lost
Service Expansion — the growth side. Note its Performance tile uses an inverted scale, so the numbers won't match by design
POC Management — the same contact-level data, in depth
Retention Forecast — what account teams expect to lose
Submit Pulse Data — your check-ins feed this model's sentiment
Glossary — churn score, priority tiers, active book