Choosing the single asset that matters most

• Which single asset or train are you proposing for this initial pilot? Options: Compressor or compressor train, Turbine or generator, Heat exchanger or exchanger network, Reactor or reformer, Pump or booster, Other
• Describe the operating range and typical duty cycles for that asset, including common load points and upset conditions.
• Tell me about any trips, derates, or unplanned stops this asset experienced in the last 12 months and what the immediate cause was.
• Who owns day-to-day performance decisions for this asset on the plant floor and who owns long‑term asset strategy? Options: Reliability engineer, Operations shift lead, Process engineer, Asset manager, Maintenance planner, Shared/other
• Estimate the annual cost impact when this asset underperforms or has an unplanned outage. Options: Less than $50,000, $50,000 to $250,000, $250,000 to $1,000,000, More than $1,000,000, Unsure

Where today's assumptions leave you exposed

• If your existing analytics only flag symptoms after they appear, how much time and cost do you typically lose before engineers can act? Options: Less than 1 hour, 1 to 8 hours, 8 to 48 hours, More than 48 hours, Unsure
• Walk me through a recent incident where earlier prediction could have prevented a shutdown, what happened, and why your current tools failed.
• Where in your current workflow do model outputs and engineering intuition most often disagree?
• On a regular week, how often does your team run manual what-if studies or offline simulations for this asset? Options: Daily, Weekly, Monthly, Quarterly, Rarely or never
• Which failure modes or performance degradations are hardest to detect today? Options: Efficiency loss or throughput decline, Vibration or mechanical looseness, Sealing or leak degradation, Heat transfer fouling, Control loop instability, Other

Who must say yes, and what keeps them awake

• Identify the single stakeholder who, if unconvinced, would stop the pilot from moving forward immediately. Options: Plant manager or site director, VP of Operations, Reliability manager, Head of Engineering, IT/OT lead, Legal/procurement
• List the other stakeholders who will be involved in evaluation, implementation, or sign-off.
• Who among your stakeholders prioritizes model explainability over raw numerical accuracy? Options: Operations and shift leads, Reliability and maintenance, Process engineering, IT/OT, Senior leadership, Mixed views
• How do senior leaders quantify the return they expect from a validated twin, in dollars, hours saved, or KPIs?
• When the pilot shows a positive result, who must approve the production rollout and what level of budget authority do they hold?

Can your data actually run the twin

• If we cannot access historian data at 1-minute resolution for the selected asset, can you still meet your pilot acceptance criteria? Options: Yes, Yes with downsampling or interpolation, No, we need higher resolution, Unsure
• List the systems that hold the data we need, for example historian, DCS, PLC, CMMS, or others. Options: Historian, DCS, PLC/RTU, CMMS/Maintenance records, Operator logs, Other
• Name who owns access and credentials to those systems and how long it typically takes to obtain read access.
• Describe the typical data quality issues you see for this asset, such as missing timestamps, sparse events, sensor drift, or inconsistent units. Options: Missing or irregular timestamps, Gaps or sparse sampling, Sensor drift or bias, Inconsistent units or tags, Rich, consistent data
• Are there regulatory reviews, legal approvals, or vendor agreements required before data can be shared for a pilot? Options: Yes, legal review required, Yes, operations approval required, Yes, vendor approval required, No approvals required, Unsure
• How many dedicated hours per week can your team commit to model calibration and validation during the 30 to 60 day pilot? Options: 0 to 5 hours, 6 to 15 hours, 16 to 30 hours, More than 30 hours

Define success in a way that actually moves the needle

• Assuming the pilot model meets your target accuracy, what exactly changes in the operations room within a day?
• Choose up to three KPIs that will determine pilot acceptance. Options: Prediction error or RMSE, Reduction in nuisance alarms, Unplanned downtime hours avoided, Fuel or energy efficiency, Throughput or yield increase, Maintenance cost reduction, Number of specific alerts, Safety incident reduction, Other
• Give the acceptance band you would require for model prediction error, selecting from common bands. Options: Less than 5% error, 5 to 10% error, 10 to 20% error, No strict band, judge by impact, Unsure
• Identify who will sign the formal pilot acceptance and what documentation or evidence they will need to be convinced.
• Assuming the twin demonstrates the stated benefit, how soon would budget holders authorize a production rollout? Options: Immediately, within 2 weeks, Within the next quarter, Within 6 months, Requires next budget cycle, Unsure
• On a practical level, how critical is avoiding false positives that trigger unnecessary maintenance? Options: Critical, cannot tolerate, Important, would prefer low false positive rate, Tolerable if true positives are high, Not a priority

What could stop this before build starts

• What single technical or organizational barrier would force you to pause or cancel the pilot?
• Name internal projects or priorities that could conflict with resource availability during the pilot window. Options: Planned turnaround or outage, IT or OT migration, Regulatory inspection, Other capital projects, No conflicting projects
• Tell me which roles on your team would need to be replaced or backfilled for this pilot to stay on schedule.
• Explain how your change control process currently treats model-driven alerts or changes to operating setpoints.
• Would contracting, data-sharing agreements, or indemnity terms typically add more than four weeks to your timeline? Options: Yes, No, Sometimes/Depends, Unsure

The alternatives you're actively weighing

• Suppose your incumbent provider delivered the same validated accuracy and integration, would you still change providers? Options: Yes, still change, No, would stay with incumbent, Depends on cost and support, Unsure
• Select the external alternatives you are evaluating now, such as other vendors, systems integrators, or internal build options. Options: Other digital twin vendors, Systems integrator or consultancy, Internal software development, Academic or research partner, No external alternatives
• Has anyone proposed an internal build instead of a vendor pilot, and if so, what timeline did they estimate?
• What would have to be true about your current monitoring tools for you to decide not to engage an external partner?
• Provide the vendor or internal capability you consider the incumbent for this use case. Options: Internal tool or team, Existing vendor, No incumbent, Unsure
• Select up to three decision drivers when choosing between alternatives. Options: Total cost of ownership, Time to value or speed, Prediction accuracy, Ease of integration with your historian, Vendor support and SLAs, Ownership of IP and models, Security and compliance

Next steps that make or break momentum

• Should the pilot prove the numbers, what approval or procurement step would still block a signed contract that week?
• Provide calendar windows that are off-limits for integration or testing because of planned outages or peak operations in the next six months.
• Please list the people who should be in the kickoff meeting from your side to remove blockers within the first week. Options: Reliability engineer, Process engineer, Operations shift lead, IT/OT contact, Data owner/historian admin, Legal or procurement rep
• Estimate how quickly you can provide the initial dataset and access after contract signature. Options: Within 1 week, 1 to 2 weeks, 2 to 4 weeks, Longer than 4 weeks
• Would you need a pilot scope that guarantees limited exposure, for example read-only access or test environment only, before approving integration? Options: Yes, read-only access only, Yes, test environment only, No, full integration acceptable, Unsure
• Finally, what's the best internal metric we should report weekly to keep your sponsors engaged?

• Solution Experience Session

• Confirm the current state and what it is costing you
• You confirm the stated current state and the operational cost that makes this urgent.
• Provide a historian extract (time range and tags) and recent operating logs for the selected asset to enable a follow-up calibration run.
• You confirm that the demonstrated what-if scenario answers your top operational question.
• Prioritize the top operational question to prove
• Deliver an initial calibration and model-versus-historian comparison for the chosen scenario before the next meeting.
• Confirm the pilot decision timeline and nominate the technical and business approvers for acceptance sign-off.
• You agree the proposed pilot acceptance criteria and the evidence that will be collected during validation.
• Walk through a what-if scenario using your context
• Show calibration and model vs historian comparison
• You identify any remaining concerns that must be resolved before pilot approval.
• Identify any integration constraints (SCADA/historian access windows, data owners, credentials) that could affect pilot start.
• Agree pilot acceptance criteria and evidence needed
• Validation check, confirm this maps to your need

• Solution Experience Session
• Solution Experience Deck
• Solution Brief

• meeting
• slides
• document

• Integrate Historian and SCADA Data Streams
• Ingest Maintenance and Asset Records
• Develop Physics-Based Asset Twin
• Train Data-Driven Model Components
• Calibrate Twin to Historical Operations
• Deploy Virtual Sensors and Estimators
• Activate Real-Time Monitoring Dashboards
• Configure Operational Alerts and Thresholds
• Enable Degradation and RUL Predictions
• Enable What-If Scenario Simulation
• Execute 30–60 Day Parallel Validation
• Integrate Twin into Operations Workflows
• Train Engineers and Operators on Twin Use
• Ongoing Model Retraining and Recalibration
• Provide APIs and Data Export Connectors

Integrate Historian and SCADA Data Streams

• Do you have a current tag list for the selected asset in your historian or SCADA (include tag names and asset ID)? Options: Yes, full tag list available, Partial tag list only, No, require discovery
• Which communication protocols and endpoints are available for the historian or SCADA for this asset (for example OPC UA server URL, REST endpoints, or file export location)? Options: OPC UA, OPC DA, MQTT, REST/HTTP, File export (CSV), Other
• Specify the required sampling cadence for key signals such as rotor speed, bearing vibration axis, inlet temperature, and main flow meter during the pilot. Options: 1 Hz or higher, 0.1 Hz to 1 Hz, Minute-level (1/min), 5 minute, Hourly
• What measurable acceptance criteria will confirm data ingestion is complete for the pilot (for example tag coverage percent, max latency seconds, and allowable missing-data rate)?
• Identify the person or role who can approve historian access credentials and provide the tag list by asset ID.

Ingest Maintenance and Asset Records

• Provide the source systems for maintenance history and asset master data such as CMMS exports, work order system, or spreadsheets. Options: CMMS export, Work order system, Spreadsheets, PDF reports, Other
• List the asset identifiers in your maintenance records that map to the selected asset nameplate, tag, or equipment number.
• How far back should maintenance logs, failure reports, and work orders be ingested to support calibration and root cause analysis for this asset? Options: 12 months, 24 months, 36 months, Custom
• Who is the CMMS administrator or maintenance engineer that can provide exported failure mode codes, work order PDFs, and spare parts lead times?
• Describe any conventions in your work orders or failure codes (for example standardized failure mode codes or custom fields) that the team should map.

Develop Physics-Based Asset Twin

• Will you provide asset blueprint artifacts such as P&ID pages, single-line diagram pages, and equipment nameplate data for the selected asset? Options: All documents available, Some documents available, No, require extraction
• Are there recent hardware changes or retrofits such as new pumps, added bypass lines, or replaced heat exchanger bundles that the physics model must reflect? Options: Yes, No
• Estimate which design parameters you can share for the twin such as heat transfer area, rotor diameter, compressor map files, or valve characteristics. Options: All available, Partial list, None available
• Indicate any proprietary OEM performance curves or vendor models that must be included or that cannot be shared due to licensing. Options: Include and share, Include under NDA, Cannot share
• State any control logic artifacts to include such as PLC ladder fragments, trip logic descriptions, or PID setpoint tables.

Train Data-Driven Model Components

• Name the historical time windows and alignment you prefer for training data for this asset (for example rolling 12 months aligned to 1 minute tags). Options: Rolling 6 months at 1 min, Rolling 12 months at 5 min, Custom
• Confirm which sensor tags and units are considered reliable for training such as main flow meter, bearing vibration axis 1 (mm/s), and exhaust temperature (C). Options: Full reliable tag list provided, Partial list, No reliability assessment yet
• Assign who will label supervised events in the dataset such as trips, maintenance windows, or known degradation episodes (include role or email).
• Select the acceptable minimum number of failure or degradation events in historical data needed to train classifiers or RUL estimators. Options: None required, 1-3 events, 4-10 events, More than 10
• Detail any data preprocessing rules you require such as unit conversions, outlier trimming, or gap interpolation methods.

Calibrate Twin to Historical Operations

• Set the list of historical operating campaigns or runs to use for calibration including start and end dates and operating modes.
• Outline the calibration performance targets you expect for key outputs (for example normalized root mean square error NRMSE or mean absolute error MAE thresholds on efficiency or flow). Options: NRMSE < 5%, NRMSE < 10%, MAE within engineering tolerance, Custom
• Measure and document the calibration thresholds and evidence that will define successful calibration for handoff to validation such as agreed residual plots, NRMSE percent, and example trace overlays.
• Report who will sign off on calibration results from operations, process engineering, and reliability stakeholders.
• Choose whether to exclude known bad-data periods from calibration such as planned outages, instrument maintenance windows, or commissioning tests. Options: Exclude outages and tests, Include everything, Exclude by custom list

Deploy Virtual Sensors and Estimators

• Enter which unmeasured physical quantities you want virtual sensors for such as tube metal temperature, internal fouling factor, or mass flow where no meter exists.
• Supply the expected output latency and cadence for estimators for operator use (for example sub-second for HMI, 1 minute for trending, hourly for reports). Options: Sub-second, 1 minute, 5 minutes, Hourly
• Clarify which existing physical sensors will serve as ground truth for estimator training and validation including tag names.
• Explain any safety or regulatory constraints on using estimator outputs for automated actions such as trips, interlocks, or permissives.
• Approve whether virtual sensor outputs may be written back to the historian under a virtual tag namespace for operational visibility. Options: Yes, write back to historian, No, read only, Write to sandbox historian

Activate Real-Time Monitoring Dashboards

• Recommend the operator views you require such as unit performance overview, per-shift KPI trends, and ranked anomaly lists.
• Allocate which user roles should have dashboard access and which role-level widgets they need such as operator, reliability engineer, or plant manager. Options: Operator, Reliability engineer, Process engineer, Plant manager, Other
• Document required display formats including control room widescreen HMI, web dashboard, and mobile summaries. Options: Control room HMI, Web dashboard, Mobile summary, All of the above
• Include the desired dashboard refresh frequency aligned to historian sampling for use cases such as anomaly detection or shift handover. Options: 1-5 seconds, 30 seconds, 1-5 minutes, Hourly
• Attach examples of KPIs and visualizations you expect such as fouling heatmap, efficiency trend lines, or vibration spectra.

Configure Operational Alerts and Thresholds

• Where are the highest priority alerting conditions for this asset such as imminent tube rupture, compressor surge, or loss of seal gas?
• When an alert triggers, do you want to use existing thresholds or should the platform propose thresholds derived from historical operating envelopes? Options: Provide thresholds, Platform propose, Hybrid
• Why should specific notification channels be used for each alert type and which channels do you require such as control room HMI, email, SMS, or CMMS ticket creation? Options: HMI alarm, Email, SMS, CMMS ticket, Other
• Ensure you name the role responsible for triage when a model generated alert is raised and the expected SLA for initial response.
• Do you require alert suppression windows for safe-operational periods such as startup, shutdown, or commissioning? Options: Yes, No

Enable Degradation and RUL Predictions

• Which components and failure modes do you want remaining useful life predictions for such as bearing wear, tube fouling percentage, or catalyst activity decay?
• Specify historic failure records, failure to failure timelines, or work order durations you can provide to seed RUL models.
• What prediction horizon is specific for your planning horizon such as 24 hours, 7 days, or 30 days before planned maintenance is required? Options: 24 hours, 7 days, 30 days, Custom
• Provide whether RUL outputs should map to spare parts lead times and whether the model should trigger procurement actions. Options: Map to spares and trigger procurement, Map to spares only, No procurement integration
• List the recipients and downstream systems to notify on RUL alerts such as CMMS, maintenance planner email, or shift supervisor.

Enable What-If Scenario Simulation

• How do you want to parameterize typical what-if scenarios such as increased feed rate percent, reduced coolant flow liters per minute, or higher ambient temperature C?
• Who should be able to run scenarios and change inputs such as process engineers, operations managers, or authorized simulation users? Options: Process engineer, Operations manager, Reliability engineer, Other
• Are there constraints to scenarios such as maximum allowable pressure or temperature per P&ID and safety interlocks that must be enforced? Options: Yes enforce P&ID limits, No limits required, Custom constraints
• Identify which control actions should be adjustable in scenarios for example setpoint changes, valve positions, or feed composition.
• Describe whether scenario results must include probabilistic uncertainty bands or only deterministic outputs for decision making. Options: Deterministic only, Probabilistic only, Both

Execute 30–60 Day Parallel Validation

• Name the planned start date and duration for the parallel validation window and note any blackout periods during which comparisons are not permitted. Options: 30 days, 45 days, 60 days, Custom
• Confirm which live operating signals and virtual outputs will be compared during validation and the comparison cadence such as 1 minute or 5 minutes.
• Set the decision metrics and numeric thresholds that will confirm pilot success during the 30 to 60 day validation for example NRMSE percent, anomaly detection true positive rate, or lead time accuracy.
• Report who will run daily or weekly validation reviews and where validation evidence will be stored such as a shared drive, project workspace, or audit log.
• Choose whether planned operating campaigns such as load ramps or maintenance events must be included in the validation window. Options: Include planned shutdowns and ramps, Include normal operation only, Include both

Integrate Twin into Operations Workflows

• Enter which operational workflows should consume twin outputs such as shift handover reports, maintenance planning, or HMI annotations.
• Supply whether automated ticket creation in your CMMS is required when the twin flags an issue and the desired ticket template. Options: Yes integrate with CMMS, No integration, Manual ticket creation
• Clarify the handover artifacts required for operations including runbooks, updated standard operating procedures, and accepted test scripts.
• Explain who will be the day to day operations owner for the twin and who will escalate model issues to engineering.
• Approve whether model outputs used for operational decisions must produce an auditable trail for regulatory review. Options: Yes audit trail required, No audit trail required

• desired_state
• current_state
• stakeholders
• gaps
• success_criteria
• decision_readiness
• desired_state
• decision_readiness
• current_state
• success_criteria
• gaps
• stakeholders
• desired_state
• success_criteria
• stakeholders
• gaps
• current_state
• decision_readiness
• decision_readiness
• decision_readiness
• decision_readiness
• decision_readiness

• Master Services Agreement (MSA)
• Statement of Work (SOW)
• Subscription Agreement & Order Form
• Data Processing & Data Sharing Agreement (DPA)
• Acceptance Sign-Off Certificate
• Service Level Agreement (SLA)
• Governance & Rollout Addendum
• Pricing & Expansion Addendum
• Security & Compliance Addendum (industry conditional)
• IP & Model Rights Addendum
• Change Order Agreement

• Go-live Health Check (weeks 1-4)
• First Measurement Review (weeks 4-10)
• Operational Realization Review (90-day production window)
• Quarterly Success Review (ongoing cadence)

• Publish the prioritized quarterly backlog with owners and target completion dates.
• Validate that operational outcomes meet or are explainably moving toward the targets recorded in Pilot Validation, or document remaining gaps and closure dates.
• Establish a retraining and monitoring cadence to maintain model accuracy across operating regimes.
• Publish a concise lessons-learned and runbook update for operations and engineering.
• Publish the lessons-learned document with identified changes to the operational runbook.
• Define and schedule the model retraining cadence and required data collection windows.
• Close any remaining high-priority remediation tickets or extend timelines with rationale and dates.
• Trend review of core metrics
• Ensure no metric regressions in model in-calibration uptime or alert SLA compliance without assigned remediation plans.
• Prioritize the backlog for the next quarter with clear action items and resolution timelines.
• Confirm monitoring, historian/SCADA access, and support coverage remain intact for continued operation.
• Close low-priority tickets older than 90 days or reclassify with rationale and new timelines.
• Confirm continued access to historian/SCADA endpoints and document any upcoming maintenance or window changes.
• Reconfirm success criteria and owners
• All deployment integrations verified as either healthy or assigned a remediation action with target date.
• Early adoption signals recorded and baseline usage metrics established for follow-up measurement.
• Critical blockers documented with remediation actions and owners.
• Document remaining data feed gaps and schedule fixes with expected completion dates.
• Publish an early-adoption summary showing initial user activity and alert acknowledgements.
• Create a short runbook for first-response to model alerts and distribute to operations.
• Present first measurement data
• Determine whether mean absolute error and percent-within-threshold are trending toward the targets recorded in Pilot Validation.
• Identify the dominant root causes for any gap and agree concrete calibration or data actions.
• Set a clear timeline for fixes and the date for the next measurement checkpoint.
• Update the calibration dataset and run a targeted recalibration of the model with specified operating regimes.
• Correct historian field mappings and document any missing signals required for model inputs.
• Schedule the follow-up measurement review and circulate required data extracts in advance.
• Present operational outcomes vs Pilot Validation targets
• Deployment and integration verification
• Sustained model accuracy review
• Backlog and enhancement triage
• Root-cause analysis for gaps
• Monitoring and support coverage check
• Calibration and data fixes
• Early adoption and usage signals
• Close persistent blockers and confirm mitigations
• Timeline to operational stability
• Capture lessons learned and operational playbook updates
• Short action backlog and next steps
• Open issues and blockers
• Agree immediate remediation actions