Connecting the Dots: Exploring the Power of Network Visualization for Maintenance & Reliability Metrics

I spent some time this week, when I would have been attending the SMRP Annual Conference , feeling in the spirit of SMRP and exploring the Best Practices Metrics. My focus was on network visualization of the metrics, a method to better understand the interrelationship between different metrics.? This approach offers an interactive way to see how certain metrics share components and definitions, which can be useful for consistency checks.

Network visualization is a potential benefit of having metrics available in a digital format, beyond the limitations of an unstructured document.? As we move further into an era of increased digitalization, there’s a growing need for standardized metrics that are straightforward to integrate across generalized data and with data-driven models.? While I’ve done previous work in making recommendations to generalize SMRP metrics and make them more readily computable from real world data, visualization of the metrics is new to me.

In this blog post, I’ll share what I did and what I discovered.? My hope is that, by the end, you’ll have a better sense of the potential digitally formatted metrics hold for improving consistency and offering deeper insights into their structure and interdependencies.

About the Best Practices Metrics

The SMRP Best Practices are a well-established collection of guidelines for metrics designed to measure aspects of maintenance and reliability processes within an industrial organization.? These metrics are widely used in manufacturing, particularly in process industries.? Now in its 6th Edition, the Best Practices are structured around the SMRP Body of Knowledge (BoK) which is based on 5 pillars: 1 – Business Management, 2 – Manufacturing Process Reliability, 3 – Equipment Reliability, 4 – Organization & Leadership and 5 – Work Management.

The metric definitions are constructed such that each metric is calculated using components, which are lower-level foundational metrics such as Crew Capacity and Contractor Labor Hours.?? Many metrics use the same component metrics.? For example, Total Number of Maintenance Craft Workers is a component which is used in both 5.5.1 CRAFT WORKER TO SUPERVISOR RATIO and 5.5.2 CRAFT WORKER TO PLANNER RATIO metrics.

Network Diagrams 101

Network diagrams, or graphs, shows interconnections between a set of entities.? Each entity is represented by a Node.? Edges are the connections between two nodes.? A directed graph is a network where edges are represented as directed arrows.

In a network diagram representing SMRP metrics, each metric or metric component is represented as a node. The diagram forms a directed graph based on parent-child relationships between metrics: if one metric serves as an input to another, it is considered the "child" of that metric. Metrics with multiple definitions (such as those calculated by count or by percentage) were included once per definition, which slightly increased the total number of nodes over the number of SMRP metrics.

I got the idea to experiment with network visualization from Michael P. Brundage , who proposed this approach for metrics visualization using ISO 22400 , which are 34 standardized metrics for manufacturing operations management.? I converted the Best Practices to an interactive network visualization using a python package called pyvis to visualize the network, but there are many packages available.? ?

An example below shows a network of 6 nodes comprised of 3 SMRP metrics Pillar 5 (colored violet) which share 3 component definitions (colored gray):

There are a total of 6 edges defined by the metric definitions: Operating time and Number of Failures are input components to both 3.5.1 MTBF and 3.5.5 MTTF.? ?Operating Time has another edge pointing to 3.5.3 MTBM as it is also used in the definition along with Number of Maintenance Actions.

Results

The Best Practices network has 179 total nodes split into 73 metric notes and 106 component nodes.? The full visualization is shown below, where the nodes are colored by the Pillar as indicated in the legend:

While you can’t see anything specific in this picture, from the big picture we can immediately make a couple observations:

Work Management dominated. There is so many more Work Management metrics (blue dots) than metrics from the other 4 pillars.? Anyone who has engaged with the Best Practices knows this is true!

Connectivity.? Observe that the Best Practices metrics are not entirely interconnected into one large, tangled network. Instead, there are smaller, disconnected clusters, some consisting of just 1-3 full metrics, alongside a few much larger clusters.

In total, there are 21 groups, with an average of 8.4 nodes per group. However, this average is skewed by a few groups with a large number of nodes, the largest containing 56 nodes, as reflected in the median of 3 nodes per cluster. Let’s take a closer look at some of the larger clusters.

Distinct metric groups form distinct clusters.

The clusters containing 15 and 11 nodes both contain 5 metrics from Pillar 5.? The cluster with 11 nodes includes the two definitions for 5.3.6 Planner Productivity (by Job Plans and by Labor Hours), their input components, along with related metrics 5.4.8 PLANNED BACKLOG, 5.4.9 READY BACKLOG and 5.4.3 ACTUAL HOURS TO PLANNING ESTIMATE).? ?

The second cluster, containing 15 nodes, features many of the storeroom-related metrics, as shown below.? Since these metrics form a distinct group within the Work Management Pillar, it makes sense for them to exist as a separate cluster. ?Similarly, there is a standalone cluster for the Pillar 4 Maintenance training metrics.

Overall Equipment Effectiveness (OEE) Metrics

The second-largest cluster, containing 21 nodes, contains the OEE metrics. At the center are the Pillar 2 OEE metrics with connections to Performance and Quality on the left and to Downtime on the right, which are in violet as part of Pillar 3.? The relationships between these values are clearly defined in the Best Practices, particularly in the OEE Components figure and Guideline 2.0: Understanding Overall Equipment Effectiveness (OEE).

Work Types Mega Cluster

The very largest cluster is centered around the Work Types, Labor Hours and Costs metrics. 26 of the 73 SMRP metric nodes are part of this network, driven by Total Maintenance Labor Hours and Total Maintenance Cost, spanning across three pillars.

Labor Hours and Cost are frequently used as normalization factors across many metrics. Additionally, well-defined relationships between metrics in this cluster can be attributed to Guideline 5.0: Maintenance Work Types, which thoroughly defines and explains the different work types and their interconnections.

What do the smaller clusters tell me?

While we learn something from the larger clusters, we can also learn from looking at the smaller clusters.?

1. Need for clear definitions

One observation I made is that the larger, more interconnected clusters are often the result of clearly articulated concepts. Many of the smaller clusters, on the other hand, share component definitions that lack the same level of clarity seen in metrics like Work Types or OEE.? For example, take the concept of "Wrench Time." While it's calculated using "Total Work Time" and "Wrench Time," the specific factors that influence these—such as activities or personnel—are not as clearly defined as the OEE factors or work types. This lack of definition makes the relationship between these components less clear for specific calculation guidance.

2. Need for clear definitions around counting personnel.

Another issue is the inconsistency in how maintenance employees are counted across the metrics. In 5.5.1, we have metrics for the number of Maintenance Craft Workers and Supervisors, while in 4.2, the total number of maintenance employees is used. It’s unclear whether these figures are meant to represent the same population. To further complicate things, 5.5.3 introduces the term "Maintenance Employees (company or owner resources)," while 4.2.1 uses "Number of Internal Maintenance Employees (both salaried and hourly)." The inconsistent terminology and lack of clarity in how personnel are counted creates confusion and needs better qualification.

2. Need for clearer guidance on Predictive Maintenance

SMRP does not formally differentiate between ?“Predictive Maintenance” and “Condition-based maintenance”.? In fact, the Best Practices explicitly states: “The terms condition based maintenance (CBM), on-condition maintenance and predictive maintenance (PdM) can be used interchangeably.”? ?From a network perspective, this leads to some interesting behaviors. Inconsistent connections appear where one term is used in place of the other, and a tangled web forms when both are used simultaneously.

This inconsistency highlights a larger issue: the need for more developed guidelines around predictive maintenance. The scattering of predictive maintenance metrics throughout the network, as seen below, further reflects this gap. Some recommendations for rethinking the terminology and how to measure the effectiveness of predictive maintenance programs can be found here .

Conclusion

Network visualization serves as a valuable tool for dynamically engaging with metrics, showcasing one of the key benefits of digitalizing standards content. Not only does it help identify related metrics and promote consistency in definitions, but it also reveals areas where interrelationships could be further developed. This capability highlights the strength of digital tools in managing and navigating high volumes of complex content.

What do you think?

***

Update: if you are an SMRP member and want to download and play with the interactive graph, I uploaded a copy to the SMRP Exchange. If you do play with it, I want to hear what you think!