A new technician onboard in the trades costs the shop $42,000-$78,000 fully loaded in the first 90 days and produces $14,000-$28,000 of net revenue in the same window. The 90-day gap is the bet the owner makes when she signs the offer. Most shops manage that bet with a one-page checklist, a clipboard, and a handful of "ride along with Marco for two weeks" promises that decay by day 21. By day 60 the new tech is on calls she should not be on, the recall rate is climbing, and the service manager is fielding callbacks she can not trace. By day 90 the owner is having the "is this the right hire" conversation with the service manager — six months too late to recover the bad placement and three months too early to defend a redeployment to the floor. This lesson is the named workflow that fixes that arc. The AI-built 30/60/90 technician onboarding tracker. Day-1-through-day-90 milestone targets on six metrics — MPR (Membership Penetration Rate), average ticket, recall percentage, financing presentation rate, membership pitch fidelity, ride-along grade — with AI building the per-tech tracker every Sunday night so the service manager walks into Monday with the new tech's exact position on the ramp curve, the next week's specific coaching focus, and the cumulative evidence base for the 30/60/90 review decisions. Shops running this workflow in 2026 hit day-90 veteran-band metrics on 65-75% of new hires versus 25-35% in unaudited shops, redeploy or exit wrong-fit hires at day 90 rather than day 270 (saving $30K-$90K per wrong-seat decision), and produce a recruiting differentiator that compounds across two and three years of hires.

Why 90-Day Onboarding Is the Bet Most Shops Lose

The math is brutal and the math is consistent across HVAC, plumbing, electrical, and drain. A new service tech earns $28-$42 per hour base, runs $55-$85 per hour fully loaded with payroll taxes, vehicle, fuel, phone, uniform, and tool allocation, and consumes another $1,200-$2,800 per month of service-manager time during the ramp window in ride-alongs, debriefs, scorecard reviews, and recall investigations. Across 13 weeks the fully-loaded cost lands between $42K and $78K depending on shop pay band and ramp intensity. Against that cost, a new tech produces $14K-$28K of net revenue in the first 90 days — they are running fewer calls per day than a veteran, closing partial-replace work at a lower rate, attaching memberships at 8-15% versus a veteran's 35-50%, and presenting financing on 12-18% of qualifying jobs versus a veteran's 28-40%. The gap between cost and revenue in the ramp window is the bet.

Shops that win the bet run the new tech to veteran metrics by day 90 — average ticket within 15% of the floor median, MPR within 5 points of floor median, recall percent within 1 point of floor median, financing presentation in band. From day 91 forward the new tech is net-margin-additive and the ramp investment recovers in months 4-9. Shops that lose the bet still have the new tech at 60-70% of veteran metrics at day 90, do not have the evidence base to defend a redeployment decision, and either keep paying for an underperformer for another 90-180 days or fire and re-recruit at a $30K-$90K cumulative cost. The decision-making cost of losing the bet exceeds the operating cost.

The bet is not a hiring problem. The bet is an onboarding-discipline problem. The same new hire placed into a shop running the 30/60/90 tracker outperforms by 20-30 points across every metric at day 90 versus the same hire placed into a shop running a clipboard. The named workflow is what closes the discipline gap.

The Six Metrics the Tracker Owns

The tracker tracks six metrics. Not 12. Not 18. Six. The discipline is to pick the six that predict 90-day veteran-band attainment and ignore the noise. Empirically, across deploying trades shops in 2026, these are the six.

First, average ticket. The single most-watched tech metric on every dashboard. Day-30 target $380-$450 (the new tech is running diagnostic-fee-heavy days with limited repair authorization), day-60 target $520-$620, day-90 target within 15% of floor median ($580-$720 at most $5M-$15M residential shops). Second, MPR — Membership Penetration Rate. Day-30 target 12-18%, day-60 target 22-30%, day-90 target 32-42%. Third, recall percent. Day-30 target under 6% (the new tech has noisy data; tolerate higher), day-60 target under 4%, day-90 target under 2.5% (within 1 point of floor median). Fourth, financing presentation rate on qualifying jobs (jobs over $5K total). Day-30 target 18-25%, day-60 target 32-40%, day-90 target 45-55%. Fifth, membership pitch fidelity — percentage of qualifying calls where the AI-summarized tech voice notes show a structured membership pitch was delivered. Day-30 target 60-70%, day-60 target 80-88%, day-90 target 92%+. Sixth, RGA — Ride-along Grade Average from Rilla virtual ride-alongs scored against the shop's RGA rubric. Day-30 target 50+, day-60 target 65+, day-90 target 75+ (veteran band 80+).

Why These Six and Not Others

Other plausible metrics get dropped from the tracker deliberately. RPT (revenue per truck) is correlated with average ticket and call volume; tracking both creates noise and double-counts the same underlying behavior. Close rate (percentage of repair-or-replace pivots that close) belongs on the Comfort Advisor scorecard, not the service tech tracker — service techs hand off to advisors at the kitchen-table pivot. Callback rate is a subset of recall percent and tracked inside the same number. Customer-satisfaction NPS surveys lag the call by 14-21 days and feed back too slow to coach. Call-volume-per-day is a dispatch decision, not a tech behavior, and the dispatcher controls the loading.

The six metrics chosen are leading indicators of veteran-band performance, controllable by the tech in the moment of the call, observable through AI-instrumented data sources (ServiceTitan voice notes, Rilla ride-along recordings, ServiceTitan ticket history, financing-portal pulls), and uncorrelated enough that movement on each tells the manager something distinct. The six are the right six.

The AI-Built Tracker — The Sunday-Night Build

The tracker builds Sunday night. The service manager does not touch it. The shop's locked AI prompt pulls the trailing 7-day data from ServiceTitan (tickets, average ticket, MPR, financing presentation, recall flags), Rilla (ride-along recordings and RGA scores), and the shop's membership-pitch fidelity scoring from AI voice-notes (the same voice-note structure introduced in L2 Chapter 4). Output: a one-page per-tech tracker landing in the manager's shared folder by 9 p.m. Sunday so the manager can read it Monday morning over coffee before stand-up.

Per-tech tracker fields. Top of page: tech name, hire date, day-N indicator (day 27 of 90), ramp-curve position (on-target / behind / ahead of milestone band). Block 1: six metrics with current 7-day value, week-over-week trend arrow, distance from current-milestone target, distance from day-90 veteran target. Block 2: AI-quoted exchange from the prior week's calls — one cleanest pitch, one missed opportunity (AI surfaces evidence; the manager writes the coaching, same constraint as the CSR audit). Block 3: this-week's single coaching focus (the binding constraint metric, written as a behavior change). Block 4: 30/60/90 milestone status — on-target, attention, or extended-ramp-trigger flag. Block 5: cumulative-evidence summary for any milestone decision (day-30 review, day-60 review, day-90 review) due in the next 14 days.

Why Sunday Night and Not Monday Morning

Sunday night beats Monday morning for three reasons. Monday morning the service manager has stand-up, dispatch escalations, the after-hours callback list, and the weekend recall flags competing for attention. AI-built tracker output landing at 9 p.m. Sunday gives the manager 30-45 minutes Sunday or 20 minutes Monday over coffee to read each tracker and walk into Monday with a plan rather than reacting to the data live. The tracker is also re-readable on the manager's phone Sunday night — same spaced-repetition logic as the CSR coaching cards. Manager retention of the per-tech plan walking into Monday with overnight processing is 30-40% higher than reading the same data live at 7 a.m. amid the morning fires.

The Day-30 Milestone and the Coaching Focus Shift

Day 30 is the first formal review. Service manager, new tech, and (optional) lead tech sit for 30 minutes. The AI tracker has produced four weeks of data; the conversation is structured around the data, not anecdote. The format mirrors the CSR audit huddle: cleanest moments first, missed opportunities second, focus for the next 30 days third, commitments fourth.

Day-30 target hits expected on 4 of 6 metrics. Average ticket $380-$450, MPR 12-18%, recall under 6%, RGA 50+. Financing presentation and membership pitch fidelity are leading indicators and may lag (target band but not strict gating). Pattern of hits: 4-6 of 6 = on-target ramp, proceed to day-60 cadence; 2-3 of 6 = extended-attention ramp with weekly check-in instead of biweekly; 0-1 of 6 = early-warning flag, intensive ride-along week with senior tech, conversation about fit but not a wrong-seat decision (too early). The new tech has 60 more days to either close the gap or accumulate the evidence that the seat is not the right one.

The Day-30 Coaching Focus Shift

The first 30 days of coaching focus on diagnostic discipline and the system-condition narrative — the foundation behaviors that gate every downstream metric. Day 30 forward, the coaching focus shifts to repair-or-replace pivot mechanics and membership pitch fidelity. The shift is named so the new tech understands the ramp arc: weeks 1-4 we coach foundations, weeks 5-8 we coach pivots and pitches, weeks 9-12 we coach financing presentation and design-vs-base attach. The named arc lets the tech see where she is on the curve and what to expect next. Without the named arc the new tech feels coached randomly and the trust in the workflow erodes.

The Day-60 Milestone and the Binding Constraint Test

Day 60 is the binding-constraint test. By now the new tech has 56 days of data; trend signals are clean; the AI tracker shows whether the ramp is converging on veteran-band or diverging. Day-60 target hits expected on 5 of 6 metrics. Average ticket $520-$620, MPR 22-30%, recall under 4%, financing presentation 32-40%, membership pitch fidelity 80-88%, RGA 65+. Pattern of hits: 5-6 of 6 = strong trajectory; 3-4 of 6 = focused-intervention ramp (single metric coaching intensified, paired with senior tech for 4 shadow shifts); 2 or fewer of 6 = wrong-seat conversation begins but no decision until day 90 with the full evidence base.

The binding-constraint test surfaces the single metric most likely to gate day-90 veteran-band attainment. If average ticket is in band but financing presentation is at 18% versus 32-40% target, the binding constraint is financing — and the day-60 to day-90 coaching focuses entirely on financing presentation mechanics, soft-pull pre-approval discipline, and the kitchen-table financing pivot. If average ticket is at $440 versus $520-$620 target but everything else is in band, the binding constraint is repair-or-replace pivot timing — and the coaching focuses on the diagnostic-to-options handoff. The tracker surfaces the constraint; the manager runs the focused coaching for 30 days; the day-90 review tests whether the focused coaching moved the metric.

The Day-90 Milestone and the Decision

Day 90 is the decision. The full evidence base is built — 12 weeks of AI tracker output, 12 ride-along sessions, 6 milestone reviews, ~85-100 service calls completed. Day-90 target hits expected on 5-6 of 6 metrics. Average ticket within 15% of floor median, MPR 32-42%, recall under 2.5%, financing presentation 45-55%, membership pitch fidelity 92%+, RGA 75+. Pattern of hits drives the decision.

5-6 of 6 metrics in band = veteran seat confirmed; new tech moves into standard floor cadence and standard biweekly scorecard. 3-4 of 6 metrics in band = extended ramp authorized; manager defends an additional 30-60 days of focused intervention with documented expectation that the named binding constraint moves into band; if it does, veteran seat at day 120-150; if it does not, exit conversation with the full evidence base. 2 or fewer of 6 = wrong-seat decision with cumulative evidence base; redeployment options surfaced (install support, lead-development specialist, customer-success role, parts-and-warehouse) before exit.

Why the Day-90 Decision Is More Accurate Than the Day-270 Pre-AI Decision

Pre-AI shops made the wrong-fit-tech decision at month 9-12 because the data was sparse — monthly close-rate reports, 1-2 in-person ride-along observations per quarter, anecdotal manager assessment. By month 9 the team had absorbed underperformance through reduced lead routing (self-fulfilling underperformance spiral); the cost of the bad seat was $60K-$120K fully loaded plus team morale damage. AI-built tracker at day 90: 12 weeks of daily data, ~85-100 observed calls, 6 metrics tracked weekly, RGA trend across the full window, cohort comparison against prior new-hire cohorts, coaching-responsiveness signal embedded in the tracker history.

Empirically across deploying 2026 shops, the false-positive rate (wrong-seat decision incorrectly made) drops from ~25% pre-AI to ~8% with the tracker; the false-negative rate (wrong-fit retained too long) drops from ~40% to ~15%. Data density compensates for the shorter window. The day-90 decision is both faster and more accurate than the day-270 decision; the manager has 2-4x the evidence base in 30% of the time. Honest data earlier protects the tech (cleaner redeployment), the team (less morale damage), and the P&L ($30K-$90K saved per wrong-seat decision).

Cross-Coordination with Recruiting and the Cohort Effect

The tracker also feeds the recruiting function. Each new-hire tracker becomes a per-hire dataset. Across 12 hires over 24 months, the cumulative tracker archive surfaces patterns — which recruiting source produced the highest day-90 veteran-band attainment rate; which interview signal correlated with day-30 RGA above 55; which prior-shop experience produced fastest financing-presentation ramp. The recruiting workflow updates against the tracker patterns: posting copy emphasizes the named onboarding structure, interview rubric weights the signals correlated with success, reference-check questions target the pattern that emerged.

The cohort effect compounds. New hires in cohort year-2 enter a shop with a refined recruiting signal, an established 30/60/90 tracker culture, and senior techs who have themselves been through the tracker and can mentor with structured language. Cohort year-2 day-90 veteran-band attainment runs 75-85% versus 65-75% in cohort year-1. The named workflow becomes a recruiting moat the longer it runs. PE buyers in 2026 diligencing trades shops ask about onboarding workflow durability — the named tracker with multi-year archive is the artifact that defends the answer.

The Manager's Deliverable for This Chapter

By the end of this lesson the service manager builds four artifacts. First, the locked AI prompt for the per-tech tracker — versioned in the shared folder, names the six metrics, names the milestone-band targets per day-30/60/90, names the output fields (5 blocks named above), constrained to AI-surfaces-evidence with manager-writes-coaching discipline. Second, the per-tech tracker template — Google Doc or Notion page format with the five blocks, sized to one page per tech per week. Third, the 30/60/90 milestone review template — structured agenda for the three formal reviews, with cumulative-evidence summary auto-pulled from the trailing tracker weeks. Fourth, the recurring calendar entries — Sunday 9 p.m. tracker build (automated), Monday 8 a.m. manager read window, day-30, day-60, day-90 reviews calendared from each new hire's start date with the owner copied for day-60 and day-90 reviews.

The four artifacts convert the 30/60/90 onboarding from an ad-hoc clipboard process into a named workflow. Within the first hire cycle the manager runs the workflow and produces the day-90 decision with evidence. Within four hire cycles (typical 12-18 month window at a six-tech shop) the recruiting differentiator emerges and the cohort effect begins. Within 24 months the shop has a multi-year tracker archive, a refined recruiting signal, a known day-90 veteran-band attainment rate, and a named workflow that PE buyers diligence as enterprise-value protection. The next lesson covers the same operating discipline applied to the parts-pricing problem: AI-built landed-cost and tariff-impact tracking, with daily supplier-price pulls feeding the flat-rate pricebook so margin recovery happens in 5 minutes per week of approval rather than 5 weeks of margin slippage.

Key Takeaways

• The 90-day onboarding bet is the math most shops lose. $42K-$78K fully loaded cost against $14K-$28K net revenue in the first 90 days. Shops that hit day-90 veteran-band metrics recover the ramp investment in months 4-9; shops that do not either keep paying for an underperformer or face $30K-$90K wrong-seat cost. The bet is an onboarding-discipline problem, not a hiring problem.
• The tracker owns six metrics, not 12. Average ticket, MPR, recall percent, financing presentation rate on $5K+ jobs, membership pitch fidelity (AI-scored from voice notes), and RGA from Rilla. Each is a leading indicator of veteran-band performance, tech-controllable in the moment, AI-instrumented, and uncorrelated enough that movement on each tells the manager something distinct.
• The AI tracker builds Sunday night. Locked prompt pulls trailing 7-day data from ServiceTitan, Rilla, and voice-note fidelity scoring; one-page tracker per tech lands in the shared folder by 9 p.m. Sunday. Manager reads Sunday night or Monday over coffee — 20 minutes total — and walks into stand-up with a plan rather than reacting live.
• Day-30 milestone tests 4-of-6 metrics in band. Day-30 targets: average ticket $380-$450, MPR 12-18%, recall under 6%, RGA 50+, financing and pitch fidelity leading indicators. Coaching focus shifts at day 30 from foundations (diagnostic discipline + system-condition narrative) to pivots (repair-or-replace mechanics + membership pitch).
• Day-60 milestone identifies the binding constraint. Day-60 targets: average ticket $520-$620, MPR 22-30%, recall under 4%, financing 32-40%, pitch fidelity 80-88%, RGA 65+. The binding-constraint test surfaces the single metric most likely to gate day-90 veteran-band attainment; coaching focuses entirely on that constraint for the next 30 days.
• Day-90 milestone drives the decision. 5-6 of 6 = veteran seat. 3-4 of 6 = extended ramp 30-60 days. 2 or fewer = wrong-seat with redeployment options. Day-90 decision false-positive rate drops from ~25% pre-AI to ~8% with the tracker; false-negative rate drops from ~40% to ~15%. Faster and more accurate than the day-270 pre-AI decision; saves $30K-$90K per wrong-seat decision.
• The deliverable is four artifacts. Locked AI prompt, per-tech tracker template, 30/60/90 milestone review template, recurring calendar entries with owner-copied 60- and 90-day reviews. Year-2 cohort effect lifts day-90 veteran-band attainment from 65-75% to 75-85% as the tracker archive refines recruiting signal and senior techs become structured mentors. The named workflow becomes a recruiting moat and PE-diligence artifact.