Read on Twitter
2 - Part of becoming a good forecaster is learning how to spot a trend happening.
Two months ago, you looked at the data from the Southeast United States ... and you looked at the trends on a weekly basis, and you could see that trouble was brewing.
Two months ago, you looked at the data from the Southeast United States ... and you looked at the trends on a weekly basis, and you could see that trouble was brewing.
3 - Now, when you look at the data in real time every day, you don't really see the trends. That's part of the problem. You have to step out of real time information and analyze via longer time horizons.
4 - When you compared weekly data, you could easily see that trouble was brewing.
Slow-brewed, of course, but brewing.
Slow-brewed, of course, but brewing.
5 - As a forecaster, you then have to be very careful with "what you say and do next".
You might have data that looks like this by week:
24 ... 36 ... 54 ... 60 ... 79
What is the next point in the series?
You might have data that looks like this by week:
24 ... 36 ... 54 ... 60 ... 79
What is the next point in the series?
6 - Somebody will say "Everything is fine ... we had growth rates of 50% for two straight weeks but since then we have growth rates that are about 25% ... so the growth rate will die out and it's over.
7 - Somebody will say "Everything is a catastrophe ... 79 will become 94 then 114 then 156 and in three weeks we will be at 3,000 and we'll be overrun.
8 - Each thesis is likely to be wrong.
9 - Somebody will tell you that you are wrong because they don't trust your ability to analyze information, or they don't trust your data source, or they don't trust that the data is accurate, or they don't think you are analyzing enough data.
10 - Ignore that person. There are always people who will use that line of defense to diffuse you to get you to shut up.
Don't let them silence you. Ignore them.
Don't let them silence you. Ignore them.
11 - Part of this is also observing with your own eyes.
Here in Arizona, behaviors in mid-May looked "normal" ... few masks, taverns full, growth rates very slow. Behaviors didn't pass the eyeball test, the data showed slow growth, behaviors validated growth trajectories.
Here in Arizona, behaviors in mid-May looked "normal" ... few masks, taverns full, growth rates very slow. Behaviors didn't pass the eyeball test, the data showed slow growth, behaviors validated growth trajectories.
12 - So then things come down to how you communicate your forecast.
Unsurprisingly, your choice of words will inflame people who disagree with you.
Don't let it stop you from communicating. Try to be fair, but communicate nonetheless.
Unsurprisingly, your choice of words will inflame people who disagree with you.
Don't let it stop you from communicating. Try to be fair, but communicate nonetheless.
13 - Finally, correct yourself when you are wrong.
