It is 3:47 p.m. Friday. The owner is between two appointments. The service manager has 47 minutes before the owner walks out for the weekend. The one artifact that gets read between now and Monday morning is the Friday Owner Recap — one page, seven numbers, three wins, two problems, one ask. It is the highest-leverage written communication in the shop's week. It tells the owner what moved, what did not, and what decision the owner has to make before Monday morning's tech stand-up. The recap is the L3 service manager's deliverable to the L3 owner; it lives at the intersection of L3 Ch6 marketing attribution, L3 Ch4 tech scorecards, L3 Ch5 RC&D triage, L3 Ch3 dispatch yield, L3 Ch2 missed-call recovery, and the AI-built data flow that makes a 21-minute Friday afternoon assemble enough operating signal for an owner to run the shop on Monday. This lesson is the build. The recap's seven named numbers, the three-win two-problem one-ask format, the AI assembly chain, the marketing-manager / service-manager edit pass, the 4-minute owner read, the failure modes that turn the recap into noise, and the quarterly evolution that surfaces seasonality and PE-readable signal. The shops that run this Friday recap discipline operate with owner-level attention to the right numbers; the shops that skip it operate by gut every Monday morning.

Why One Page, Seven Numbers

The owner does not read a five-page report on Friday afternoon. They scan a one-page artifact in four minutes between the 3:15 advisor call and the 4:30 supplier call. Anything that does not fit on one page does not get read. Anything not in the top seven numbers does not move the owner's Monday decisions. The discipline is reductive on purpose — the owner's bandwidth is the constraint, not the data's availability. The L3 service manager who locks the format at one page and seven numbers produces a document the owner reads cover-to-cover every Friday; the one who tries to fit twelve numbers and three pages produces a document the owner skims at the headline and ignores at the body.

Seven is the calibrated number. Five misses average ticket and GLSA ROAS, which together explain the marketing-spend question the owner asks every Monday. Ten is too many — the owner stops reading at six. Seven hits the cognitive limit for scan-read while covering the seven decision surfaces the owner allocates capital and attention against. Each maps to a Monday morning decision (where the CSR coaching focus goes, which tech needs a ride-along, which membership campaign needs reinforcement, which financing tier needs talking-points refresh, which recall needs root-cause attention, which truck needs a re-route, which marketing channel needs more or less spend).

The three-wins-two-problems-one-ask wrapper turns the seven numbers into action. Wins re-anchor the team on what worked. Problems force prioritization — two forces ranking, not listing. One ask surfaces the decision the owner actually has to make. Together: numbers (where the shop is), wins (what to double down on), problems (what to fix), ask (what the owner decides). The recap is the operating loop in one page.

The Seven Numbers, Named

The recap's seven numbers are not negotiable. They are the L3 owner's standard read across every shop in the trades — Nexstar, BDR, Service Champions, CertainPath, and the PE platform set (Wrench Group, Authority Brands, Apex Service Partners, Sila Services, Path Light Pro, Redwood Services, ARS/Rescue Rooter) all reference variants of the same seven. The L3 service manager builds to this set so the recap reads instantly to any owner, coach, peer group, or PE reviewer who lands on it.

Number One: Booking Percent

The CSR floor's primary metric. Target 80-85%, baseline 65%. The number that moves week-to-week the most because CSR training, scripts, and AI Avoca handoffs all land here. AI-recovered missed-call leakage, after-hours capture, abandoned-call recovery — every L3 Ch2 workflow output flows into this one number. The owner's Monday decision off this number: where does the CSR coaching focus go next week. The L3 service manager pulls from CallRail / ServiceTitan / Sera / HCP / Avoca aggregator dashboards.

Number Two: MPR (Membership Penetration Rate)

The tech scorecard's primary recurring-revenue metric. Target 35-50%, top quartile 60%+. The number that drives 18-30-month LTV through Step 37 (the 11-month renewal touch). The owner's Monday decision off this number: which tech needs the membership-pitch coaching huddle; which membership campaign (basic to premium upgrade) needs the next AI-built outbound. Pulled from FSM membership module reporting.

Number Three: Financing Close Percent

The Comfort Advisor's kitchen-table metric on $5K-plus tickets. Target 28-40%, baseline 14%. Wisetack / GreenSky / Synchrony soft-pull adoption, the financing tier matrix, the Skip-the-Quote door soft-pull workflow all feed here. Owner's Monday decision: which financing tier talking-points refresh, which advisor needs the Rilla scorecard pass on the financing pivot moment. Pulled from FSM financing reporting and the lender portal exports.

Number Four: Recall Percent

The shop's quality-and-margin canary. Target under 2%, industry median 4-7%. Recall (true same-problem-same-system within 30 days) vs. callback vs. warranty, surfaced by L3 Ch5 SLA-bound triage. The owner's Monday decision: which root-cause cluster gets attention (tech, part, install date, symptom); which warranty exception needs owner sign-off. Pulled from FSM ticket data with the L3 Ch5 classifier tags applied.

Number Five: RPT (Revenue Per Truck)

The dispatch board's economic output. Targets per Built on Tenth, Level CFO, MarginPlug 2026 data: shop baseline $1,600-$2,400/day for residential service trucks, top-quartile $2,800-$3,500/day; replacement trucks $2,500-$4,000/day top-quartile; commercial replacement $5K+/day. Pulled from the dispatch board's daily output ledger. Owner's Monday decision: which truck needs re-routing; which tech needs the dispatch yield ride-along.

Number Six: Average Ticket

The shop's per-job revenue. Residential service target $450-$600 per 2026 trades benchmarking. The number that catches the photo-annotated proposal lift ($180-$320 documented), the good/better/best presentation discipline, the repair-vs-replace conversation outcome. Owner's Monday decision: which tech needs the photo-tagging coaching; which jobsite-photo discipline needs the Friday huddle. Pulled from FSM ticket revenue divided by ticket count.

Number Seven: GLSA ROAS (and the Marketing Channel Mix)

The marketing spend's accountability number. Google Local Service Ads baseline 3-4x ROAS with mature optimization, target lift 30-50% on top of baseline using AI bidding (Ryze AI 2026 documented uplift band). The recap reports GLSA ROAS as the headline and references the channel mix (GLSA, PPC, Hatch, NiceJob, mailer) in the wins or problems section if the mix shifted. Owner's Monday decision: which channel gets more spend, which gets less, what the marketing manager is testing next. Pulled from CallRail / GLSA / Google Ads / Hatch dashboards aggregated by AI into the recap's number.

The AI Assembly Chain

The seven numbers do not assemble themselves. The L3 service manager runs a named AI workflow that pulls from six data sources, normalizes, computes period-over-period deltas, surfaces the largest movers, and drafts the wins/problems/ask narrative the L3 service manager and marketing manager edit. The workflow is the L3 Ch6 Lesson 3 deliverable — the AI-built Friday recap.

The data sources: ServiceTitan / Sera / HCP / FieldEdge / BuildOps (booking %, MPR, average ticket, RPT, ticket data for recall %); the lender portal exports (Wisetack, GreenSky, Synchrony for financing close %); CallRail Conversation Intelligence (call volume, abandoned %, after-hours capture); GLSA dashboard plus AI bidding tool (cost per booked call, GLSA ROAS); Hatch (nurture conversion rate, stale-lead reactivation); NiceJob / Podium AI Employee / Birdeye AI Employee (review velocity, response rate). The AI workflow pulls each source's API or CSV export, normalizes to the recap's units, computes the week-over-week and quarter-over-quarter deltas, ranks movers by absolute and relative change, and drafts the recap.

The AI's draft includes the seven numbers in headline form, three candidate wins (highest positive movers), two candidate problems (largest negative movers or threshold breaches), and one candidate ask (the decision the owner has to make this week given the data). The L3 service manager and marketing manager review and edit; the AI's draft is rarely shipped verbatim, but it shortens the L3 service manager's recap-build from 3 hours to 21 minutes. The 21 minutes is the calibrated cost — long enough to edit substantively, short enough to fit into the Friday afternoon between other obligations.

The 21-Minute Friday Build

The recap-build is a named workflow on the L3 service manager's Friday calendar. 3:30 p.m. start, 3:51 p.m. delivery. The workflow has four steps; the L3 service manager has run it weekly for at least a quarter before the rhythm holds.

Step One: AI Pull and Draft (7 min)

The L3 service manager triggers the AI workflow — Zapier, n8n, or a custom script that pulls the six data sources, normalizes, computes deltas, and produces the draft recap. The draft lands in the recap template (one-page Google Doc, Notion page, or PDF template). 7 minutes is the AI's processing time plus the L3 service manager's confirmation that the pull completed cleanly.

Step Two: Service Manager Edit (8 min)

The L3 service manager reads the AI's seven numbers and confirms accuracy against the L3 Ch8 Lesson 3 weekly AI output audit's accuracy dimension — every number cross-referenced against source-of-truth before the recap ships. Wins, problems, and ask are edited for shop-specific context the AI cannot have (Joe's tech ride-along scheduled Monday, the customer recovery on the Bel Air install, the supplier shortage on the R-454B coils). 8 minutes for substantive edits at the service-manager level.

Step Three: Marketing Manager Edit (4 min)

The marketing manager edits the GLSA ROAS commentary, the channel mix language, and the ask if the ask touches marketing spend. The marketing manager's edits also bring the brand voice up — the recap reads in the shop's voice rather than the AI's voice. 4 minutes for marketing-manager-specific edits.

Step Four: Format and Deliver (2 min)

Final formatting — title (Friday Recap — [Shop Name] — Week of [Date]), one-page layout confirmed, PDF generated, delivered to the owner via email and Slack pinned post. 2 minutes for the mechanical finish. The owner has the recap by 3:51 p.m. Friday, before the 4 p.m. supplier call.

The Four-Minute Owner Read

The recap is designed for a four-minute read. The owner skims the seven numbers in 90 seconds (~12 seconds per number), reads the three wins in 30 seconds, the two problems in 60 seconds, and the one ask in 30 seconds. Total: four minutes. The owner has read every number and decided whether the ask requires action before Monday or can wait until the Monday morning tech stand-up.

The format that supports the four-minute read: numbers in a horizontal row at the top, each as a delta (booking % 81 / +3 vs. last week / +5 vs. plan), wins as three bullet lines (max 18 words each), problems as two bullet lines (max 22 words each, slightly longer for context), ask as one bolded sentence at the bottom (max 25 words, in imperative form). The visual hierarchy is tuned for scan-read; the L3 service manager and marketing manager who format-deviate without owner approval are the reason most owners stop reading their service manager's recap by week six.

The owner's Monday morning is faster because the recap is read Friday. The 7 a.m. Monday tech stand-up references the recap's wins and problems; the dispatcher pre-positions the day's board around the recap's ask; the CSR floor knows where the coaching focus is this week because the recap named it Friday. The recap's leverage compounds across the week — Friday's read produces Monday's tighter operating tempo. Without the recap, Monday morning begins with the owner pulling numbers from five dashboards, asking the L3 service manager three times the same question, and rebuilding the operating signal from scratch.

The Three Wins, Two Problems, One Ask Pattern

The narrative wrapper is the recap's interpretive layer. Numbers without narrative produce data dumps the owner ignores; narrative without numbers produces vibes the owner cannot act on. Together: numbers anchored to action.

Three Wins — What Worked, Investable

Three wins per week — not two, not five. Three forces the L3 service manager to find three (most weeks there are five candidates) and to rank, not list. Each win is one bullet of 12-18 words: the tech, the metric movement, and the action implied. "Joe's MPR moved 31% to 44% on the new condenser-call membership pitch — coaching huddle Wednesday." "AI Avoca recovered 11 missed calls Saturday night — $4,200 booked revenue from after-hours leakage." "GLSA ROAS hit 4.8x on Tuesday's bid-adjustment — keep the bid the way it is for next week." Each win names the lever; the owner sees what to double down on by Monday.

Two Problems — What Did Not Work, Action-Required

Two problems per week — not one, not four. Two forces the L3 service manager to rank, not list. Each problem is one bullet of 18-22 words: the metric, the cause hypothesis, the proposed action. "Recall % climbed 2.1% to 5.4% — three Goodman coils on the Bel Air install batch failing same way — root-cause cluster Monday 9 a.m." "Marco's RPT dropped to $1,400 from $2,100 — three training calls on the schedule plus a callback that should have been on Jose — Rilla ride-along Tuesday." Each problem names the hypothesis; the owner sees what to fix by Friday next week.

One Ask — What the Owner Has to Decide

One ask per week. One only. The L3 service manager who lists three asks produces a recap the owner skims; the L3 service manager who locks at one ask produces a recap the owner answers by Friday night. The ask is the single decision that requires owner judgment this week — financing tier change, vendor approval, supplier shortage handling, customer recovery sign-off above threshold, a comp plan adjustment, a hiring decision. Bolded, one sentence, 20-25 words. "Approve the Wisetack tier-2 talking-points refresh for the August AC replacement push — marketing manager to ship by Tuesday." The owner reads, replies (often within an hour), and the L3 service manager and marketing manager ship Monday.

Failure Modes of the Friday Recap

Four failure modes recur across L3 shops that adopt the Friday recap workflow. The L3 service manager watches for each.

Failure Mode One: scope creep. The recap grows from one page to two, then three. Seven numbers become twelve. Wins multiply. Problems multiply. The owner's read time grows from four minutes to fifteen and the recap stops getting read. Detection: the recap exceeds one page in any week. Fix: hard format constraint — the template is one page and the L3 service manager cannot ship beyond it. Anything not fitting moves to a separate appendix the owner reads on demand.

Failure Mode Two: vague asks. The ask drifts from a decision-requesting sentence to a status update ("we are watching the GLSA ROAS situation"). The owner has nothing to decide. The recap loses its leverage as a decision-forcing artifact. Detection: ask does not contain a verb requesting owner action. Fix: ask template — the ask must contain "approve," "decide," "confirm," "authorize," "review and respond" or similar imperative.

Failure Mode Three: AI voice drift. The AI's draft ships through the L3 service manager and marketing manager edits with chatbot phrasing intact ("I'd be happy to elaborate," "Please don't hesitate to ask," "This recap aims to provide an overview"). The owner reads the AI's voice rather than the L3 service manager's. Detection: L3 Ch8 Lesson 3 weekly audit's voice dimension scores the recap below 4. Fix: brand-voice system prompt enforcement on the recap workflow; L3 service manager and marketing manager rewrite passes catch chatbot tells.

Failure Mode Four: number drift. The AI's pull misses a data source's update, the number is wrong, the recap ships with bad data, the owner makes a decision off the bad data. Detection: the L3 Ch8 Lesson 3 weekly audit's accuracy dimension on the recap workflow scores below 4. Fix: every number in the recap is cross-referenced against source-of-truth by the L3 service manager before ship; the AI's pull is automated but the verify is human; bulletin-board entry filed when the verify catches a drift.

Quarterly Evolution and PE-Readable Signal

The recap evolves across quarters. Quarter 1: format holds, the seven numbers tracked, the team learns the cadence. Quarter 2: seasonal patterns emerge — booking % rises into the summer cool-down peak, recall % rises in spring AC startup, financing close % rises when 0% promo windows open. The L3 service manager adds seasonal commentary without expanding the format. Quarter 3: year-over-year deltas surface. Quarter 4: the quarterly summary feeds the L3 governance review; the owner walks into the PE quarterly meeting with the recap series as the operating-discipline artifact.

PE reviewers in 2026 value documented operating cadence at premium multiples. A Wrench Group portfolio review, an Authority Brands franchisor audit, an Apex Service Partners QBR, or an M&A diligence pull asks the same question: how does this shop know what is happening week-to-week? The 52-week recap series — one page each, seven numbers consistently tracked, wins/problems/ask narrative, owner sign-off — is the artifact that answers. Same EBITDA, different operational discipline visibility, different exit multiple.

How the Recap Integrates with the L3 Stack

The Friday recap is the L3 Ch6 capstone artifact, drawing from every prior L3 chapter and feeding the L3 governance review. L3 Ch2 missed-call recovery feeds booking %. L3 Ch3 dispatch feeds RPT. L3 Ch4 ride-along coaching feeds MPR, average ticket, and financing close %. L3 Ch5 RC&D triage feeds recall %. L3 Ch6 Lesson 1 marketing attribution feeds GLSA ROAS. L3 Ch7 advanced workflows feed the wins narrative. L3 Ch8 prompt discipline, structured output, weekly audit, and documentation feed the recap's quality. The recap is the cross-chapter integration artifact.

The recap also closes the L3-to-L4 handoff. The L4 owner runs the Daily 8-Minute AI Routine and the Weekly 25-Minute Review. The Friday recap is the input artifact to the L4 owner's weekly 25-minute review — read Friday at 3:51 p.m., revisited Sunday or Monday morning for the deeper review. Without the recap, the L4 owner pulls numbers from five dashboards and synthesizes from scratch.

The Friday Recap and the Shops That Survive AI

The Friday Owner Recap is the highest-leverage written artifact the L3 service manager produces. 21 minutes per week to build; 4 minutes for the owner to read; 52 weeks of compounding operating discipline. The shops that run it have owners who walk into Monday morning knowing what moved, what did not, and what they have to decide. The shops that skip it have owners who pull numbers from five dashboards every Monday at 7 a.m. and rebuild the operating signal from scratch.

Lock the seven numbers. Hold the three-wins-two-problems-one-ask structure. Run the 21-minute Friday build. Deliver to the owner by 3:51 p.m. Edit for voice, accuracy, and one-page constraint. Watch the four failure modes. Compound the 52-document archive. The Friday Owner Recap is L3's weekly operating system made readable.

Key Takeaways

• One page, seven numbers, three wins, two problems, one ask — The owner's 4-minute Friday read. Format is reductive on purpose; the owner's bandwidth is the constraint, not the data availability.
• The seven numbers, named — Booking %, MPR, financing close %, recall %, RPT, average ticket, GLSA ROAS. Each maps to a different lever and a different Monday morning owner decision.
• The AI assembly chain — Six data sources (FSM platform, lender portals, CallRail, GLSA dashboard, Hatch, NiceJob/Podium/Birdeye) pulled, normalized, deltas computed, recap drafted. Cuts L3 service manager build time from 3 hours to 21 minutes.
• The 21-minute Friday build — AI pull and draft (7 min), service manager edit (8 min), marketing manager edit (4 min), format and deliver (2 min). 3:30 p.m. start, 3:51 p.m. owner has the recap.
• The 4-minute owner read — 7 numbers in 90 seconds, 3 wins in 30 seconds, 2 problems in 60 seconds, 1 ask in 30 seconds. Visual hierarchy tuned for scan-read; format deviation kills owner engagement by week six.
• Three wins discipline — Each win is 12-18 words: tech, metric movement, action implied. Forces ranking, not listing. Primes the owner to invest in what is working.
• Two problems discipline — Each problem is 18-22 words: metric, cause hypothesis, proposed action. Forces ranking. Primes the owner to fix by next Friday.
• One ask discipline — One decision per week, bolded, 20-25 words, imperative form ("approve," "decide," "confirm," "authorize," "review and respond"). The single decision the owner has to make this week.
• Four failure modes — Scope creep (recap exceeds one page), vague asks (no decision-requesting verb), AI voice drift (chatbot tells survive edits), number drift (data source pull misses update). Each has a detection trigger and a fix.
• Quarterly evolution — Q1 format holds, Q2 seasonal commentary, Q3 YoY deltas, Q4 quarterly summary feeds L3 governance memo. PE-readable signal compounds across the 52-document archive.
• Integration with the L3 stack — Pulls from Ch2-Ch7, audited by Ch8, hands off to L4 owner's weekly 25-minute review. The cross-chapter integration artifact.
• Discipline is small, leverage is large — 21 minutes per week, 18 hours per year. The L3 service manager's highest-leverage written communication; the owner's Monday morning operates on Friday afternoon's discipline.