At Google, we have a standard postmortem template that allows us to consistently capture the incident root cause and trigger, which enables trend analysis. We use this trend analysis to help us target improvements that address systemic root-cause types, such as faulty software interface design or immature change deployment planning. Table C-1 shows the breakdown of our top eight triggers for outages, based on a sample of thousands of postmortems over the last seven years.
|User behavior change||9%|
|Service provider change||5%|
Table C-2 presents the top five contributing root-cause categories.
|Development process failure||20.23%|
|Complex system behaviors||16.90%|