Deterioration Modelling – Why It Needs a Dedicated Role
Why Infrastructure Deterioration Modelling Needs a Dedicated Role—and Why ‘Doing It Ourselves’ Often Fails
Why Infrastructure Deterioration Modelling Requires a Dedicated Analyst
An Infrastructure Deterioration Model (IDM) is a complex tool with many moving parts. After many years of working with clients, both as consultants and trainers, we have come to the conclusion that configuring and running an IDM is a task that requires a dedicated person with strong data analysis skills, and possibly some programming expertise—not an engineer or manager already tasked with a wide range of other responsibilities.
We often hear clients express a desire to “run the model ourselves.” While this is understandable, our experience shows that even with thorough training, such attempts almost always prove to be impractical.
The main reason is that many of these clients are asset engineers or asset managers who already shoulder a heavy workload. Adding another complex responsibility—even with the best training—means it will be very difficult for them to become truly proficient in running and fine-tuning deterioration models to deliver good outcomes.
There are three key reasons why IDM work needs a dedicated modelling role:
1. Infrequency of the Modelling Task
If you are a manager in charge of a single infrastructure network, you will probably only run your model and analyse outputs once or twice per year. With this limited frequency, each time you sit down to run the model it will feel like “starting to learn all over again.” This lack of continuity leads to frustration and errors, both for you and for the Lonrix support team.
A dedicated modeller, by contrast, will work with the model repeatedly throughout the year, steadily building mastery of the many interrelated steps involved in the process. People drawn to this type of work tend to have strong analytical and technical skills, and often enjoy the detailed problem-solving involved—including the data handling and occasional programming tasks that can be off-putting to the average civil engineer.
2. Complexity of Data Preparation
In our experience of more than 20 years running deterioration models, we have found that preparing the model input set and the committed treatment set typically makes up at least 50% of the total time required to complete the modelling task, from the first planning meeting to the final report and Forward Works Plan upload.
Tools like the Juno AMS Data Join Tool can greatly simplify the process of merging data from different sources. However, even with advanced tools, most real-world projects still require a variety of custom project-specific calculations, such as flags based on local policies, to ensure the data is suitable for modelling.
For example, you may need to develop scripts to identify unusually long or large modelling elements that could overwhelm the available budget unless split into smaller units. This is often an iterative process requiring strong data manipulation skills.
Missing data is a consistent theme in the preparation of a modelling input set - this requires dedicated, domain specific oversight and attention. For example, you may miss structural capacity information on parts of your network. You can then use data processing skills to assign a best guess estimate based on domain knowledge (e.g. consider rut depth and mesh cracking and pavement age).
The preparation of the committed treatment list is another area where careful work is essential. Treatment names or codes are often inconsistent between source systems and the model setup. These mismatches must be resolved to ensure the model correctly interprets the committed treatment data.
A more challenging issue is mapping committed treatments to modelling elements, which rarely aligns neatly. For example, a single committed treatment may span parts of two separate elements. Decisions must be made about how to allocate such overlaps—whether both elements should be flagged or only those where the overlap exceeds a defined threshold. (You can read more about this topic in this blog post.)
3. Need for Model Tuning
Every infrastructure network has unique characteristics, and those characteristics change over time as assets age, conditions evolve, and budgets shift. Getting a deterioration model to produce optimal, actionable results almost always requires tuning. This involves conducting sensitivity analyses and carefully adjusting thresholds and rates.
Even if someone carefully calibrates your model this year, the next cycle will bring new input data, new committed treatments, and new constraints. Before you can trust the outputs, you’ll need to analyse early runs and make informed adjustments to keep the model relevant and accurate. This requires a person who works with the model regularly and knows how each parameter affects the outcome.
To appreciate the potential complexity of this tuning work, consider the Cassandra Default Road Network Model: it contains more than 200 adjustable parameters, over 50 of which are critical and must be checked and adjusted for every project or network.
What This Means for You
If you are reading this, you are likely either:
(a) someone with a strong data and analysis background who is interested in developing skills to run and maintain infrastructure deterioration models, or
(b) an asset manager or engineer looking for someone to take on this role to support your planning efforts.
If you want to become a dedicated modeller:
Lonrix is rolling out a comprehensive training programme for people who want to build expertise with Juno Cassandra and similar tools. Visit our Getting Started page on the Juno Cassandra documentation site to learn more.If you are a client looking for modelling support:
Please reach out to Lonrix or ASC Consultants or Juno Services if you need help developing a long-term asset management plan, conducting budget-scenario analyses, or creating a Forward Works Plan. While we are actively training more dedicated modellers, our current capacity is limited, so please allow enough time for planning and scheduling.
That’s it for this post. Thanks for reading, and feel free to share or subscribe.