Run regular post-mortems
Ozan Onay — 29 Mar 2011
In Sharks are Dangerous, #2 Ash referred to the policy of revisiting failed acquisitions at Johnson & Johnson. I thought it was worth quoting Munger’s description of this:
At most corporations if you make an acquisition and it turns out to be a disaster, all the paperwork and presentations that caused the dumb acquisition to be made are quickly forgotten. You’ve got denial, you’ve got everything in the world. You’ve got Pavlovian association tendency. Nobody even wants to even be associated with the damned thing or even mention it. At Johnson & Johnson, they make everybody revisit their old acquisitions and wade through the presentations. That is a very smart thing to do.
The point is that there are very compelling emotional reasons why an organization might naturally avoid revisiting failures, but we should do so anyway in order to improve our future decision-making abilities.
If we recognize this, we might say “well if we fail, let’s be sure to do a thorough post-mortem”. Unfortunately, by that stage we are already affected by confirmation bias so might not adequately recognize the failure. Furthermore, somebody must initiate the post-mortem, and that person will feel the full effect of both Pavlovian association tendency and fear of the Persian Messenger Syndrome.
One solution to this problem is to institute a regularly scheduled retrospective. Like many Agile teams we do this at the end of every iteration of our product (so each week). Since the meeting occurs at most 5 days after a failure, memory bias is reduced. Since we measure the week’s performance against criteria that were made explicit at the start of the iteration, failures are manifestly evident, which reduces Persian Messenger Syndrome. This is also reduced by the fact that the meeting doesn’t need to be booked ad hoc by any individual—it’s just a part of every iteration. Finally, running retrospectives so often could theoretically improve our abilities to retrospect over time, however I’m not yet confident that this is the case.