Start Here: What's keeping you up about quality right now?

• Briefly, what event triggered this conversation today (e.g., scrap spike, audit finding, recall, recurring customer complaint)? Options: Scrap rate exceeded target, Customer audit non-conformance, Product recall, Unexpected defect spike, Proactive improvement initiative, Other
• How long has the issue been present and how did you first notice it? Options: <1 month, 1–3 months, 3–6 months, 6–12 months, Over a year
• Who on your team is most impacted day-to-day by this problem? Options: Quality Engineer(s), Process Engineer(s), Production/Line Lead, Continuous Improvement Manager, Plant Manager, Other
• If nothing changes in the next quarter, what tangible impact do you expect (scrap %, throughput loss, customer penalties, etc.)?
• What's one thing you hope a new quality analytics tool would immediately prove or resolve for you?

Are we confusing signal with symptom?

• When you see a defect spike today, how confident are you that you can identify the true root cause within a single shift? Options: Very confident, Somewhat confident, Occasionally, Rarely, Never
• Tell me about a recent investigation where the team thought they found the cause but later discovered something else—what happened and what was missed?
• Which data sources do you currently use to investigate defects (pick all that apply)? Options: MES logs, Vision inspection output, Operator defect logs / spreadsheets, OEE / downtime data, Environmental sensor data, Lab/test results, Other
• How often do you find correlations that feel spurious or driven by noise rather than a real process issue? Options: Frequently, Sometimes, Rarely, Never, Not sure
• Describe a known root-cause relationship you've already validated on your line (e.g., material lot → scratch rate). How did you confirm it?

Who holds the keys to making this change?

• If we propose a pilot, who are the decision-makers that must be aligned to proceed? Options: Quality Engineer, CI Manager, Plant Manager, IT/MES Lead, Operations Manager, Procurement/Finance, Other
• Which stakeholder will sign off on pilot acceptance (technical and commercial)? Options: Quality lead, Plant manager, CI lead, IT/MES owner, Cross-functional committee, Other
• What budget envelope or approval process governs purchases in this category? Options: Operational improvement budget, Capital expenditure, Departmental discretionary, Requires executive approval, Undecided / unknown
• How do stakeholders prefer to evaluate success—hard financial ROI, defect-reduction % targets, audit-readiness, or operator adoption? Options: ROI / cost-savings, Defect % reduction, Time-to-root cause, Audit compliance, Operator ease-of-use, Other
• Has your organization run external pilots before? If yes, what made them feel like a success or failure?

If we could rewind: what would perfect look like for your line?

• Imagine the pilot succeeds—what three specific outcomes would convince leadership to roll this out plant-wide?
• Which success signal matters most to you personally: faster RCA, measurable scrap reduction, fewer customer complaints, or operator time saved? Options: Faster root-cause analysis, Reduced scrap/waste, Fewer customer complaints, Improved audit readiness, Reduced operator/admin time, Other
• What magnitude of scrap reduction or cost saving would you need to see in the pilot to justify expansion? Options: <10% reduction, 10–25% reduction, 25–50% reduction, >50% reduction, Unsure / need to model
• How would a measurable win change your team's daily experience or priorities?
• If adoption were perfect, what would a day on the line look like differently compared to today?

What's missing from your data story that keeps you blind?

• Can you provide the specific data sources and access level we can get for a pilot (e.g., 90 days of MES records, vision system exports, SPC logs)? Options: Full database access (read), Flat-file extracts (CSV), API access, No access / manual export only, Unsure yet
• Which of the following are available for the pilot line today? Options: Timestamped defect records, Part serial or lot IDs, Machine/process parameters, Vision inspection images or metadata, Changeover logs, Environmental sensor logs, None of the above
• Have you historically struggled with data quality—missing timestamps, mis-labeled defects, or inconsistent part IDs? How long has that been true? Options: Not at all, Occasionally, Frequently, Systemic / always
• If data gaps exist, who would be the owner responsible for filling them or creating a workaround during a pilot? Options: MES/IT, Quality team, Process engineering, Line maintenance, Vendor/3rd party, No owner assigned
• Describe any export, security, or privacy constraints that might limit sharing vision images, sensor data, or MES records.

Can operators make this real in under 10 seconds?

• How are defects currently logged at the point of occurrence (paper, MES, tablet, verbal, spreadsheet)? Options: MES line interface, Tablet / handheld app, Paper log, Spreadsheets after the shift, Operator notes / verbal, Other
• On average, how long does it take an operator to log a defect today, and how disruptive is that to cycle time? Options: <5 seconds / negligible, 5–10 seconds / manageable, 10–30 seconds / noticeable, >30 seconds / disruptive, Not measured
• We often see defect taxonomies that are either too granular or too vague—what level of detail do operators currently use (broad categories, specific defect codes, image-linked labels)? Options: Broad categories (e.g., scratch, dent), Moderate (code + short note), Very granular (many specific codes), Image-linked labels, No consistent taxonomy
• If we needed to refine your taxonomy for the pilot, who would be the best SME to design and approve it? Options: Quality Engineer, Process Engineer, Line Supervisor, Production Operator rep, Cross-functional team
• What resistance or concerns do operators typically express about new data-entry steps or inspection workflows? Options: Adds to cycle time, Too complicated, Not trained, Won't be followed, Privacy/monitoring concerns, Other
• Share an example of a taxonomy or logging change you've tried before—what worked and what didn't?

What would make leadership reach for the contract right now?

• What is your ideal pilot timeline and decision window (from kickoff to go/no-go)? Options: 2–4 weeks, 4–6 weeks, 6–8 weeks, Longer than 8 weeks, Undecided
• Which acceptance criteria will you use to judge pilot success (pick up to three)? Options: Detect known root-cause correlations, Reduce scrap by X%, Operator logging <10s, Integrates with MES/vision, Actionable alerts with low false positives, ROI payback within Y months
• What ROI horizon does your leadership expect for quality improvement investments? Options: <6 months, 6–12 months, 12–24 months, Longer / strategic
• Who will be responsible for tuning alerts and SPC thresholds during the pilot? Options: Quality Engineer, CI Manager, Vendor with support, Joint team, Not decided
• If the pilot identifies a complex root cause involving multiple systems, how willing is leadership to fund the follow-up remediation work? Options: Very willing, Somewhat willing, Only if ROI is clear, Unlikely without additional approval, Unsure

Red flags, past pilots, and how we mitigate them

• Have you run a pilot that failed to deliver—what were the top three reasons it didn't work?
• Which of the following risks concern you most for a pilot with our platform? Options: Data access gaps, Taxonomy too complex, Operator non-adoption, Alert fatigue / false positives, Integration with MES, Security/compliance
• What contingency or escalation process do you expect if the pilot uncovers data integrity issues that block analysis? Options: MES/IT remediation, Manual data collection workaround, Narrower scope, Pause pilot, Other
• How do you prefer risk to be shared during a pilot—fixed-scope trial, joint success metrics, or vendor-backed remediation commitments? Options: Fixed-scope trial, Shared success milestones, Vendor remedy commitments, Flexible approach
• What level of vendor transparency and documentation do you expect during pilot handovers and final acceptance? Options: Detailed technical docs & runbook, Executive summary + highlights, Hands-on training + recordings, All of the above

Next steps: deciding, scheduling, and who we loop in

• If we proposed a 4–6 week pilot, when would you like to start? Options: Immediately (next 2 weeks), Within 1 month, 1–2 months, 3+ months, Undecided
• Who should be included in the kickoff meeting from your side (names and roles, or roles if names unknown)?
• What would make you say 'no' to starting a pilot in the near term? Options: No budget, Insufficient data access, Stakeholder objections, Higher priority projects, Unsure of value, Other
• What's the single most important question we can answer for you between now and kickoff to make your life easier?
• Finally, please list any compliance, security, or procurement constraints we should factor into the proposal (file sharing rules, NDA needs, approved vendor lists, etc.).

• Pilot Kickoff & Problem Framing
• Data & MES Integration Workshop
• Defect Taxonomy & Operator Workflow Workshop
• Pilot Execution Plan & Acceptance Criteria
• Pilot Validation & Handover Review

• Document risks and remediation gates that can pause or adjust the pilot if needed.
• Map taxonomy terms to MES codes and platform fields to enable joined analytics.
• Validate operator logging workflow and confirm median logging time meets the <10s target.
• Define initial SPC/alert thresholds to use during the pilot to avoid alert fatigue.
• Seller to publish the pilot taxonomy into the sandbox and map to MES codes.
• Customer to run operator timing trials with the prototype UI and report median/90th-percentile times.
• Both to finalize the taxonomy change-control process for the pilot (who can edit, how changes are approved).
• Agree an SPC tuning cadence for the pilot and assign owners for alert triage.
• Pilot runbook & weekly cadence
• Agree on an executable pilot runbook with dates and responsibilities.
• Define objective acceptance criteria that include detection of known root-cause correlations.
• Set measurable ROI targets and define how they will be calculated and reported.
• Introductions & objectives
• Seller to schedule and perform the 90-day data load and initial model runs.
• Customer to provide the canonical list of known correlations and the acceptance-check owner.
• Both to finalize and sign the Pilot Acceptance Checklist containing pass/fail criteria.
• Establish weekly pilot check-in meetings and assign a pilot coordinator.
• Pilot execution summary
• Demonstrate with evidence that the platform surfaced the customer’s known root-cause correlations.
• Obtain explicit validation (accept/reject) from the customer for each known correlation.
• Verify MES integration worked end-to-end for the pilot and operator logging meets performance targets.
• Agree on acceptance outcome and next-steps (remediation plan or full-rollout proposal).
• Seller to deliver a Pilot Validation Report with screenshots, data extracts, and correlation evidence for audit.
• Customer to provide formal pilot acceptance decision and any remediation requests tied to the acceptance checklist.
• If accepted, seller to prepare a full-rollout proposal including timeline, effort estimate, and ROI projection.
• If gaps remain, both parties to agree remediation tasks, owners, and a revalidation date.
• Agree on a single, crisp current-state sentence that frames the pilot.
• Quantify the business consequence of the problem so urgency is explicit.
• Capture 2–5 known root-cause examples that must be reproduced by the pilot.
• Define the future-state outcome and 3 success signals the pilot will prove.
• Lock pilot scope, timeline (4–6 weeks), and stakeholder RACI for execution.
• Customer to provide the one-sentence current-state and documented consequence metrics (scrap %, $ impact, audit findings).
• Customer to deliver a list of validated 'known correlations' (examples with timestamps and context) to be used for validation.
• Seller to circulate pilot charter (scope, timeline, roles) within 24–48 hours for sign-off.
• Schedule Data & MES Integration Workshop and Taxonomy Workshop as follow-ups.
• Prework verification
• Confirm 90 days of required data sources are accessible in a format the platform can ingest.
• Finalize field mappings and document any transformations required.
• Establish MES integration approach and confirm access/owners for workstream execution.
• Execute a test ingest and produce a short data-quality report with remediation actions.
• Agree contingencies and timelines for resolving missing or low-quality data.
• Customer to provide required credentials, sample exports, and a data dictionary for all sources.
• Seller to complete test ingest, run data-quality checks, and deliver a DQ report with prioritized fixes.
• Customer IT/MES owner to grant temporary access to a non-production endpoint or provide scheduled exports.
• Both parties to finalize a Data Acceptance Checklist that triggers the full 90-day load.
• Review current capture and terminology
• Finalize a pilot taxonomy that is analytically useful and operable by shop-floor staff.
• Roles, responsibilities & RACI
• Proof: Detected correlations demo
• Single-sentence Current State
• One-sentence data readiness state
• One-sentence taxonomy goal
• Customer validation of known examples
• Consequence Quantification
• Draft pilot taxonomy & level-of-detail
• Field mapping & schema alignment
• 90-day data load plan and verification
• MES integration & operator logging metrics
• Map taxonomy to MES codes & data model
• MES integration method & auth

• Ingest 90 Days of Production and Defect Data
• Configure Defect Taxonomy and Operator Labels
• Integrate MES via API/OPC Connectors
• Ingest Machine-Vision Images and Metadata
• Stream Process Parameters and Sensor Data
• Deploy Automated SPC Rule Engine
• Train and Deploy Image-Based Defect Classifier
• Run Root-Cause Correlation Engine and Reports
• Deploy Pareto and Trend Dashboards
• Activate Real-Time Shop-Floor Defect Entry UI
• Configure Alerts, Thresholds, and Suppression Logic
• Link Material Lots and Traceability Records
• Enable Export to ERP/MES and CSV Reports

Ingest 90 Days of Production and Defect Data

• Do you have a contiguous 90-day export available for the target line? Options: Yes, No, Partial (less than 90 days)
• Which formats will the historical data be provided in? Options: CSV/Excel, SQL database export, MES-native extract, Flat files (JSON/NDJSON), Other
• What key fields are present in your export (select all that apply)? Options: Timestamp, Part/Batch ID, Serial Number, Station/Operation ID, Defect Code/Label, Operator ID, Cycle Count, Shift
• Are timestamps synchronized with line equipment (single clock) or do records come from multiple systems with different clocks? Options: Single synchronized clock, Multiple clocks - consistent timezone, Multiple clocks - inconsistent
• What is the average volume of rows/records per day for the selected line? Options: Less than 1k, 1k-10k, 10k-100k, 100k+
• Are there known gaps, missing fields, or data-quality issues in the historical data we should expect? Options: None, Minor (occasional missing rows), Moderate (frequent nulls/fields), Severe (large time ranges missing)

Configure Defect Taxonomy and Operator Labels

• Do you currently use a defect taxonomy? If yes, how many unique defect codes/labels exist? Options: No taxonomy, Yes - 1-10 labels, Yes - 11-50 labels, Yes - 51+
• Who will own taxonomy decisions (role/title)?
• Do operators currently log defects on the line? If so, what is the typical label granularity they can reliably select within <10s? Options: No operator logging, Coarse (1-5 choices), Moderate (6-15 choices), Fine-grained (16+ choices)
• Which taxonomy style do you prefer for the pilot? Options: Operator-friendly coarse labels, Detailed engineering-level labels, Hybrid (coarse at capture, mapped to detailed offline)
• Are there existing mapping tables between legacy codes (MES/ERP) and desired taxonomy? Options: Yes - ready, Yes - requires mapping work, No mapping exists
• List any bottlenecks or concerns about operator adoption or label consistency (e.g., language, shift changes, temporary rework codes).

Integrate MES via API/OPC Connectors

• Which MES(s) are in use on the target line(s)? Options: Siemens MES, Rockwell/FactoryTalk, SAP ME/MII, Dassault/Apriso, Custom/Other
• Do you have API access or OPC-UA endpoints available for the pilot line? Options: API endpoints available, OPC-UA available, Only database access, No programmatic access currently
• Who is the contact for MES access and firewall/VPN approvals (role/title)?
• Are there security or change-control policies that will restrict connector installation or data access windows? Options: No restrictions, Planned maintenance windows only, Strict change-control - requires approvals
• Which MES entities do you need synced (select all that apply)? Options: Work orders, Serial/lot traceability, Defect events, Tool IDs, Operator IDs
• Do you require read-only integration or bi-directional updates (e.g., writebacks of defect labels)? Options: Read-only, Bi-directional (writebacks)

Ingest Machine-Vision Images and Metadata

• Are machine-vision images captured and stored for the target line? Options: Yes - images stored centrally, Yes - images available on camera system, No images available
• What image capture frequency / file size / retention should we plan for? Options: Frame per part (high volume), Periodic sampling (e.g., 1 in N), Event-triggered only
• What image formats and metadata are available (e.g., JPEG/TIFF, camera timestamp, exposure, ROI)?
• How will we access images for ingestion (network share, API, direct camera stream)? Options: Network share/SFTP, Camera vendor API, Direct stream (RTSP), Other
• Are there privacy/IP constraints on shipping images to a cloud service or must processing stay on-prem? Options: Cloud OK, Hybrid (metadata cloud, images on-prem), On-prem only
• Estimate average images per day for the pilot line and peak throughput requirements.

Stream Process Parameters and Sensor Data

• Which process parameters/sensors should be streamed (e.g., temperature, pressure, spindle RPM, humidity)?
• What is the sampling frequency for key sensor streams? Options: Sub-second, 1s-10s, 10s-1min, 1min+
• Are sensor streams available via PLC/OPC, historian (PI/OSIsoft), or separate IIoT gateways? Options: PLC/OPC, Historian, IIoT gateway, Not currently available
• Do you require short-term buffering/edge-storage to handle network interruptions? Options: Yes, No, Unsure - consult required
• What retention window is required for raw sensor data during the pilot? Options: 7 days, 30 days, 90 days, Custom
• Identify any proprietary sensor protocols or custom drivers that may need development.

Deploy Automated SPC Rule Engine

• Which SPC rules do you currently use or want to enable (e.g., Shewhart, CUSUM, EWMA, zone rules)? Options: Shewhart (Xbar/R), CUSUM, EWMA, Zone/Western Electric, Custom rules
• Do you have control limits defined or should the system compute baseline limits from historical data? Options: Limits provided, Compute from historical 90 days, Mix - compute per product family
• What cadence do you want SPC evaluated (real-time per part, per subgroup, hourly)? Options: Per part (real-time), Per subgroup, Hourly/batch
• Who should receive SPC alerts and what escalation path should be used?
• Do you require suppression windows or feedback loops to reduce alert fatigue (e.g., mute alerts during known changeovers)? Options: Yes - suppression needed, No - all alerts
• Are there regulatory or audit requirements for SPC records retention and export? Options: Yes, No, Unsure

Train and Deploy Image-Based Defect Classifier

• Do you have labeled image datasets for each defect class and a baseline of non-defect images? Options: Yes - ample labeled data, Yes - limited labels, No labeled images
• Approximately how many labeled images per defect class are available? Options: <50, 50-200, 200-1,000, 1,000+
• What inference latency target is required for on-line classification (ms per image or max seconds per decision)?
• Will classification run on edge devices, on-prem servers, or cloud GPUs? Options: Edge device, On-prem server, Cloud GPU, Hybrid
• Do you require model explainability outputs (e.g., heatmaps, bounding boxes) for operator validation? Options: Yes - heatmaps/boxes, No - only label, Optional
• Are there seasonal/lot-based visual variations that may require per-lot or per-tool models? Options: Yes, No, Potentially - needs assessment

Run Root-Cause Correlation Engine and Reports

• Which root-cause relationships are highest priority to detect (material lot, tool wear, environment, operator, machine state)? Options: Material lot, Tool wear, Environment, Operator, Machine state, Other
• What statistical confidence or minimum sample size do you require before actionable correlations are surfaced? Options: High (p<0.01, large sample), Medium (p<0.05), Low - exploratory
• Do you want automated periodic correlation reports (daily/weekly) and ad-hoc investigator-driven queries? Options: Both automated and ad-hoc, Automated only, Ad-hoc only
• Are there known ground-truth correlations (provide examples) we should use to validate pilot effectiveness?
• Should correlation outputs include recommended corrective actions or just hypotheses for engineering review? Options: Include recommended actions, Hypotheses only, Both
• Which stakeholders should be subscribed to correlation alerts and reports (roles/titles)?

Deploy Pareto and Trend Dashboards

• Which KPIs should appear on the pilot dashboard (scrap rate, DPPM, defect counts by class, yield)? Options: Scrap rate, DPPM, Defect counts by class, Yield, Cycle time, Other
• Who are the primary dashboard users and what permission levels are needed? Options: Operators, Quality engineers, Plant managers, CI teams, Other
• What time windows and aggregation levels are required for trend analysis (real-time, hourly, shift, daily, weekly)? Options: Real-time, Hourly, Shift, Daily, Weekly
• Do you need exportable reports or scheduled PDF/CSV deliveries to stakeholders? Options: Yes - PDF/CSV, Yes - scheduled dashboard snapshot, No
• Are there custom visualizations required (e.g., stacked pareto by shift or heatmaps by machine)? Options: Yes, No
• Do dashboards need to be embedded into existing portals (MES/SharePoint) or accessed via the platform UI? Options: Embed in MES, Embed in SharePoint, Platform UI only, Multiple

Activate Real-Time Shop-Floor Defect Entry UI

• Will operators use tablets, station PCs, or integrated HMI screens to enter defects? Options: Tablet, Station PC, HMI screen, Mobile phone
• What is the target defect logging time per event (e.g., <10 seconds)? Options: <5s, <10s, <20s, No specific target
• Do you require multi-language support or icons to speed operator selection? Options: Yes - languages, Yes - icons, No
• Should the UI allow attaching images or auto-linking camera captures to the defect record? Options: Yes - attach images, Auto-link only, No
• Do you need kiosk-mode/locked-down UI to prevent accidental navigation during line operation? Options: Yes, No
• Which authentication method will operators use (badge scan, SSO, local PIN)? Options: Badge/Proximity, SSO (SAML/OAuth), Local PIN, Other

• Statement of Work (SOW)
• Pilot Agreement (4–6 week timeline)
• Commercial Terms & Payment Schedule
• Roles & Responsibilities (RACI)
• Data & Integration Commitment
• Taxonomy Governance & Alert Tuning Plan
• Acceptance Criteria & Success Metrics
• Risk Mitigation & Contingency Plan
• Change Order Agreement
• Security & Data Processing Agreement (DPA)
• Support & Escalation Plan
• Pilot Sign-Off & Handover
• Termination & Exit Terms

• Success Review — Outcomes vs Success Signals
• Lessons Learned Retrospective
• Enhancement Backlog & Roadmap Prioritization
• Operational Handover & Support Channel Setup
• ROI Review & Scale Decision (Executive Summary)

• Confirm training and documentation are scheduled and accessible to frontline operators.
• Schedule operator refresher training and a follow-up verification session.
• Inventory of requests and incidents
• Produce a prioritized backlog with clear must-haves for rollout and an agreed timeline.
• Assign owners and commitments for critical fixes before expansion.
• Establish a single source of truth for enhancement status and communication cadence.
• Publish prioritized backlog into CustomerNode or dev tracker with impact/effort scores and owners.
• Schedule recurring biweekly product syncs to report progress against the agreed items.
• Set up the shared enhancement channel and invite all stakeholders with defined governance rules.
• Runbook overview and ownership
• Operational team has a clear runbook and named owners for day-to-day monitoring and support.
• Establish and activate a shared channel with governance so issues and enhancement requests are tracked transparently.
• Current-state recap (1 sentence)
• Publish the final runbook and contact list to the shared channel and document repository.
• Create the shared channel, define triage rules, and onboard participants.
• Schedule operator and support-team training sessions and a 30-day follow-up health check.
• Executive one-sentence current state & consequence
• Validate realized and projected ROI and its alignment with plant financial targets.
• Secure executive approval and budget commitment for the agreed rollout option.
• Assign an executive sponsor and set a high-level rollout timeline.
• Deliver a one-page business case and secure sign-off from the plant manager/CFO for the selected rollout option.
• Create a high-level rollout plan with milestones, resource assignments, and a preliminary budget.
• Schedule the first quarterly business review (QBR) post-rollout to review KPI tracking and continuous improvement.
• Establish whether pilot met each acceptance criterion with supporting evidence.
• Quantify pilot impact versus ROI targets and surface any major gaps.
• Agree a clear next step (approve rollout, require remediation + re-test, or close pilot) and owners for follow-up.
• Produce an 'Acceptance Report' summarizing pass/partial/fail for each success signal and attach supporting exports.
• List unresolved gaps with owners, due dates, and success criteria for any required remediation.
• If approved, create initial rollout project charter (scope, timeline, budget) and schedule kickoff.
• Set scope and retro rules
• Document the top operational and technical lessons with evidence and root causes.
• Create a short, prioritized remediation list with owners and timelines.
• Agree changes to pilot->rollout process to avoid repeat issues.
• Open remediation tickets in the shared workspace for each prioritized issue, with acceptance criteria.
• Update the defect taxonomy guide to reflect agreed simplifications and operator examples.
• ROI and KPI results vs targets
• Escalation paths and SLAs
• What went well (data-driven)
• Consequence summary vs baseline
• Impact vs effort scoring
• Shared channel creation & governance
• Success signals checklist
• Agreement on must-haves for rollout vs backlog
• What didn't go well (specific examples)
• Risks, mitigations, and required investments
• Proof through validated cases
• Timeline and owner assignment
• Training, documentation, and operator checklists
• Scale options, timeline, and resource needs
• Root-cause analysis of top 3 problems
• Communication & shared channel governance
• Operational performance review
• Improvement ideas & countermeasures
• Contingency plans for data/integration failures
• Decision & approvals
• Prioritize fixes and assign owners
• Acceptance decision & next steps