Airlines often have the challenge of adapting OEM data to meet their own maintenance requirements. One type of adaptation is called a Customer Originated Change (COC). Although it requires significant effort to manage COCs, it is usually worth the effort. COCs are derived from airline-specific business processes and experience, and are helpful in capturing and sharing best-practices and knowledge that is acquired over time. But some airlines have been tempted to change the actual data model of the documentation that was received from the OEM. This can increase costs, in unexpected ways.
Airlines receive a parts catalog and a maintenance manual in SGML/XML format from aircraft manufacturers, structured according to aviation industry standards (the ATA DTDs). Many airlines decide to transform this data structure to simplify their own publishing process, usually with only minor changes.
For example, an aircraft’s printed parts catalog contains a PNR table at the end, listing each part number numerically—similar to an index. This concept exists in the XML data as well and allows for attributes of each part to be listed in one location, instead of repeated throughout the body of the document each time the part is referenced.
When publishing the XML data (whether to paper, or electronically), this structure requires a more complex stylesheet, because the attribute data is often not sitting exactly at the location that it needs to appear. To reduce stylesheet complexity, some airlines may perform a transformation, copying and inserting this attribute content throughout the textual area of the main body (PCDATA inside the tag).
The immediate result seems wonderful: A faster transformation during the publishing process, with a stylesheet that is easier to create. However, the long term consequences far outweigh these benefits.
As I mentioned at the beginning of this post, the airline is also making COCs – changing content in the document according to their own specific needs. An example of a common COC would be if the airline’s engineering team decided to use an alternate part from an approved spare parts manufacturer (but there are many other reasons as well).
Using the PNR example and a modified data model, if an engineer wanted to update a part number s/he must now do so in multiple locations throughout the parts catalog, instead of just once within the PNR index. This is only a minor problem—search and replace can do the trick—but the real headache, and the real cost to the airline, comes as a result of the ongoing cost for managing these COCs.
Every three months, the OEM sends an updated version of the parts catalog. It is the airline’s responsibility to reconcile all the COCs with the new OEM revision, so that all the best-practices knowledge is re-incorporated. The COC comparison can be a massive resource drain on airline engineering teams, sometimes resulting in a six month delay in adopting new data. (Enigma 3C Revision Manager can dramatically reduce the pain of reconciling OEM revisions and COCs, but that doesn’t alleviate the underlying issue.) During this delay, new efficiencies are not being adopted, and in the worst case, regulatory mandates must be tracked and implemented manually.
Sometimes it’s easier for an airline to just generate a new COC. The innocent data transformation that made the publishing process ‘more efficient’ has resulted in many more COCs being required, with the associated review and approval process as well. I’ve seen cases where an airline had eight times as many COCs as a result of this type of modification.
The lesson learned is that data models are very sensitive to change. Often, the original DTDs do not seem to make sense so the temptation to make a simple change can be overwhelming. Airlines need to resist the temptation, as the long-term costs will usually far outweigh the short-term benefits. A second lesson is that airlines should seek publishing platforms that can handle complex DTDs—with strong stylesheet engines—and that include intelligent COC and revision management capabilities.