Zendesk organization health score report: predict at-risk accounts from ticket data
Customer success teams spend most of their energy on accounts that are already upset. A Zendesk organization health score report lets you shift that effort earlier. Instead of waiting for a renewal conversation or an angry escalation, you build a composite view of each account’s support experience and watch for the patterns that precede churn.
This guide shows how to build a health score from Zendesk ticket data, which signals belong in the model, and how to act on the results. Pair it with your support metrics dashboard and Zendesk tickets per organization report so account health becomes a regular part of the ops review.
What an organization health score should capture
For the metric definition, see organization health score in the glossary. In practice, a useful health score answers three questions about each account:
- Is this account generating more support work than normal? Rising ticket volume or tickets per organization relative to that account’s own baseline is a leading signal.
- Is the experience getting worse? Climbing resolution time, falling CSAT, or increasing reopen rate all point to friction even when the account is not yet complaining directly.
- Are there escalation or severity signals? Tickets tagged with high priority, urgent, or escalation topics carry more weight than routine questions.
A health score is not a single metric. It is a weighted combination of signals that, together, give you a direction: stable, improving, or deteriorating.
Which signals to include
Start with signals that are already available in your Zendesk instance. Adding complexity later is easy; getting the team to trust and act on the score is the hard part.
Tier 1 — start here:
- Ticket volume trend — Compare tickets created in the last 30 days to the previous 30 days. A significant increase for a single account is the simplest early warning.
- Average resolution time — Longer resolutions for an account may mean harder problems or less effective support routing. Use resolution time at the organization level.
- Reopen rate — Accounts that reopen tickets more often are experiencing incomplete resolutions. See reopen rate.
- CSAT score — If you collect satisfaction ratings, per-organization CSAT is a direct sentiment measure. See CSAT.
Tier 2 — add when ready:
- Escalation count — Tickets routed to senior agents or managers. See escalation rate.
- Repeat contact rate — Customers who come back with a new ticket about the same issue. See repeat contact rate.
- First reply time — Accounts that consistently wait longer for a first response may feel deprioritized. See first response time.
- High-touch frequency — Accounts requiring multiple agent touches per ticket. See touches per ticket.
How to build the report in Zendesk
Zendesk Explore does not have a built-in health score widget, so the report is assembled from component metrics grouped by organization.
- Start with the tickets dataset and filter to a rolling window (last 60 or 90 days gives enough trend data).
- Group by organization so every metric rolls up to the account level.
- Add columns for each signal: ticket count, average resolution time, reopen count or rate, average CSAT.
- Create a calculated metric for volume trend by comparing the current period to the prior period. A simple formula is
(current_period_tickets - prior_period_tickets) / prior_period_tickets. - Score each signal. The simplest approach is a 1-3 scale per signal: 1 = healthy range, 2 = warning, 3 = at-risk. Define thresholds based on your own data. For example, if median resolution time across all orgs is 8 hours, an org consistently above 16 hours scores a 3.
- Sum or weight the scores to produce a composite. Equal weighting is fine to start. Adjust later when you see which signals actually predict trouble.
If you want a single dashboard view, create a table visualization sorted by composite score descending. The accounts at the top of the list are your priority.
For the organization-level ticket view that feeds into this, see Zendesk tickets per organization report.
How to read the report
A health score is a triage tool, not a verdict. Use it to answer:
- Which accounts need proactive outreach this week? Sort by worst score and review the top 5-10.
- Is the at-risk list growing? If the number of accounts scoring “at-risk” is climbing week over week, you may have a systemic quality or capacity problem, not just individual account issues.
- What is driving the score for a specific account? Drill into the component signals. An account with high volume but good CSAT is a different conversation than one with rising reopens and falling satisfaction.
Do not treat the score as a precise prediction. Treat it as a prioritization layer. The goal is to get your team talking to the right accounts before those accounts start talking to your competitors.
Connecting health score to action
The report only creates value if someone acts on it. Build a lightweight workflow:
- Weekly review: Include the top at-risk accounts in your weekly support ops review. Assign an owner for outreach on each flagged account.
- Cross-functional handoff: Share the at-risk list with customer success or account management. Support sees the friction first; CS owns the relationship.
- Feedback loop: When an account churns or escalates, check whether the health score flagged it early. If it did not, adjust the thresholds or add a missing signal. If it did but no one acted, fix the workflow.
- Re-score monthly: Thresholds that made sense six months ago may not match your current ticket mix or team size. Revisit the scoring model regularly.
Common mistakes
- Including too many signals too early. Start with 3-4 signals your team already trusts. A complex model that no one understands will not drive action.
- Using absolute thresholds across all accounts. A 50-seat enterprise account generating 30 tickets a month is different from a 5-seat SMB generating 30. Normalize by account size or baseline where possible.
- Ignoring accounts with no tickets. Zero tickets can mean the account is self-sufficient or it can mean they have stopped trying. Cross-reference with product usage if available.
- Scoring once and forgetting. A static snapshot loses value quickly. Automate the refresh or build it into a recurring review.
FAQ
Can I build a health score entirely in Zendesk Explore? You can build the component metrics and a basic composite in Explore. For more sophisticated weighting or integration with product usage data, teams often export to a spreadsheet or BI tool. TicketBoard can surface these signals without the export step.
How many accounts should I flag as at-risk? Keep the list actionable. If you flag 50 accounts and your CS team can only reach 10, the list loses credibility. Start with the top 5-10 and expand as your outreach capacity grows.
Should I include CSAT if our response rate is low? Low response rates make per-org CSAT noisy. If fewer than 20-30% of tickets get a rating, weight CSAT lower in the composite or use it only as a tiebreaker. Volume trend and reopen rate are more reliable with sparse survey data.
How does this relate to churn prediction? A health score is a lightweight churn signal built from support data. For a broader look at how support metrics connect to churn, see Support Metrics That Predict Customer Churn Before It Happens.