Best Practices for Categorizing Your Data in Aviation SMS

What Are Classifications in Aviation Risk Management

In SMS, it’s extremely useful to organize reported issues by classifying them with classifications.

A classification is simply a unit of organization. By “unit of organization,” we are talking about a short phrase that describes the safety issue in one way. This issue can be described in terms of:

• It’s “type” of problem, such as “Aircraft damage in hangar”;
• It’s a central issue (hazard), such as "Runway excursion";
• The central human component (Human Factor) that contributes to the situation, such as "Overtime hours";
• The root cause(s), such as "Lack of flight training"; and
• Other factors, such as classifying the type of relevant procedure, policy, task, etc.

When you are managing safety issues, you assign classifications to the issue to organize it for later data mining. Over time, it allows you to gather data about common:

• Types of issues;
• Hazards;
• Root causes;
• Human Factors;
• And so on.

Classifications provide the best means of practicing meaningful data mining and data analysis that we have seen.

Purpose of Classifications in Aviation Safety Management

Classifications are the bread and butter of your SMS program’s safety data organization. Making safety decisions without well-constructed classification systems is like riding a bike with flat tires – you won’t get anywhere. In fact, one of the primary markers of a functional aviation SMS program (versus one that is a farce) is a well-organized set of classifications.

Yet as important as they are, we see aviation safety programs all over the globe struggle with building classification systems that:

• Are easy to navigate;
• Comply with standards of oversight agencies; and
• Provide specific data organization (but not TOO specific) that allows for meaningful data analysis.

Good use of classifications provides the following benefits:

• Understand the biggest safety needs;
• Understand safety performance;
• Data mine the safety program for lots of different data; and
• Establish trends in safety issues.

Classifications are fundamental to the efficacy of a safety program. Programs cannot functionally persist or improve without them.

Create Classifications Trees to Organize Classifications

Classification trees, described above, are perhaps the best organized into a classification tree. A “classification tree” allows you to expand and collapse each category and sub-category. It successfully condenses large amounts of information into a manageable format.

Compare this to something like a table. If you are looking at Hazard Classifications, for example, you might have 200 or more classifications:

• As a table, this information would be the equivalent of about 4 pages long with small print;
• As a table, this information would also have much redundant text because every line item would include the name of the category and sub-category; but
• As a classification tree, this information would be about 10 lines long (minimum) to 20 lines long (category fully extended), or about the equivalent of about half a page (small print); and
• Moreover, the classification tree has no redundant text.

However, having a classification tree means having a computer-based product, such as aviation safety software or a similar product. Classification trees cannot be used in other mediums, such as Excel or Microsoft Word-based classification organizational systems.

Related Aviation SMS Data Analysis Articles

• How to Organize Data for Good Data Analysis in Aviation SMS [Best Practices]
• How to Set Up Classifications in Aviation SMS Implementations
• What Is Importance of Data Mining in Aviation SMS Implementations

Organize Classification Trees 3 Levels Deep

A very good practice for creating your classifications trees is to make classifications with three levels:

1. Category - system related classifications, such as
   • Flight Ops,
   • Ground Ops,
   • Maintenance
2. Sub-Category to organize system classifications into relevant groups, such as
   • Landing,
   • Takeoff,
   • ATC,
   • On-Board Aircraft
3. Classification to identify a specific type, such as
   • Rejected Landing,
   • Aborted Landing,
   • Unstable Approach

Common Problem: A big mistake we see operators make is that they will have classification trees that are:

• Only one or two levels deep, which tends to provide classifications that are vague and meaningless; or
• Four or more levels deep, which quickly becomes extremely unmanageable.

Solution: three levels in a classification tree have been shown over and over to be the most effective for getting specific, meaningful, and manageable data.

How Specific Classifications Should Be

One questions we often get asked is: how specific should classifications be? As far as how specific your classifications should be, you want to ensure that each classification is:

• Specific enough to be useful for making safety decisions and establishing trends;
• Not so specific that you need many, many classifications to capture the essential idea; but
• General enough to be useful for capturing the main idea, such as grouping highly similar situations.

Here are some examples of good specificity for hazard classifications:

• Flight Ops>> Wildlife>> Bird strike;
• Flight Related>> Wildlife>> Wildlife on runway; and
• Flight Related>> Wildlife>> Rodent in aircraft.

These examples are specific enough to capture the essential piece of information (“wildlife” and “runway”) but not so specific that you will need to assign a couple of classifications just to capture one idea. The point is: one type of problem, one type of classification.

Less effective use of classifications are listed below – and it’s probably the most common mistake (along with having classifications that are too vague) of using classifications in the aviation industry:

• Flight Ops>> Wildlife>> Moose on runway;
• Flight Related>> Wildlife>> Bear on runway;
• Flight Related>> Wildlife>> Deer on runway;
• Flight Related>> Wildlife>> Mouse in aircraft; and
• Flight Related>> Wildlife>> Rat in aircraft.

You can see that the above classifications are less effective because you need so many more classifications to capture the essential information. Classification trees built in this manner would include many hundreds of classifications, many of which would never be used and would only cause clutter**.

Common Problem: Classifications that are much too specific, leading to highly unmanageable classification trees

Solution: Be only as detailed as you need to be to capture the main idea.

**The above situation would only make sense if your organization was in a dense wildlife area, and it was actually helpful to understand what your top wildlife concerns were. However, most operators would not fit into this description, and the above examples would be unnecessarily verbose.

Don’t Repeat Yourself (DRY)

DRY is a very important term in many industries. The goal is to ensure:

• Organizational best practices;
• Reduce useless clutter; and
• Reduce extra work.

In the context of classifications, it means don’t include the same classifications in different classification trees. This would cause you to have extra classifications and have to classify the same issue twice.

Common Problem: Having the same classifications in multiple trees.

Solution: If you have a classification in multiple trees, remove it from one of the trees – 99.9% of the time it doesn’t need to be there.

Separate Classifications by Type

Different types of classifications should be organized into their own classification trees. Some common types of classifications we see used by aviation service providers are:

1. Type of issue
2. Hazards
3. Human Factors
4. Root Causes
5. Risk Controls (i.e., create a classification tree that includes all risk controls)
6. Policies/Procedures (i.e., create a classification tree that includes all policies/procedures)
7. Job Duties
8. Locations

You can literally create any type of classification tree you want that best fits your organization. The point is that different types of classifications are separated.

Common Problem: Trying to use only one tree to manage all classifications.

Solution: Break up different classifications into different trees depending on which type of data you are trying to capture.

Number of Classifications to Assign Per Issue

As discussed, it is very useful to assign different types of classifications to an issue to organize it. To recap, some of the most common types of classifications are:

• Type of issue classifications;
• Hazard classifications;
• Root cause classifications;
• Human Factor classifications;
• Safety policy/procedure classifications

Assuming you are using multiple types of classifications, here are some loose guidelines that we suggest how many of each type of classification to apply to issues when managing them. Remember, these are loose guidelines, and there is not just one right way to do things – different issues call for different actions:

• Type of Issue: one is ideal, but sometimes two are necessary
• One Hazard: there is only one primary dangerous condition per safety issue;
• Root causes: multiple is good, as there are usually several root causes
• Human Factors: apply only the most important ones per issue, usually 1-3; and
• Safety Policy/Procedure: usually one or less, but up to two is okay

Remember, these are not hard rules, they are simply best practices – i.e. “soft rules”. Some issues will require you to be flexible.

Common Problem: Way over-classifying issues.

Solution: The only hard rule is to apply classifications to issues very thoughtfully because important safety decisions are made based on them!

Last updated August 2024.