I got an interesting piece of feedback from my Seven Signs and the Jena Six post about some ways you can identify a potentially viral rumor or message that would damage your organizational reputation. Essentially, the comment boiled down to a question about the ranking of those threat factors, because Emotion tends to turn even the most judicious of thinkers into a hair-trigger auto-forwarding machine.
Here was my response:
I agree, if I were to rank those seven, Emotion would be in the lead. If I wanted to get really sophisticated with it, I could give each one different weight toward calculating a score.
One day, I will do that.
But the more complexity you throw into the mix, the more training it required to properly code them (and the longer it takes, which works against you in a real 24/7 crisis.) What is optimal in theory is often not workable in practice.
There are constant changes to the recommendations on rescue breathing, beats-per-minute, and number of compressions per cycle. But the mathematician in me knows that the optimum is most likely not a round number, like 10, 15 or 20.
When you put someone into action, though, you want them to actually execute and not freeze up. Hesitation kills. Trying to remember if it was 17 or 19 is counter-productive.
Just for kicks, do a search for Gavin de Becker and his MOSAIC programs. He has made a science of ranking pre-incident indicators of violence, and makes frighteningly accurate predictions about threat assessments. He can look at a seemingly creepy fan letter, and in moments disregard it as not a problem. Rather, his staff can, because there is a formula at work involving a couple of dozen factors, weighted from 1-4 points each. And MOSAIC runs this automatically, as software.
I’m not ready to build MOSAIC for my company — but I am working on something similar for crisis communications monitoring, yet even more basic than the Seven Signs.
So what I really want is eHarmony.com for rumors.
The Quick and the Dead
To work, it has to be something you can implement quickly yet still provides value. The best way to develop such a system involves complex regression analysis, which takes a large pool of already-coded instances and compares them for contributing factors. You dump in the known quantities, and let the software find the relationships.
But until that time, I want something we can use on the fly, with multiple people doing the encoding in a manner that’s uniform enough to give us good results. If I can get the mix just right, we can have a way of prioritizing incoming messages that will be imperfect yet functional. Who cares if something was scored a 2 instead of a 3, when the 4′s are a bigger danger? And seeing a handful of 4′s and a thousand 3′s, maybe volume trumps worst-case-scenario thinking.
In modern crisis management, you can’t take forever to be right.