Support Reporting
Part of Support quality tools
Measuring whether quality reviews improve customer outcomes
Trace a QA finding through a support change to a defined customer outcome, without treating review scores as proof of improvement.
To assess whether quality reviews help customers, connect a review finding to a change in support practice, then measure the customer result that change was meant to improve. A higher review score alone does not show that customers received better answers.
Describe the expected change
Start with one problem, such as agents omitting an eligibility condition. Record the intended path: reviewers find the omission; guidance changes or agents receive coaching; later answers include the condition; fewer customers need to ask again about the same issue. Observe each step separately.
If the guidance was never changed, do not credit it for a later customer trend.
Choose an outcome close to the problem. Possibilities include repeat contact about the same issue, a correction to an answer, a customer rating or whether a customer had to restate information after handover. Define how the outcome is identified.
A new ticket from the same person is not automatically a repeat of the earlier issue.
Show what each measure counts
| Measure | Count and qualification |
|---|---|
| Review finding | Eligible conversations, reviewed conversations and cases with the named error |
| Action taken | Findings assigned, guidance changed and affected agents reached |
| Later behaviour | Eligible later conversations checked for the same error |
| Customer outcome | Eligible customers or issues, measured results and cases without a result |
Customer ratings need coverage alongside the score. Help Scout's Happiness report shows rating counts and the share of customers who rated a reply; its email rating links can apply to individual replies. Intercom offers conversation ratings, with configuration affecting which conversations receive a request. These results describe respondents, not every customer.
Quality reports also count different units. Zendesk QA distinguishes manually submitted reviews from reviewed conversations. Intercom Monitor reporting counts scorecard evaluations and criterion evaluations. Identify whether an analysis needs conversations, reviews or customer issues before joining data.
Compare periods and other explanations
Set a baseline before the action and a follow-up after agents could use the change. Keep the issue definition, channels, outcome window and rating request method consistent.
Record other changes, such as staffing, a product fault, a policy revision or a new bot route. A simple before-and-after trend may be useful, but those changes can also explain it.
Where practical, compare similar teams or issue groups over the same calendar period, with the approach agreed before looking at results. Differences in case mix or workload can still mislead. Do not delay a necessary customer correction to create a comparison.
State what the evidence supports
If later answers include the condition but repeat contact has not changed, the response process may have improved without a measured customer benefit. If repeat contact falls while answer quality is unchanged, investigate another cause. Report both observations.
Describe a simple trend as an association. A stronger claim that the review program contributed to an outcome needs a suitable comparison and a check of other explanations. Record remaining data gaps and the owner of the next change.
