There is a culture present in incident retrospectives. Participants expect things to go a certain way, take a certain amount of time, and abide by certain rules.
The following are a collection of ideas of how to go about changing from a culture of quick, shallow retros focused on identifying a few contributing factors and a small number of very time bound corrective actions to a culture that seeks deep learning and more systematic changes to the systems and the organization that produced it:
- start with small experiments
- let the results speak for themselves
- those that are interviewed should tell the story, not the interviewer
- promote multiple points of view; the more differing the better because contrast is data; that is where you will find the greatest miscalibrations; capture uncertainty and trace it
- the greatest obstacle as an investigator is what you know; don't assume others know what you know; be the unknowing person; understand how others understood
- involvement in the incident taints the interviewer so use someone with enough background to not have to delve into basics, but not too close to the situation
- these are not single meeting events; expect to have multiple rounds of interviews for a single incident and even multiple interviews of a single person!
- there will be follow up meetings
- do incident metrics as required by the company
- track the uptake of the information published by the more in depth incident analysis to gauge impact