COBie Handover Is Not Reliability Readiness: A Maximo Acceptance Contract
Importing a building handover can create an impressive asset register, but a populated register is not automatically ready for reliability decisions. This guide defines the evidence and ownership needed between import, commissioning and dependable asset performance analysis.
COBie Handover Is Not Reliability Readiness: A Maximo Acceptance Contract
A capital project can hand operations a building model, a set of spreadsheets and a successful Maximo import, yet leave the reliability team unable to answer a basic question: which equipment is maintainable, where is it, and what evidence supports its maintenance strategy? An import proves that data moved. It does not, by itself, prove that the data describes the commissioned installation or that a future failure can be assigned to the right asset.
An IBM Community hands-on lab makes this distinction concrete. Its author imports COBie workbooks for the Barton Keep sample building into Maximo Manage 9.3 with demo data. The lab reports that three workbooks produced locations, assets, systems, specifications, job plans, master preventive maintenance templates and company records. It then walks through commissioning and a subsequent export. Those results belong to that demonstration, not to every project or every Maximo release.
The larger opportunity is to use that handover as the beginning of a reliability evidence chain. A building component, a location, an asset record, a work order and a failure history all need stable relationships. A maintenance strategy also needs review by someone who knows the operating context. None of that follows automatically from receiving a workbook, and none is guaranteed by a successful import session.
IBM's Maximo Health and Predict dashboard documentation illustrates why the distinction matters. It describes asset information such as installation dates, maintenance history, costs, work orders and meters in the context of different dashboard measures. Some measures have explicit data requirements. The point is not that every building needs every score. It is that a downstream decision can only be as interpretable as the data, definitions and ownership behind it.
I would treat handover as an acceptance contract between the project team and the people who will maintain the building. The contract should establish identity, operating context, maintenance readiness and the evidence needed to correct mistakes. The workflow below is a proposed approach, not a claim that Maximo automatically enforces these gates.
Make the import a traceable baseline, not a declaration of truth
COBie organizes building handover information into structured worksheets. In the Barton Keep lab, the author describes facility, floor and space information alongside components, types, systems and attributes. The Maximo import creates records that maintenance teams can use, but the existence of a record is not evidence that someone has verified the corresponding physical item. A contractor's model can describe design intent, while the installed configuration can change during construction or commissioning. Treat the imported data as a baseline to be checked against the accepted installation.
Start the handover package with a source manifest. Record the workbook names and versions, responsible supplier, delivery date, building or project scope, and the model or source identifiers used for matching. Keep the original files in an approved repository. Define how corrected files are submitted and who can approve a new baseline. Avoid relying on filenames alone: a corrected workbook with the same name can otherwise be mistaken for the version that was accepted earlier.
The lab describes a model identifier retained on imported records to help apply later data drops. Confirm that the corresponding identifiers exist and behave as expected in your supported version and configuration. For each sampled asset, link the source component identifier to the target asset and its location. The business owner should be able to answer which source produced the record, which import session processed it and which reviewer accepted its physical identity. If any of those links cannot be reconstructed, document the gap before moving the data into a decision workflow.
A useful reconciliation has several layers. Compare the delivered component list with the import results, review validation warnings, then sample the resulting Maximo records against commissioning evidence. Check missing assets, duplicated identities, unexpected locations, incomplete descriptions and records created outside the agreed scope. Count comparisons help identify exceptions but do not establish correctness: a component can be present twice while another is absent, and a correctly counted asset can still be associated with the wrong room.
The lab reports 189 locations and 136 assets from its sample import, together with 15 systems and 4,167 asset specifications. Those are useful observations about the lab, not acceptance thresholds. The same source reports a different number of exported attributes; imported specifications and exported attributes are different measures and should not be compared as though they were identical. In a real handover, define each count precisely, name its source and explain exclusions before drawing conclusions.
Keep import, merge, update and export decisions distinct. The community lab describes these as separate session types, and their consequences should be exercised on a nonproduction copy before a live correction. Decide what constitutes a new item, what counts as an update, and how a failed or partial session is investigated. A reliable baseline is one that can be explained and repaired, not simply one that produced a green completion message.
Connect maintainable identity to location and operating context
A reliability analysis begins with a consistent unit of observation. Is the maintained item the whole air-handling unit, an individual fan, a motor, or a component within the motor? The right answer depends on the organization's maintenance and failure analysis practices. A construction model may contain components that are useful for design coordination but not appropriate as separate maintainable assets. Conversely, an item that matters to maintenance may need an identifier and hierarchy that were not supplied in a convenient form at handover.
Agree the boundary with the reliability engineer and maintenance planner before accepting the mapping. For each selected asset class, document the physical item, unique identifier, location, parent or system relationship, and the person responsible for the record. Include the attributes needed for the first maintenance decisions, such as manufacturer or model when available and verified, rather than importing every attribute and assuming it will be useful. Distinguish an unknown value from a default value. A placeholder can make a field look complete while concealing an unresolved decision.
Location is equally important. The lab describes a hierarchy from facility to floor and space, with component-related operating locations beneath the relevant space. It also describes how a component whose space reference is missing can be attached higher in the hierarchy. That is a useful recovery behavior for a demonstration, but it may not satisfy a plant's maintenance standard. A record that points to a building rather than a room may be enough for an initial import and still be inadequate for a technician searching for a specific pump.
Create a sample-based field verification plan. Select high-consequence equipment, representative equipment from each major system, and records with missing or suspicious references. Walk the route with operations staff and compare the tag, asset description, room, parent relationship and installed configuration against the record. A discrepancy should have an owner and a closure condition. Record whether the issue belongs to the source handover, the mapping, the Maximo record, or the physical installation. That classification helps prevent the same error from returning in the next data drop.
Do not assume all assets within a system have identical consequence or duty. A standby pump and a continuously operated pump may be the same model while requiring different maintenance decisions. Capture the operating context that makes a strategy meaningful: duty, redundancy, environment, access constraints and the role the item plays in a larger service. These are examples of decision inputs, not a claim that every one is provided by the import. Establish the data owner and permitted values for the fields your organization actually uses.
The result should be a controlled asset identity register, not a perfect digital twin claimed on day one. Some records may remain provisional while commissioning continues. That can be acceptable if the status, limitations and responsible owner are visible. What is unsafe is presenting provisional hierarchy data as an authoritative basis for maintenance planning or risk ranking without telling the decision maker what remains unverified.
Turn job plans and failure history into a learning loop
A handed-over job plan is a starting point for a maintenance strategy, not proof that its interval, task sequence or safety assumptions fit the site. In the sample lab, the import creates job plans and master PM records. Before activating a real maintenance program, have the planner and equipment owner compare the proposed work with commissioning evidence, supplier instructions, operating duty, access arrangements and applicable safety procedures. Confirm which team can revise the plan and how changes are approved. Do not treat a source workbook as an operating authority merely because its contents imported successfully.
Work through a fictional pump as an example. The handover provides a component identifier, its model, a room and a proposed inspection task. The field team confirms the installed pump and its tag, then checks whether the task describes a safe, observable action. The planner identifies required access and any isolation process through the site's approved procedures. The reliability engineer records which failure modes the task is intended to detect or prevent. If a key detail is unknown, the team assigns an owner and a due date rather than inventing a plausible value.
Next establish how actual failures will be recorded. An IBM Community reliability article explains the value of recording problem, cause and remedy information in corrective work. IBM's failure-data documentation for Maximo for Aviation describes failure hierarchies containing those elements. That product-specific documentation illustrates the concept; confirm the exact fields, configuration and behavior in the Manage environment being deployed.
Do not force technicians to choose a definitive cause before investigation is complete. A symptom can be observed immediately, while its cause may be uncertain or may change after engineering review. Define how the organization represents unknown or pending analysis in its configured codes and workflow. The remedy should describe what was actually done, not what the plan expected someone to do. Retain a narrative where it adds context, and establish who is allowed to revise the structured coding after review.
A monthly review can compare recurring symptoms, investigations and completed actions for the selected assets. The purpose is not to create a sophisticated dashboard before the data is trustworthy. It is to decide whether a task should be changed, whether a recurring issue requires deeper analysis, and whether the equipment hierarchy is capturing the problem at the right level. Document each decision, the evidence behind it and the next review point.
Check whether a work order is being counted as a failure for the business purpose at hand. A corrective task, an inspection finding and an actual loss of function are not automatically the same event. If multiple work orders relate to one incident, the reporting method must say how they are treated. Otherwise a change in work management practice can look like a change in reliability. A disciplined taxonomy and documented event definition are more valuable than a precise-looking metric whose numerator nobody can explain.
Protect downstream scores from false confidence
The Maximo Health dashboard documentation describes several different indicators, including health, criticality, maintenance-to-replacement ratio, remaining useful life and mean time between failure. These indicators are not interchangeable. A health score is not automatically a failure probability, a missing value is not a healthy condition, and a maintenance cost ratio is not a business case by itself. The documentation also notes product and channel distinctions for some features, so validate the exact application and release before promising a particular view.
The same IBM documentation defines its displayed MTBF calculation as asset age since installation divided by the number of unscheduled work orders during that period. That is a specific product definition. It should not be silently equated with every organization's engineering definition of operating time between confirmed functional failures. An installation date that is absent or wrong, or a work-order classification that is inconsistent, can therefore change the interpretation of the displayed value. Before using the measure in a decision, inspect its definition and the quality of its inputs in the actual configuration.
Use a scorecard readiness review for each proposed decision. Identify the decision owner and the action they may take, then list required input fields, source systems, refresh expectations, and what happens when a value is missing. For a replacement discussion, for example, the owner might require verified installation information, a credible replacement cost, maintenance cost scope, and evidence of actual condition. The exact policy is an organizational choice, not a universal Maximo requirement. If the data is incomplete, mark the decision as provisional and specify the missing evidence.
Create a small set of test cases that make data defects visible. Use a nonproduction asset with a missing installation date; another with a deliberately incorrect date; a third with a missing replacement cost; and a fourth with a work order coded differently from the agreed event definition. Compare what the application displays with the documentation for the deployed version and the expectations agreed by the business. Record whether the score is absent, changes, or requires investigation. Do not assume the output in advance: the configuration, product and release determine the behavior.
Trace a selected dashboard result back to its source. The reviewer should be able to identify the relevant asset, the record or input driving the result, the date of the evidence, and the owner responsible for correction. If a data point came from handover, preserve its link to the accepted source version. If an operations team corrected it after commissioning, the correction should not be accidentally overwritten by the next project data delivery. Establish precedence rules and test an update on a controlled copy before applying it to an operational record.
Finally, make model and score limitations visible to the people using them. A precise number may be useful for prioritization, but its precision is not a guarantee of correctness. A reliability engineer should be able to challenge an unexpected score, inspect the underlying data and refer the issue to a named owner. The feedback path from decision to source correction completes the evidence chain that began at handover.
Practical Implications
Treat the COBie lab as a practical example of what a configured import can create, not a promise that every customer will get the same records or timing. Identify the version and import capabilities actually available in your environment, and rehearse the workflow with representative but nonproduction data.
Before commissioning, agree a minimal acceptance pack: source-file manifest, import results and warnings, sampled source-to-asset mappings, physical verification of high-consequence items, unresolved exceptions and named owners. The maintenance planner should review imported job plans and preventive maintenance templates against the site's work and safety practices. The reliability engineer should establish how problems, causes and remedies will be recorded and reviewed.
After commissioning, define who owns each class of asset data and which system is authoritative when a corrected project workbook conflicts with an operational change. Test merge or update behavior before the next handover drop. Retain an audit trail of accepted changes and decisions instead of assuming the latest spreadsheet is always right.
For any Health or Predict dashboard, document the actual product definition, data prerequisites and business purpose of each measure. Run missing-data and incorrect-data tests before using the output for prioritization. Where information is provisional, make that visible to the decision maker and assign an owner to close the gap. These are governance controls the team must design and verify, not claims of automatic product enforcement.
Bottom Line
A building handover is the first chapter of a reliability record, not the last step of a data migration. The Barton Keep demonstration shows how a structured COBie import can create an operating foundation in Maximo Manage. The reliability team still needs to establish whether the right physical items were created, whether their locations and relationships are correct, and whether proposed maintenance work fits the actual installation.
I would release the imported data into operational decision making only after the project and maintenance owners agree on evidence, exceptions and accountability. That does not mean every record must be perfect before work begins. It means the team knows which facts are verified, which are provisional, and how a later correction reaches the right record without undoing an accepted operational change.
The payoff is a traceable path from construction evidence to asset identity, maintenance work and eventual reliability review. A dashboard can then support a decision because its inputs and meaning are understood, not merely because it displays a number. A successful import is the start of that chain; commissioning and disciplined feedback make it useful.