Get Serious About Fixing Bad Data

Recognize the Problem

A. Bad data -- what is it, and how do you prevent it

B. Identify data purpose, i.e. why enter this data altogether? What is it exactly you want to do with this data if it was accurate? This could be the more significant topic.

Even Watson Knows

Leadership Has a Responsibility

It is important to recognize the problem. Once in place, then create a plan of attack. The best plans involve a combination of strategies which include clear roles, process, and technology. The CMMS is a critical business system.

There are Many Ways to Get Bad Data

CMMS Data Needs to be Reviewed and Continually Refined

The maintenance backlog should be periodically reviewed for accuracy, relevance, and prioritization. Occasional plant enhancements can make some work no longer relevant. As to PM-jobplans, these should be refined (i.e. continually improved) along with RCM-suggested strategies. The ideal design has work order failure mode data being systematically compared to the RCM analysis failure mode. Lastly, formal work order feedback can be used to improve safety plans, PM job steps, and PM frequency. The absence of these review processes leads to an inaccurate database.

A Data Quality Plan Leaves Nothing to Chance

Here's an example of what the CMMS community is not thinking of:

Not only should there be a CMMS utilization plan, but there should be an SOP that describes how to insert a new record. On the surface, this sounds simple enough. But consider the asset screen. There can be 20-30 fields on that screen. Which ones are necessary for a new record? Which ones are mandatory? How should the asset description be worded? How do you identify criticality, spare parts, or PM strategy? And what if, in the heat of the battle, a technician discovers a missing asset and wants to "get it quickly into the CMMS" so that he can link the work order to the proper place for actuals reporting? There should be a minimum setup defined (as to what fields must be populated) and then a follow-on action that states, "These additional fields should be populated by the role/person responsible for asset accuracy" which can be at a later date. In that scenario, the asset would be given a status indicating it still needs an update.

Understand Output Requirements as this Impacts Input Requirements

Standard reports are easy to understand, but it is the requested reports (i.e. analytics) by management and others that require reverse engineering. The data may not even exist, or it does, but no procedure is in place requiring population.

In some cases, a KPI is identified as being important, but no one takes the time to study the SQL (and data) necessary to populate this metric.

You know a problem exists when the user community's first instinct is to solve the problem by using spreadsheets outside of the CMMS. Not only is this inefficient but it is also error prone. Typically, this indicates a lack of response by IT or the business analyst, or even the end-user who failed to formally document the request.

Focus on Bad Actors to Manage by Exception

The following curriculum studies asset management from a CMMS perspective including the CMMS utilization plan and Data Quality Plan.