--- name: sales_coaching description: Coach B2B sales reps with evidence over a repeating cycle: score recorded calls and meetings against a behaviour rubric, pick one focus skill per cycle, run practice drills tied to a live deal, and re-score later calls to check whether the behaviour actually changed. Use when a manager or peer coach reviews a rep's calls, builds a coaching plan, prepares a coaching one to one, runs a role play for an upcoming meeting, or checks whether past coaching worked. For running your own discovery call use discovery_call_playbook. For general one to one coaching use coaching_conversation. --- # Sales Coaching ## Purpose Most sales coaching fails in the same three places. Feedback is given from memory instead of from the call. The rep leaves with six things to fix and fixes none of them. Nobody checks a later call, so there is no way to know whether anything changed. This skill runs coaching as a short cycle with evidence at both ends: 1. Score real calls against a rubric, with timestamps and buyer quotes. 2. Pick one focus skill for the cycle, chosen from the pattern across calls, not from one bad call. 3. Coach it in a session where the rep does most of the diagnosing. 4. Drill it against a live deal the rep is about to work. 5. Re-score the next comparable calls and record whether the behaviour moved. Scope boundary: this is the manager or peer coach working on a rep's selling behaviour. Deal strategy for one opportunity belongs in `deal_risk_review`. Qualification scoring belongs in `meddic_qualification`. The rep's own call preparation belongs in `discovery_call_playbook`. The method rests on four verified bases, used plainly: - Deliberate practice research (Ericsson, Krampe and Tesch-Romer, 1993): improvement comes from work on a specific subskill, on a well defined task at the right difficulty, with informative feedback and repetition with refinement. Hours on the job alone do not produce it. Most sales drills have no expert designed curriculum behind them, so this skill calls them purposeful practice, which is the honest label. - Feedback Intervention Theory (Kluger and DeNisi, 1996 meta-analysis): feedback improved performance in about two thirds of cases and worsened it in about one third. Feedback aimed at the task and at how to do the task helped. Feedback aimed at the person, such as trait labels, often hurt. That is why every score in the rubric points at a behaviour and a timestamp, never at a trait. - Call data at scale (Gong, 2025 update, 326,000 calls of at least 10 minutes): the average seller talked 60 percent of the time, closed won calls averaged 57 percent seller talk and lost calls 62 percent, and high performers held a steady talk share win or lose while low performers swung about 10 points. Use this to coach consistency, not to chase one magic ratio. The older 43 percent talk figure still circulates. Gong's own update supersedes it. - SPIN research (Rackham, 1988, from observation of more than 35,000 sales calls): stronger sellers in larger sales developed needs with Problem, Implication and Need-payoff questions before presenting a solution. The rubric scores question quality in those terms. Full rubric: `references/rubric.md`. Drills: `references/drills.md`. Worked reviews: `references/worked-examples.md`. Templates: `assets/call-review-template.md` and `assets/coaching-plan-template.md`. Scorer: `scripts/score_review.py`. ## Workflow ### 1. Build the evidence pack before forming a view Collect two or three recorded calls from the last two weeks. Mix call types only if the focus is a general behaviour such as listening. For a stage specific skill, such as discovery, use calls from that stage. For each call, note the stage, the buyer roles present, the outcome so far, and whether the rep knew it would be reviewed. Add the deal context from the CRM in two lines. Do not coach a call you have not listened to or read in transcript. Coaching from a rep's summary produces coaching about their summary. Consent and storage rule first: use only calls recorded under your team's recording policy and local consent law, and keep reviews inside the systems your team already uses for call material. ### 2. Score each call against the rubric Score all eight dimensions in `references/rubric.md` from 0 to 2, using the anchors there. Every score needs evidence: a timestamp and what was said or done. A score with no timestamp is an opinion, and the rep is right to discount it. The eight dimensions: opening and agenda, discovery depth, implication and impact, listening and interactivity, objection handling, decision process and access, value articulation, next step. Write the review into `assets/call-review-template.md`, then run the scorer: ``` python3 scripts/score_review.py review.md ``` The scorer checks that all eight dimensions are scored, that at least two timestamped evidence points are present, that exactly one focus skill is named, and that a follow up date exists. It prints the total, the lowest dimensions, and any hard gaps. Score the behaviour, not the outcome. A call can score well and still lose to budget. A call can score badly and advance because the buyer had already decided. Outcome data matters at the pattern level in Step 3, not as a grade for one call. ### 3. Pick one focus skill for the cycle Read across the two or three scored calls and pick the focus by three tests, in this order: - Recurrence: the gap shows up in at least two calls, or twice in one long call. A one off miss is a note, not a coaching cycle. - Leverage: fixing it would change the most downstream moments. Weak implication work starves value articulation and next step quality, so it usually outranks a weak opening. - Readiness: the rep has a live deal in the next two weeks where the skill will be used for real. A focus with no live rep is homework, and homework gets skipped. Name the focus as a behaviour a camera could catch. "Ask at least two implication questions per main problem before presenting" is a focus. "Improve discovery" is a topic. "Be more confident" is a trait, and coaching it breaks the feedback rule in the Purpose section. Record the focus, the baseline evidence, and the target in `assets/coaching-plan-template.md`. One focus per cycle. Park other gaps on the plan as later items, in writing, so the rep knows they were seen and deliberately deferred. ### 4. Run the coaching session in 30 to 40 minutes Run the session in this order. The rep talks first at each stage, the coach adds evidence after. - Playback (10 minutes). Play or read two short moments: one where the focus behaviour was missing and one where the rep did part of it well. Keep each clip under two minutes. Ask "What do you notice?" before you say what you noticed. - Diagnosis (5 minutes). Agree on what happened and why, in behaviour terms. Separate three causes, because each takes a different fix: the rep did not know, the rep knew and could not do it under pressure, or the rep chose not to because something else felt safer. Knowledge gaps take teaching. Skill gaps take drills. Choice gaps take a conversation about the trade off, often inside a GROW shape: goal for the next calls, reality from the evidence, options, and a commitment the rep words themselves. - Model (5 minutes). Show the behaviour done well once. Use a short clip from a strong call your team has permission to share, or do it live yourself in role. One model, then stop talking about it. - Drill (10 to 15 minutes). Run a drill from `references/drills.md` matched to the focus, built on the rep's real upcoming deal. See Step 5. - Commitment (5 minutes). Close with three lines the rep writes, not you: the behaviour, the named deal and call where they will use it, and the date you will both check the recording. If there is no named call and date, the session was a chat. Coach talk discipline: ask before telling, and when you do tell, tie it to a timestamp. Praise counts only when it is as specific as the correction. "Good discovery" teaches nothing. "You followed the outage answer with two implication questions at 18:40 and the buyer gave you the cost figure" teaches the rep what to repeat. ### 5. Drill against a live deal, not a generic scenario Generic role play transfers badly because the rep rehearses lines for a buyer who does not exist. Build every drill from a real upcoming call: the actual account, the actual people, the problem as currently understood, and the gap the last call left open. Run drills in short rounds. One round is 3 to 5 minutes on one moment, not a full call replay. After each round give one piece of task feedback, tied to the focus behaviour only, and run the round again at once. Two or three rounds on one moment beat one pass through a whole call. Raise difficulty across rounds: a cooperative buyer first, then the buyer's real objection, then time pressure or a second stakeholder. Stop conditions: stop a round when the rep is flailing with no foothold and reset with a hint or a model line. Stop the drill block when the rep lands the behaviour twice in a row at full difficulty. That is the practice target, and pushing past it into fatigue teaches sloppiness. Keep drills private or in a small trusted pod. Never drill a named rep's weakness in front of the whole team. Team sessions work on shared clips that the rep involved has agreed to share. ### 6. Follow up on the next comparable calls On the date set in the session, score the next one or two comparable calls on the focus dimension only, plus a quick pass on the other seven to catch regressions. Compare against the baseline in the coaching plan. Record one of three verdicts, with evidence: - Moved: the behaviour appears at the target level in a real call. Name the timestamp. Close the cycle or raise the bar on the same skill. - Partial: the behaviour appears but inconsistently, or only when the buyer makes it easy. Run one more drill round and check one more call, with a new date. - Not moved: the behaviour is absent again. Diagnose before repeating yourself. Wrong focus, no real chance to use it, drill too far from the real call, or a choice gap you coached as a skill gap. Change one of those. Repeating the same session louder is not a plan. After two or three cycles on the same rep, read their plans together. Patterns across cycles are the real coaching record: which skills stuck, which keep relapsing, and whether relapse follows a call type, a buyer type, or workload. That read feeds territory, pairing and enablement decisions, not just the next session. ### 7. Calibrate with other coaches monthly If more than one person coaches, score the same call independently once a month and compare. Dimensions where coaches differ by a full point need sharper anchors, so write down the example that settled the argument and add it to your local copy of the rubric. Uncalibrated rubrics drift into grade inflation within a quarter, and inflated scores make Step 3 pick the wrong focus. ## Output Contract For a call review, deliver the completed review: all eight dimension scores with timestamped evidence, the scorer output, the single worst gap, and the behaviour the rep did well and should repeat. For a coaching cycle, deliver the completed coaching plan: one focus skill stated as an observable behaviour, the baseline evidence from at least two calls, the drill used and how it was built from a live deal, the rep's commitment with a named call and check date, and, after follow up, a moved, partial or not moved verdict with evidence. For a session prep request, deliver a session outline in the Step 4 order, with the two playback moments chosen by timestamp, the drill selected, and the commitment lines left blank for the rep to write. Report scores as behaviour evidence, not as a ranking of people. Never present one call's total as a performance grade. ## Operating Rules - No evidence, no score. Every rubric score and every coaching claim points to a timestamp, a quote, or a CRM fact. - One focus per cycle. A second focus dilutes the first. Park the rest in writing. - Feedback names behaviours and task choices, never traits, character or effort as a personality judgement. This follows the feedback evidence in the Purpose section, and it is also what makes feedback safe to hear. - The rep self assesses first in every session. A coach verdict delivered before the rep has spoken trains dependence, not skill. - Drills use real upcoming deals, run in short rounds, and end after two clean landings at full difficulty. - Every cycle ends with a dated follow up check on a real call. A cycle with no follow up check stays open until that check happens. - Do not present call data figures from memory. If you cite the Gong talk share or question count patterns, cite them as Gong's published analysis with its sample, and note that they describe correlation in recorded calls, not a quota for one rep. - Respect recording consent, buyer privacy and your team's data rules. Do not move call recordings or transcripts into systems that are not approved for them.