“Clarity: 12/20” tells an agent an evaluator saw a gap. It does not say which conversations, what good looks like, or what to practise on the next shift.
A coaching plan turns a repeated QA pattern into one situation, one observable behaviour, one practice method, and one review date. The rest of this article walks a single illustrative case—an agent whose Clarity scores drop mainly in refund and goodwill-credit conversations—so you can copy the artefact, not just the advice.
Supportman turns QA scores into per-agent insights, so your coaching plan writes itself from real conversations instead of gut feel.
Start with a pattern
Do not build a plan from one awkward chat. Look across enough comparable work that a miss is a habit, not a bad day.
A workable bar for most teams:
- At least 8–12 conversations in the same situation over 2–4 weeks
- The gap appears in roughly half or more of that sample
- The behaviour is within the agent’s control
- The miss costs the customer something real (confusion, a follow-up ticket, delayed money)
Also check issue mix, channel, and factors outside the agent’s control before you coach. A Clarity dip that coincides with a broken billing tool is an operations problem first.
Illustrative pattern:
Over three weeks, Jordan’s Clarity attribute averaged 12/20 against a team median of 17/20. In 6 of 9 refund and goodwill-credit conversations reviewed, the reply confirmed a credit would be issued but never stated the amount, the eligibility basis, or when the customer would see it. Three of those customers opened a follow-up asking where the money was.
That is coachable. “Needs to communicate better” is not.
Step 1: name the task
Name the situation narrowly enough that practice can look like real work:
Refund and goodwill-credit conversations where Support can approve the credit in-product.
Avoid coaching “clarity” across every queue. If the gap only shows up in refunds, practise refunds.
Step 2: describe the performance gap
Attach feedback to the work, with a concrete miss:
The reply said “I’ve gone ahead and processed a credit for you” and closed the conversation. It did not name the amount, explain why the customer qualified, or say when the credit would appear on the statement.
Keep the diagnosis behavioural. Personality labels (“careless,” “not empathetic,” “lacks ownership”) are hard to practise and easy to resent—see why some agent feedback backfires.
Step 3: define the target behaviour
Write an action the agent can take before the conversation closes:
Before closing a refund or goodwill-credit conversation, state all three: (1) credit amount, (2) eligibility reason in one sentence, (3) when the customer should see it.
If the agent cannot see amount or timing in the tools, that is a workflow fix, not a coaching plan.
Step 4: show a strong example
I’ve issued a $48 goodwill credit because the outage blocked your team for two billing cycles. It should appear on your statement within 3–5 business days. Reply here if it has not landed by next Wednesday and I’ll chase it.
The example removes ambiguity. It is a model, not a script to paste unchanged into every ticket.
Step 5: practise deliberately
Deliberate practice needs a defined task, a clear gap from the target, focused repetition, and feedback soon after the attempt.
A study evaluating written feedback through a deliberate-practice framework found that specific performance gaps appeared in only 3.9% of feedback encounters and action plans in 13.7%. The authors argued that feedback improves when it names both the gap and a concrete learning plan. Read the feedback-and-learning-plan study.
For Jordan, useful practice in the next five working days:
- Rewrite three of the weak refund replies using the three-part close
- Annotate two strong teammate examples for amount, reason, and timing
- Have the manager spot-check the first five live refund chats after the coaching conversation
Skip generic “be clearer” homework. Practise the same situation you will score.
Step 6: set a measurable goal
This is where most QA coaching plans go soft. “Raise Clarity” is a wish. A coaching goal should be specific enough that next Friday’s review can pass or fail it.
A practical review of goal-setting for performance management restates the core finding: specific, challenging goals outperform vague “do your best” instructions, drawing on a large body of laboratory and field research. The same review is careful about the conditions that make goals work—or fail. Results depend on goal commitment, task complexity, goal framing, team goals, and feedback. Read the goal-setting evidence review.
Apply those conditions inside a support coaching plan:
- Specific and challenging: Name the situation, the behaviour, the sample, and the date. “Include amount, reason, and timing in every refund close for the next two weeks” beats “improve clarity.” Challenging means a stretch relative to the agent’s current hit rate—not an impossible 100% on day one if the current rate is near zero.
- Commitment: Agree the goal in the conversation. Ask what would make it hard. If the agent does not believe the standard is fair or feasible, compliance will be cosmetic.
- Task complexity: If the behaviour requires new product knowledge, tool access, or policy judgment, teach or unblock that first. A hard goal on a task the agent cannot yet perform produces frustration, not skill.
- Framing: Frame the goal as development against a customer outcome (“fewer ‘where’s my refund?’ follow-ups”), not as punishment for a score.
- Team vs individual: If half the queue misses the same close, fix the macro, snippet, or playbook. Save individual goals for patterns that are truly individual.
- Feedback: Schedule the review when you set the goal. Without a checkpoint, the goal is a speech.
Illustrative goal for Jordan:
In refund and goodwill-credit conversations from 12 May to 23 May, include amount, eligibility reason, and timing before closing. Mid-point check: sample five conversations on 16 May. Final check: sample five more on 23 May. Success = at least 4 of 5 in the final sample include all three elements. Clarity score is secondary context, not the goal.
Prefer a behavioural hit rate over “raise Clarity from 12 to 16.” Scores can move for mix reasons; the behaviour is what you coached.
Step 7: follow up
At each checkpoint:
- Sample comparable refund/credit conversations only—do not dilute the review with unrelated chats
- Score the target behaviour present / absent, then discuss one example together
- Ask what felt awkward or blocked in the tools
- Recognize a clean three-part close by name
- Extend, tighten, or retire the goal
If the behaviour does not move after two honest practice cycles, investigate before writing a harsher plan:
- The target was still too vague
- The agent cannot see amount or timing in the UI
- Policy authority is unclear
- Practice did not resemble live refund work
- Feedback arrived days after the conversations
A worked coaching plan
Copy this shape. The filled version is illustrative; replace the bracketed blanks for the next agent.
| Field | Jordan (illustrative) | Your agent |
|---|---|---|
| Rubric attribute | Clarity — 12/20 over 3 weeks (team median 17/20) | [Attribute + recent score vs team] |
| Situation | Refund and goodwill-credit conversations Support can approve in-product | [Specific conversation type] |
| Pattern | 6 of 9 reviewed chats confirmed a credit without amount, reason, or timing | [Repeated evidence + sample size] |
| Customer impact | Follow-up tickets asking where the refund is | [Why this matters to the customer] |
| Target behaviour | Before close: state amount, eligibility reason, and timing | [Observable action] |
| Strong example | See Step 4 reply above | [One model reply] |
| Practice | Rewrite 3 prior replies; annotate 2 strong examples; manager spot-checks first 5 live refunds | [Exercise + live-work plan] |
| Success measure | ≥4 of 5 sampled refunds on final check include all three elements | [Behavioural hit rate, not “raise score”] |
| Checkpoints | 16 May (mid), 23 May (final) | [Dates] |
One plan, one behaviour. If you leave the meeting with three goals, you left with none.
Use AI to prepare, not replace, coaching
AI is useful for finding the pattern across every eligible conversation, pulling representative refund examples, and drafting a first-pass plan. The manager still validates the evidence, hears the agent’s context, removes tool or policy blockers, and owns the follow-up.
How to keep that division of labour—and how often to refresh plans—is covered in how AI can help managers coach without generic advice. This article stops at the plan itself.
Turn evaluation into improvement
Supportman surfaces per-agent rubric patterns with linked conversation evidence, so the coaching conversation starts from a repeated miss rather than a gut feel or a single score.
See how Supportman turns QA into coaching →
Frequently asked questions
How many coaching goals should an agent have at once?
One active behavioural goal is the default. Add a second only when the first is automatic in live work, or when a separate workflow blocker needs a parallel operations fix.
How long should a coaching goal run?
Long enough for deliberate practice and a comparable sample—often one to two weeks for a frequent situation, longer if the conversation type is rare. Change earlier if the diagnosis was wrong or the tools make the behaviour impossible.
Should the goal be a QA score or a behaviour?
Prefer the behaviour and a hit rate on sampled conversations. Use the attribute score as supporting context. Scores move with ticket mix; the behaviour is what you taught.