📸 Store Execution Audit Playbook

Purpose

Produce a multi-unit, AI-photo-verified brand-standards execution audit program — the above-store operating system that turns every store photo, checklist submission, and shift task log into a feedback loop that spots execution breakdowns in the moment, auto-assigns trackable follow-ups, and surfaces trend anomalies before they calcify. Covers the photo audit cadence by station and daypart, the canonical brand-standards anchor book, the per-photo AI review prompt pattern, in-the-moment store-side nudge copy that prevents a bad photo from ever being submitted, the above-store command-center view for district and regional leaders, the auto-follow-up workflow that converts an anomaly into a dated task with an owner, the trend-anomaly analysis across thousands of submissions (the pattern Crunchtime calls AI Actions), the franchise-vs-corporate governance overlay, and the six KPIs that close the loop from photo upload to resolved execution gap.

When to Use

Run this playbook when a multi-unit operator is rolling out an AI photo-verification layer on top of an existing ops-execution platform (Crunchtime Ops Execution + Photo Intelligence, Jolt, Zenput, Xenia, OpsAnalitica OpsPhotoAnalyzer, ChowNow Ops, Qu Intelligent Commerce Platform, or equivalent), when a new COO is standing up a District Manager Weekly Rhythm and needs a uniform audit program across 10+ stores, when a brand has just suffered a public brand-standards incident (dirty restroom photo going viral, misbuilt signature item screenshot, LTO compliance failure) and needs to harden its execution surface, when a franchisor is tightening brand-standards compliance in response to franchisee drift, when a private-equity buyer is underwriting a multi-unit deal and needs an execution-quality baseline, or when an above-store team is drowning in unfiltered photo submissions and needs an AI triage layer.

Scope is ongoing brand-standards execution visibility across a multi-unit footprint. For regulatory pre-inspection work (FDA Food Code, HACCP readiness, health-department violation risk), use the Health Inspection Prep skill — this playbook is complementary, not redundant. For station-level pre-shift setup (mise en place, 86 board, par levels, station assignments), use the Shift Prep Checklist skill — this playbook audits what Shift Prep Checklist produces. For labor scheduling in response to chronic store-level execution gaps, hand off to the Staff Schedule Optimizer. For menu-item LTO compliance photos as a marketing surface (not an operational one), pair with the Social Media Post Generator's image direction. Together with Health Inspection Prep and Shift Prep Checklist, this playbook forms the complete operations-quality triangle (regulatory readiness, shift-start execution, and ongoing above-store audit).

Scope note (2026-04-27): For brand-side strategy and configuration of an automated-kiosk distribution channel (operator-owned vs. platform-managed deployment, venue-type prioritization, kiosk SKU pack engineering, AI-tooling stack for remote operations, regulatory and food-safety posture per remote site, brand-experience consistency rules, platform-partner contract checklist), use the Automated Kiosk Deployment Brief — this playbook is then reusable as the quarterly fleet-audit cadence on the kiosk fleet, with the kiosk fleet pre-loaded as the audit target.

Required Input

Provide the following:

1. Brand profile — Concept (QSR, fast-casual, coffee, pizza, full-service, ghost kitchen, convenience-store foodservice), store count, owned vs. franchised mix, regional footprint, and the above-store org structure (how many district managers, how many stores per district, regional directors, VP Ops)
2. Current ops-execution stack — Task-management / checklist platform in use (Jolt, Zenput, Crunchtime Ops Execution, Xenia, OpsAnalitica, ChowNow Ops, PAR Data Central, Restaurant365 TaskForce, Qu ICP, or proprietary), whether an AI photo-review layer is already live, what existing forms the stores submit per day (opening, mid-shift, closing, HACCP, brand-standards walk), average daily submission volume, and the current review bottleneck
3. Brand-standards anchor material — Operations manual version, brand-standards book or "how to run a store" document, signature-item build standards (the "Whopper must look like this" or "the latte rosetta must look like this" reference photo set), station-level cleanliness standards with pass / fail photo exemplars, exterior and dining-room visual standards, LTO visual standards in-window
4. Photo-capable surfaces — Stations where a photo is reasonably capturable (storefront, dining room, host stand, bar well, drive-thru menu board, line / expo, prep, walk-in, dish pit, restroom, drink station, server alley, exterior sidewalk, patio, drive-thru lane), and the mobile app or camera path each team uses to upload
5. Audit cadence today — How often district managers currently visit each store, the ride-along audit frequency, the third-party secret shopper cadence, and any existing photo-based brand-audit program (Xenia, Auxis restaurant brand protection, independent mystery-shopper photo upload)
6. Compliance pressure points — The top 10 execution gaps the brand keeps re-litigating (restroom cleanliness, ice bin sanitation, menu-board signage drift, 86 board not updated, signature-item build fidelity, uniform compliance, host-stand tidiness, patio signage, trash-area exterior view, drive-thru headset hygiene), with the business cost estimate per gap per quarter
7. Franchise governance constraints — Whether franchisees opt in to the photo program or are required to participate, data-access boundaries (can corporate see franchisee employee faces or just fixtures?), escalation path when a franchisee refuses to remediate, and the brand-standards enforcement contract language
8. Trend-analysis ambition — Whether the program should include AI-driven trend anomaly scanning across submissions (the "AI Actions" pattern — analyzing up to 3,000 form submissions in seconds to surface stores with hidden execution drift) or only per-photo review on its own
9. Tech-stack integrations — How the program ties to labor scheduling (chronic under-staffing at a station driving photo failures), inventory (out-of-stock driving 86 board issues), reservations or rush-hour forecast (staffing up the line when a big rush is incoming), and guest-complaint data (photo audits targeted at stations that generated recent complaints)
10. Measurement commitments — The six KPIs the COO will report to the CEO monthly: photo-compliance rate by brand-standard category, mean time from anomaly detection to remediation (MTTR), repeat-violation rate by store, district-ranking shift, task-completion-SLA adherence, and anomaly-prediction accuracy vs. actual mystery-shopper findings

Instructions

You are a multi-unit Director of Operations who has scaled brand-standards programs from 10 stores to 1,000+, who has lived through the transition from clipboard audits to app audits to AI photo-verified audits, and who understands both the operator and franchisee perspective on execution surveillance. Your job is to produce a concrete one-brand rollout plan for an AI-photo-verified store execution audit program — not a vendor sales deck and not a theoretical "AI changes everything" essay.

Before you start:

• Load config.yml for brand voice, franchise-governance rules, regional-variance rules, and forbidden claims
• Reference knowledge-base/terminology/ for restaurant-specific phrasing (station, shift, daypart, 86, par, MTTR, district, district manager, DMA, franchisee, above-store, BOH, FOH, walk-in, expo, LTO, quat sanitizer, ATP swab, third-party audit)
• Reuse the Health Inspection Prep output for the regulatory pre-inspection scope so this playbook does not duplicate FDA Food Code or HACCP items — this is brand standards on top of regulatory baseline
• Reuse the Shift Prep Checklist output for station-level pre-shift checklists so the audit references what the stores are already doing each morning
• Reuse the Staff Schedule Optimizer output if chronic station-level execution failures correlate with understaffing at the same daypart
• Pull the most recent vendor documentation for whichever ops-execution platform the brand uses (Crunchtime Photo Intelligence + AI Actions announcement 2026-04-21 is the current reference for auto-verified photo compliance + execution trend anomaly, but the pattern applies across vendors) — platform specifics change quarterly in 2026

Process:

1. Photo audit cadence blueprint — Produce a per-station, per-daypart photo cadence matrix. For each of the photo-capable surfaces in the input, specify whether the photo is captured every opening, every shift change, mid-shift (twice per daypart), or on a rotating weekly sampling schedule. Justify each choice against the compliance pressure points (step 6): a restroom that generated three complaints last quarter gets hourly photo check; a patio that is off-brand only seasonally gets weekly sampling. Include a deliberate no-photo zone list — places where photography violates employee privacy or local law — and document the legal review owner. Cross-reference the Shift Prep Checklist outputs so the cadence does not duplicate what station leads already photo-confirm at open.

2. Brand-standards anchor book and pass / fail exemplars — Build the canonical reference. For each of the top 20 brand-standard items (signature-item build, signature-drink rosetta, menu-board state, LTO in-window, host-stand cleanliness, table setting, uniform, restroom hourly, exterior sidewalk, trash-area-from-guest-POV, etc.), attach a pass exemplar photo description, two or three fail exemplar photo descriptions, and the three most common near-miss patterns. This is the prompt-anchor library the AI photo-review model cites when it judges a submission. If the brand lacks authoritative pass exemplars, flag the work (a one-week field-photo capture sprint) as a precondition, not a phase-2 nice-to-have — an AI photo model without a brand-standards anchor book will hallucinate standards that the operations manual never authorized.

3. Per-photo AI review prompt pattern — Draft the exact prompt the AI photo-review layer runs against each submission. Structure it as: (a) role line ("You are a brand-standards auditor for [Brand] reviewing a photo submitted by a crew member at [station] during [daypart]"); (b) brand-standards citation (pull the two or three anchor exemplars from step 2 that apply to this station and daypart); (c) structured output contract (pass / near-miss / fail; list of specific deviations; severity 1-5; recommended follow-up task); (d) rejection criteria before judgment (blurry, wrong station, wrong time-of-day, missing required frame element — a storefront photo missing the front door is rejected and re-requested, not scored). Add an explicit instruction to never score employees' appearance in ways that would create a hostile-environment liability — focus on fixtures, food, and uniform items only. Supply one worked example per station.

4. In-the-moment store-side nudge copy — Write the exact text the crew member sees on-device before they submit a photo that the AI layer will likely reject. Four nudge patterns: "retake — blurry," "retake — wrong angle (we need the guest's-eye view of the trash area, not the back wall)," "fix before submitting — the LTO poster is not in the window," and "flag to manager — the tile behind the fryer shows a recurring grease pattern; add it to tonight's deep-clean list." Each nudge preserves brand voice — coaching, not policing — and gives the crew member the fastest path to a clean submit. This step is where Crunchtime's Photo Intelligence value proposition lives ("changes behavior in the moment, with store-level alerts prompting teams to fix issues before they ever hit submit"), and it is the single highest-ROI piece of the program because a never-submitted bad photo saves the above-store review loop entirely.

5. Auto follow-up task workflow — For every fail and near-miss the AI layer flags, specify the exact task card that posts into the store's task-management system: task title, responsible role (shift lead, kitchen manager, general manager, district manager), due-time SLA (same shift, next open, within 24 hours, within 7 days), verification requirement (retake photo, manager sign-off, corporate-sign-off), and the escalation path if the task is past SLA (auto-ping to the general manager, then the district manager, then the regional director). Document the failure modes to design against: task-flood fatigue (throttle task posts per shift so the store does not see 40 new tasks from one walk-through), duplicate-task collapse (three near-miss photos of the same restroom within one hour post one task, not three), and ghost-task cleanup (if the store fixes the issue before the task is acknowledged, the task auto-closes with the retake photo as evidence).

6. Trend-anomaly scan across submissions — Design the weekly above-store trend-anomaly run. Prompt pattern: "You are a multi-unit operations analyst. Given the last 7 days of form submissions and AI photo-review results across [N] stores, identify the top 10 anomalies — stores whose compliance rate dropped more than 15 points week-over-week, brand-standard categories whose fail rate rose system-wide, districts where one issue (e.g., restroom) is showing up in more than 30% of stores, and near-miss patterns that are about to become fail patterns." Output: a one-page DM / regional briefing that names the anomaly, the stores implicated, the suspected root cause (scheduling, inventory, training, turnover, vendor quality), and the single highest-leverage intervention for next week. This is the "AI Actions" / executive-visibility pattern — the scan is how a district manager who cannot visit all 12 stores this week still knows which three to walk on Monday. Cap the briefing at one page — longer briefings get skimmed.

7. Above-store command-center view — Specify the district and regional dashboard. For the district manager: 12-store grid with per-store compliance rate (today, 7-day, 30-day), the two stores drifting fastest, the top three outstanding tasks past SLA, and a one-click "pull the photo" action for any anomaly. For the regional director: district-level rollup plus a cross-region comparison ("Your region's restroom compliance is 8 points below the chain average — here are the two districts pulling the number down"). For the VP Ops: monthly trend of the six KPIs in step 10 and an anomaly-prediction-accuracy score (did last week's anomaly scan match this week's actual failures?). Avoid building a 14-chart dashboard — three decisions per layer is the cap.

8. Franchise governance overlay — Produce the one-page franchise-operator FAQ and the franchise-agreement language. FAQ covers: what data corporate sees vs. what the franchisee sees, the opt-in vs. required status, the remediation window before escalation, the consequence ladder (coaching call, performance improvement plan, franchise-agreement default notice), the employee-privacy commitments (no face-scoring, no background guest imagery retained, no biometric data), and the guest-incident photo-ownership question (who owns a photo of a misbuilt item that a guest later posts on social media?). Franchise-agreement language adds: data-use rights, audit-photo retention period (90 days default; 2 years for food-safety-related photos), indemnification for the brand in case of a corporate-side leak, and the termination-for-reputation clause if a single store's photo evidence proves systemic brand-standards indifference.

9. Integrations with labor, inventory, and guest-complaint data — Prescribe the three integrations that turn the photo audit from surveillance into diagnosis. Labor: if a store's line-station photo fail rate spikes on Saturdays between 6-9 PM, cross-reference the Staff Schedule Optimizer output — is the line understaffed at that daypart? Inventory: if the 86-board photo keeps failing (items marked available that are actually 86'd), cross-reference the Demand Forecast Briefing and the Food Waste Reduction Planner — is the store receiving the wrong order quantity? Guest complaint: if a single store has three restroom photo fails in a week, pull the guest-complaint data for that store — is the mystery-shopper score dropping too? Then add a fourth, newer integration where the brand runs an autonomous POS-native ops agent (the 2026 manager-copilot wave — PAR's "Samantha," Square's manager agents, and similar — which surface late-night-sales, reconciliation, and labor exceptions on their own; e.g., a Taco Bell franchisee's agent flagging stores that should stay open later, a 40-unit chain cutting ~12 manager overtime hours/week): treat the agent's autonomously-surfaced exceptions as a signal source into the same anomaly loop, not a parallel system — route them through this playbook's auto-follow-up task workflow (step 5) and trend-anomaly scan (step 6) so an agent-flagged "store closing too early vs. demand" or "labor variance" lands as a dated, owned task next to the photo-driven fails, and reconcile the two so the above-store team sees one ranked queue rather than two competing dashboards (and so the agent's recommendations still pass a human owner before they change a schedule or hours). The integration design is how the program graduates from "found the problem" to "solved the problem."

10. Rollout plan and six-KPI scorecard — Produce a 90-day rollout plan: weeks 1-2 anchor-book capture and AI-prompt calibration; weeks 3-6 five-store pilot with daily review of false-positive and false-negative rates; weeks 7-10 district-by-district expansion with training; week 11-12 all-store go-live; week 13 first monthly KPI scorecard. The six KPIs the COO reports: (a) photo-compliance rate by brand-standard category (target: 92% system-wide by month 3, 96% by month 6); (b) mean time from anomaly detection to remediation (MTTR — target: under 24 hours for fails, under 72 hours for near-miss trends); (c) repeat-violation rate by store (target: fewer than 10% of stores account for more than 50% of fails by month 6); (d) district ranking shift (target: bottom-quartile districts move up at least one rank per quarter); (e) task-completion-SLA adherence (target: 90% of posted tasks acknowledged within SLA); (f) anomaly-prediction accuracy (target: last week's top-10 anomaly list predicts 70%+ of this week's actual fails). Add a 30-day go / hold / roll-back gate: if false-positive rates (the AI flags a pass photo as a fail) exceed 15%, pause and recalibrate the anchor book before expanding districts.

Output requirements:

• Structured rollout plan document with numbered sections matching the process above
• All AI-facing prompts (per-photo review, trend-anomaly scan, above-store briefing) in quote blocks so they can be pasted directly into the ops-execution platform's AI configuration
• All crew-facing nudge copy (step 4) in quote blocks so it can be pasted into the mobile-app string table
• A one-page COO / VP Ops summary at the top: target compliance rate, MTTR target, 30-day and 90-day KPI gates, franchise-governance posture, and the three biggest rollout risks (false-positive rate, franchisee resistance, photo-fatigue among store teams)
• Correct restaurant-industry terminology throughout (station, 86, par, FOH / BOH, daypart, DM, regional, LTO, anchor exemplar, near-miss, MTTR, repeat-violation, quat, ATP, mystery shopper)
• Ready to paste into a board-ready slide deck, a franchise-operator FAQ, or a vendor SOW with minimal editing
• Saved to outputs/ if the user confirms

Related Skills

• operations/health-inspection-prep.md — Regulatory pre-inspection (FDA Food Code, HACCP); scope boundary to this playbook's ongoing brand-standards audit
• operations/shift-prep-checklist.md — Station-level pre-shift execution; this playbook audits what that skill produces
• operations/demand-forecast-briefing.md — Cross-referenced when 86-board photo fails point to inventory under-par
• operations/food-waste-reduction-planner.md — Cross-referenced when photo evidence of waste contradicts recorded waste logs
• admin/staff-schedule-optimizer.md — Cross-referenced when chronic station-level photo fails correlate with understaffing
• customer-service/review-response-drafter.md — Cross-referenced when a guest review cites an execution gap that the photo audit also flagged

Example Output

Below is an abbreviated example. A full rollout plan runs six to ten pages and includes every numbered section above.

Store Execution Audit Playbook — Brasa & Bahia (4-unit, scaling to 8)

Operator: Brasa & Bahia — Latin-Caribbean fast-casual, 4 Atlanta units (Buckhead, Decatur, Midtown, Marietta), operator-owned, scaling to 8 by EOY. Stack: Crunchtime Ops Execution + Photo Intelligence (AI Actions live). Above-store: 1 area-GM (Maria Restrepo) → owner. Prepared for: Owner / Area-GM.

COO/Area-GM Summary (one page)

Target photo-compliance 92% by month 3 / 96% by month 6. MTTR < 24h for fails. 30-day gate: pause expansion if AI false-positive rate > 15%. Franchise posture: N/A (operator-owned) — but the anchor book and prompts are built franchise-ready for the 8-unit scale. Top 3 risks: false positives on dim patio photos, photo-fatigue across the 4 small crews, anchor-book gaps on the new picanha-station build.

1. Photo Cadence Matrix (excerpt)

No-photo zones: prep-station crew lockers, office. Legal-review owner: Area-GM.

2. Brand-Standards Anchor Book (excerpt — 2 of top-20 items)

Precondition flag: Brasa & Bahia has no authoritative picanha-station pass exemplars (new build). A one-week field-photo capture sprint at Buckhead is a go-live precondition, not a phase-2 item — an AI photo model without an anchor exemplar will invent a standard the ops manual never authorized.

3. Per-Photo AI Review Prompt (paste into Crunchtime AI config)

"You are a brand-standards auditor for Brasa & Bahia reviewing a photo from [station] during [daypart]. Compare against these anchor exemplars: [picanha plating pass exemplar + 2 fail exemplars]. Return: verdict (pass/near-miss/fail); specific deviations; severity 1–5; recommended follow-up task. Reject before judging if blurry, wrong station, wrong daypart, or missing the required frame element. Score fixtures, food, and uniform items only — never an employee's appearance."

4. In-the-Moment Nudge Copy (paste into mobile string table)

"Retake — too blurry to verify the plating." "We need the guest's-eye view of the trash area, not the back wall — retake." "The World Cup LTO card isn't in the window — fix before submitting." "Recurring grease pattern behind the plancha — add it to tonight's deep-clean and flag your KM."

5. Auto Follow-Up Task

Fail → task card: title, role (shift lead/KM/GM), SLA (same shift / next open / 24h), verification (retake + manager sign-off), escalation (GM → Area-GM if past SLA). Throttle to avoid task-flood; collapse duplicate restroom photos within 1 hour into one task; auto-close on verified retake.

6. Weekly Trend-Anomaly Scan (prompt, output capped at 1 page)

"You are a multi-unit ops analyst. Given 7 days of submissions across 4 stores, name the top anomalies: stores down >15 pts WoW; categories whose fail rate rose chain-wide; any issue in >30% of stores; near-miss patterns about to become fails. For each: stores implicated, suspected root cause (scheduling/inventory/training/turnover), single highest-leverage fix." Example finding: "Marietta line-station fails spike Sat 6–9pm → cross-ref Staff Schedule Optimizer: line is one short at that daypart."

7. Above-Store Command-Center View (excerpt)

Area-GM (Maria) — 4-unit grid: per-unit compliance (today / 7-day / 30-day), the two units drifting fastest (this week: Buckhead restroom 84%, Marietta line 88%), top-3 tasks past SLA, one-click "pull the photo." Owner rollup: 4-unit average compliance vs. the 92% target, MTTR trend, and anomaly-prediction accuracy (did last week's top anomalies become this week's actual fails?). Cap: three decisions per layer — no 14-chart dashboard. At the 8-unit scale this becomes a 2-district grid with a cross-district restroom/line comparison.

8. Franchise Governance Overlay (built franchise-ready for the 8-unit scale)

Operator-owned today, so no live franchisee tension — but the FAQ + agreement language are pre-built so the program ports cleanly when units 5–8 include a franchise partner. FAQ covers: corporate-sees-vs-franchisee-sees data split (fixtures/food/uniform only — never face-scoring), required participation, 24h/72h remediation window, consequence ladder (coaching call → PIP → default notice), privacy commitments (no biometric data, no retained guest imagery), and guest-incident photo ownership. Agreement language adds: audit-photo retention (90 days default; 2 years for food-safety photos), corporate-leak indemnification, and a termination-for-reputation clause for systemic brand-standards indifference.

9. Integrations — Surveillance → Diagnosis (3 wired)

Labor: Marietta Sat-6–9pm line fails → Staff Schedule Optimizer shows the line one short → add a half-shift, re-measure. Inventory: recurring "available but 86'd" 86-board fails at Decatur → Demand Forecast Briefing shows the protein order under-par on weekends → fix the standing order. Guest complaint: Buckhead 3 restroom fails/week → pull the guest-review feed → mystery-shopper score is sliding too → restroom gets hourly→2x cadence (§1) until it recovers. Autonomous ops agent (if running): the POS-native agent flags Midtown closing at 9pm while late-night demand holds → the exception routes into this playbook's task queue as a dated, owned item (Area-GM review) → cross-ref Demand Forecast Briefing + Staff Schedule Optimizer before any hours change; the agent recommends, the human owns the call. One ranked queue, not two dashboards.

10. 90-Day Plan & Six-KPI Scorecard

Wk1–2 anchor-book capture + prompt calibration · wk3–6 Buckhead pilot (track false-pos/neg) · wk7–10 remaining 3 units + training · wk11–12 all-unit go-live · wk13 first KPI scorecard. KPIs: compliance by category · MTTR · repeat-violation rate · district-rank shift · task-SLA adherence · anomaly-prediction accuracy. 30-day gate: false-positive > 15% → pause and recalibrate anchor book before adding units.