At 10:00 every Monday, a support lead opens last week’s numbers. Twenty-five minutes later, the team has explained three charts, debated two unusual tickets, and committed to nothing. The calendar says “quality review”; the meeting was a dashboard tour.
A useful weekly quality review has a harder finish line: the group must decide whether a meaningful change needs action, assign that action, and set the evidence they will inspect next week. Everything else can happen asynchronously.
Supportman delivers a weekly support quality report to Slack automatically, so your review meeting starts from data instead of anecdotes.
Give the meeting one job
The meeting exists to turn recent support evidence into a small number of team-level decisions. It is not the place to read the report aloud, recover an individual unhappy customer, redesign the entire QA rubric, or coach an agent in public.
Write this charter in the recurring invitation:
We leave with zero, one, or two decisions about how the support system should change. Every decision has an owner, a due date, and a check for the next review.
Zero is a legitimate outcome when the evidence is weak or performance is inside its expected range. Inventing an initiative every Monday teaches the team to overreact to noise.
The report is pre-reading. If you need help deciding what belongs in it, use the weekly customer support report template; this article covers what people do once the report arrives.
Set the cadence and invite the right people
Use the same 30-minute slot each week, ideally after ratings have had time to arrive and before the team’s planning commitments are fixed. Consistency makes unfinished actions visible; moving the review whenever the week gets busy makes quality work optional.
A practical core group is three to six people:
- Support lead: facilitates and owns operational decisions.
- QA or enablement owner: prepares the evidence and challenges weak diagnoses.
- Frontline representative: explains queue conditions and tests whether an action will work in practice. Rotate this seat.
- Operations or workforce owner: attends when routing, coverage, or tooling is regularly in scope.
Invite product, engineering, or policy owners for a specific agenda item, not as permanent spectators. Send them the question and relevant conversations in advance, then release them when the item ends.
Require pre-work
The facilitator should post the report and proposed discussion items at least one working day before the meeting. Attendees spend no more than 10 minutes reading it and comment with missing context. If nobody has opened the evidence, cancel the review or use the slot for preparation; live data discovery consumes the entire agenda.
Bring no more than two proposed discussion items. Each one needs:
- The metric or quality attribute that moved, with its denominator and comparison period.
- A local threshold or reason the movement deserves attention.
- Two to five conversations that represent the pattern, including counterexamples where available.
- Known changes in volume, staffing, policy, product, routing, or measurement.
- A one-sentence question the group can decide.
Set thresholds before seeing the result. For example, a team might discuss surveyed CSAT only when there are at least 20 ratings and a movement of five percentage points, while still escalating any severe safety, privacy, or conduct failure immediately. Those numbers are an operating rule, not a claim of statistical significance; choose thresholds that fit your volume and risk.
Run a 30-minute agenda
| Time | Purpose | Facilitator prompt |
|---|---|---|
| 0–4 min | Close last week’s actions | Was it done, and what did the check show? |
| 4–7 min | Scan for exceptions | Did the pre-read miss anything urgent or invalidate an agenda item? |
| 7–17 min | Review item one | What changed, why, and what will we do? |
| 17–25 min | Review item two | What changed, why, and what will we do? |
| 25–28 min | Recognize one useful behaviour | What should the team repeat, and why did it work? |
| 28–30 min | Read back decisions | Who owns what by when, and what will we inspect next week? |
Nominate a facilitator and a note-taker before the meeting. The facilitator protects the timebox; the note-taker updates the decision log in the room. Parking-lot calibration debates, individual coaching, and investigations that need new data.
Make decisions with three questions
- What changed? State the movement, denominator, affected segment, and comparison. “Quality was down” is not enough.
- Why do we think it changed? Separate observed evidence from hypotheses. Conversation review can expose a pattern, but a handful of examples rarely proves causation.
- What will we do before the next review? Choose the smallest observable intervention that tests the diagnosis. Name an owner and the evidence that will confirm completion.
If the group cannot answer the second question, assign an investigation rather than pretending to know the fix. If it cannot answer the third, record “no action” with the reason: below threshold, already handled elsewhere, insufficient evidence, or outside the group’s authority.
Worked example: a billing-policy gap
Imagine a 12-person support team reviewing an illustrative week. Its billing-accuracy score fell from 88 to 80 across 146 billing conversations. The QA owner sampled 14 conversations: six used an old macro promising refunds within five business days, and three customers contacted support again after that window passed. Two conversations used the updated wording and set accurate expectations.
The meeting works through the three questions:
- What changed? Billing accuracy fell eight points; the issue appeared in refund conversations, not across the whole queue.
- Why? The evidence supports a stale-macro hypothesis. It does not yet show that the macro caused the entire score movement.
- What will we do? The knowledge owner will replace the refund macro and archive the old shortcut by Wednesday. The QA owner will review 10 refund conversations next week for unsupported timing promises.
The team also recognizes the agent whose updated wording appeared in both counterexamples and links that phrasing in the action. It does not set a vague goal to “be more accurate,” nor does it turn the people attached to the sampled conversations into a public ranking.
Decide who can commit what
Ambiguous authority produces actions that disappear after the call. Agree on decision rights before the first review:
- The support lead can change support workflows, macros, queue rules, and coaching priorities within an agreed scope.
- The meeting can request product or policy work, but the relevant owner accepts the commitment and date.
- The QA owner can recommend rubric changes; material scoring changes go through a separate calibration process so the trend is not rewritten mid-meeting.
- Any attendee can challenge the evidence. The facilitator makes the final in-scope call when consensus does not emerge within the timebox.
When an issue is outside the group’s authority, the decision is who will take a documented request to the accountable owner and when the group expects a response. “Tell product” is not governance.
Write actions that survive the meeting
Goal-setting research supports specific, challenging goals over vague instructions to “do your best,” while emphasizing commitment, complexity, framing, team goals, and feedback. Read the goal-setting evidence review.
Use one shared log and complete each row during the meeting:
| Decision | Owner | Due | Completion evidence | Next-week check |
|---|---|---|---|---|
| Replace and archive the refund-timing macro | Knowledge owner | Wednesday | New macro link and old shortcut disabled | Review 10 refund conversations for unsupported promises |
Start every review with open rows. Mark each one done, not done, or superseded; then record what happened to the target behaviour. Do not demand a metric improvement after one week when volume is low or the outcome naturally lags. First verify that the intervention happened and agents used it.
Keep individual coaching out of the room
A major meta-analysis of 607 feedback-intervention effect sizes found that feedback improved performance on average, yet more than one-third of the interventions reduced it. Feedback can redirect attention away from the task and toward the self. Read the Feedback Intervention Theory paper.
Use the group review for shared practices, capacity, routing, tooling, policy, product issues, and specific recognition. Move an agent’s repeated accuracy gap, conduct concern, or development plan to a private conversation with the full context. A ticket may reveal a system problem without making its owner the agenda item.
AI-generated QA or predicted satisfaction is a lead for investigation, not a verdict. Let attendees challenge the score, inspect the underlying conversation, and record suspected rubric or model errors for calibration.
Change the cadence when weekly is wrong
Weekly works when the team has enough fresh evidence to detect a pattern and enough control to test an action before the next review. Adjust it when those conditions fail:
- Low volume: use a rolling four-week window or meet fortnightly; keep severe-case review immediate.
- High-risk operation: handle safety, privacy, financial, or regulatory failures daily and use the weekly meeting for recurring causes.
- Large support organization: run queue-level reviews, then escalate only cross-team decisions to a monthly operations forum.
- Major launch or incident: use a temporary daily checkpoint with an explicit end date, then return to the normal cadence.
Fix the meeting when it stalls
- Everyone reads charts aloud: enforce pre-reading and begin with exceptions or open actions.
- Every fluctuation becomes a project: define local thresholds and allow a documented no-action decision.
- The loudest anecdote wins: require a denominator, representative conversations, and at least one counterexample.
- The cause is debated for 20 minutes: assign a bounded investigation with an owner and due date.
- Actions recur for weeks: reduce the action limit to one and have the owner explicitly accept it in the room.
- People hide mistakes: remove public agent comparisons and discuss the workflow or behaviour the team can change.
After four meetings, inspect the meeting itself: how many decisions were completed, how many were verified, and which agenda items repeatedly produced no action. Shorten or change the forum if it is not changing work.
Bring the evidence to the team
Supportman delivers team and per-agent reports into Slack on a predictable cadence, so the facilitator can select discussion items and share the pre-read without assembling the numbers by hand.
Start your weekly quality review with the evidence already in Slack →