Before you start
- You need Client Health read access (
read_customer_success). - To change which metrics count, you need access to Scoring settings under Settings > MSP Tools > Scoring.
Check why a company got its score
1
Open the company
Open MSP Tools > Client Success, select the Health tab and click the company.
2
Check Last computed
Read the Last computed date in the page header.
3
Read Dimension Breakdown
For each dimension, compare the score, the applied weight, the evidence coverage and the observed metrics with the rules below.
4
Read Top Negative Drivers
These show which dimensions cost the most points.
5
Fix the source if evidence is missing
Reconnect the integration or fix the company mapping, wait for the sync, then click Refresh Scores. See Refresh Client Health scores.
How the overall score is formed
The default model starts with these dimension weights:
For each enabled, applicable dimension with a score, multiply its base weight by its evidence coverage. Those adjusted weights are then normalized across the observed dimensions to produce the applied weights. The overall score is the weighted average of the dimension scores using those applied weights.
Overall coverage compares the adjusted observed weights with the base weights of enabled, applicable dimensions. A missing dimension contributes no score, but can reduce coverage when it is enabled and applicable. An excluded or inapplicable dimension does not count toward that denominator.
The default model publishes a numeric score only when there are at least three scored dimensions and at least 40% overall coverage. Otherwise the result is Limited data. Available dimension scores and negative drivers may still appear.
For an illustrative configuration with only these three dimensions enabled:
Coverage is 29.5 ÷ 36, approximately 81.94%. The weighted score is (80 × 6.5 + 100 × 13 + 60 × 10) ÷ 29.5, approximately 82.03. This is a worked example, not a screenshot of customer results.
Calculated risk uses the score before the table rounds it to a whole number: Thriving at 80 or above, Healthy at 60 to below 80, At Risk at 40 to below 60, and Critical below 40. A rounded number close to a band edge can sit in the other band. A manual risk flag replaces the effective label without changing the score.
Service Quality
Service Quality compares the last four complete ticket weeks with the preceding four complete weeks. It needs all eight weeks, a ticket source synced within 14 days, and enough observed enabled evidence. The observation threshold is approximately 58.82% of enabled risk weight.
The score is 100 multiplied by one minus the triggered risk weight divided by observed risk weight. Missing observations are omitted rather than counted as healthy. Switch exclusions change which weights participate. High-priority growth and escalation evidence can appear as warnings but do not directly add scoring penalties in this model.
Engagement
Native and imported CSAT responses can contribute; CES campaigns and responses without a mapped company or normalized score do not. If the tenant has no qualifying CSAT response in the last 13 weeks, both CSAT metrics are unavailable and carry no weight. No meeting or recent NPS response means missing evidence for that metric, not a favorable default.
The response rate divides CSAT responses by tickets closed, and both are counted over the same last 13 complete weeks. Recent CSAT still averages the trailing 90 days. Scores move on the next scorecard generation, and scorecards already published are not rewritten. Some responses can come from tickets closed before the window, so read the result as a response indicator, not an exact share of closed tickets that got a response.
Financial
Financial uses profitability measures from the three complete calendar months before the evaluation month, plus invoice evidence. It requires the relevant source sync timestamps to be within 14 days. It evaluates the inputs independently, so an unavailable input need not hide every Financial metric.
The endpoint model can use devices, users or the greater of those counts. Effective hourly rate needs positive revenue and support hours; seat price needs positive revenue and endpoint count. A missing denominator gives no sample, not a zero.
Seat price earns 100 at or above target, then 80/60/40 at 90%/80%/70% of target, otherwise 20. When no target is available, fixed amount bands of 200/150/100/75 earn 100/80/60/40, otherwise 20. Review Financial Reporting settings to understand the configured target and endpoint model; these settings are separate from metric inclusion.
For overdue invoices, no open invoices means no sample. Zero overdue earns 100, up to 10% overdue earns 80, up to 25% earns 60, up to 50% earns 40, and more earns 20. Payment speed earns 100/80/60/40 for averages no greater than 15/30/45/60 days, otherwise 20. No paid invoices means no payment-speed sample.
Sales
Sales gives equal default shares to activity recency and decided win rate. Its sales source sync must be within 14 days.- Activity recency uses the latest qualifying PSA opportunity or Quoter quote activity. Within 30/90/180/365 days earns 100/80/60/40 points; older activity earns 20. No activity means no sample.
- Decided win rate uses won outcomes divided by won plus lost outcomes within the trailing year. At least three decided outcomes are required. Win rates of at least 80%/60%/40%/20% earn 100/80/60/40 points; lower rates earn 20.
Projects
This dimension uses synced PSA project records and a project source synced within 14 days. Projects created only in the Projects workspace are not scored.
Effort needs a positive estimate and an actual-hours value. A project on hold is omitted from active schedule health but may still qualify for effort adherence. Empty qualifying sets produce no sample, rather than a perfect score. Project count and project monetary values do not earn points on their own.
Standards
Standards scores deployed required checks as a percentage of applicable checks:- In a normal category, each required item is one check. Optional items do not count.
- In a Require one of group category, the category is one check if it contains items, regardless of each item’s Required switch. Any deployed option satisfies it.
- Missing assignments count as Not Started.
- A normal check marked N/A or Declined is excluded. A one-of group is excluded only when every option is explicitly N/A or Declined; an unassigned option keeps the group applicable.
- An excluded company or a catalog without qualifying checks is not applicable. If all qualifying checks are excluded for that company, there is no score sample.
Environment
Environment gives 60% to server availability and 40% to patch coverage. The device source sync must be within 14 days. Server classification uses a server type or server operating-system name. A server is observed when it has a status or last-seen timestamp. It counts as offline when its status explicitly says offline or its last-seen timestamp is over seven days old. A missing last-seen timestamp alone does not make an otherwise observed server offline. The score is the percentage of observed servers not counted offline; no observed servers means no sample. Patch evidence currently comes from mapped, non-ignored Datto RMM, ConnectWise Asio and Syncro records. Recognized fully-patched/current states count as patched. Errors, required reboots, pending or missing patches count as unpatched. Unknown states are omitted from both numerator and denominator. The score is patched records divided by records with a recognized patched/unpatched state. It is not a percentage of every device shown in the portal. Device count is context and earns no points. User activity does not feed Environment.Operational
For Planner and compliance, zero overdue earns 100; overdue shares up to 10%/25%/50% earn 80/60/40; higher shares earn 20. The percentage shown as the observed metric is the overdue rate, not the resulting health points: a smaller observed percentage can mean a better score.
Completed and Cancelled Planner items with due dates remain in the denominator but are not overdue. Archived Planner items are excluded. Compliance completion uses its completion timestamp. No qualifying due-date records or observed goal progress means no sample for that metric.
Read changes over time
Previously published scorecards keep their results. The chart separates model-revision segments, but switching metric inclusion within the same model does not guarantee a new visible segment. A score change can come from new evidence, changed metric inclusion or a different data sample, as well as changed performance. Check the settings and dimension coverage before you credit a change to your work. The portfolio stale warning uses a missing data-through timestamp or one older than 36 hours. The 14-day source requirements above are separate metric eligibility gates, so a stale warning does not mean every dimension is unavailable. Refresh Client Health explains source sync and publication.Troubleshooting
Frequently asked questions
Where can I see why a company received its score?
Where can I see why a company received its score?
Open the company from the Health tab of MSP Tools > Client Success. Review Dimension Breakdown, coverage, applied weights, Top Negative Drivers and Last computed.
Is the overall score a simple average of the dimensions?
Is the overall score a simple average of the dimensions?
No. Each enabled dimension’s base weight is multiplied by its evidence coverage, then normalized into the applied weights.
What does Limited data mean?
What does Limited data mean?
The default model needs at least three scored dimensions and 40% overall coverage. Limited data means that bar is not met. It is not a score of zero.
Why can a displayed 80 still be Healthy?
Why can a displayed 80 still be Healthy?
The display rounds the score. The risk level uses the score before rounding, so 79.6 shows as 80 and stays Healthy.
Does a manual risk flag change the score?
Does a manual risk flag change the score?
No. It replaces the displayed risk label only.
How much history does Service Quality need?
How much history does Service Quality need?
Eight complete weeks, split into two four-week periods, and a ticket source synced within 14 days.
Do high-priority growth warnings subtract points?
Do high-priority growth warnings subtract points?
No. They show as warnings next to the five scored metrics.
Which surveys count toward CSAT?
Which surveys count toward CSAT?
Mapped native or imported CSAT responses that have a normalized score. CES campaigns and unmapped responses are left out.
Which months feed the Financial metrics?
Which months feed the Financial metrics?
The three complete calendar months before the evaluation month. It is not a rolling 90 days.
Which endpoint count do the Financial metrics use?
Which endpoint count do the Financial metrics use?
The endpoint model in Financial Reporting settings uses devices, users or the greater of the two.
Will Quoter fill in when the PSA win-rate sample is too small?
Will Quoter fill in when the PSA win-rate sample is too small?
No. When a company has PSA opportunity records, PSA outcomes win even with fewer than three decided outcomes.
Does a project on hold leave every Projects metric?
Does a project on hold leave every Projects metric?
No. It leaves active schedule health but can still count for effort adherence.
Does a missing Standards assignment drop out of the score?
Does a missing Standards assignment drop out of the score?
No. It counts as Not Started.
Do unknown patch states count as unpatched?
Do unknown patch states count as unpatched?
No. They are left out. Errors, pending patches and required reboots count as unpatched.
Does a missing last-seen time make a server offline?
Does a missing last-seen time make a server offline?
No. A server counts as offline when its status says offline or its last-seen time is older than seven days.
Why does a lower Planner overdue percentage give a better score?
Why does a lower Planner overdue percentage give a better score?
The percentage shown is the overdue rate. Fewer overdue items earn more points.
Does changing a metric switch break the trend line?
Does changing a metric switch break the trend line?
Not always. Breaks follow scoring model changes, not every inclusion change.
Does the 36-hour stale warning mean the data failed scoring?
Does the 36-hour stale warning mean the data failed scoring?
No. The warning and the 14-day source checks are separate.