Every 10 years the Census Bureau releases records from 72 years ago, of individual citizens.
The 1940 Census data were released this morning.
Right now when you go to the site, you’ll see this message:
Now I don’t want to rag on the people at the Bureau or the company they partnered with to host the website (Inflection). But there was a big unveiling ceremony with the Census director trying to look up his grandfather, and he hit a wall trying to do the demo as 22 million other people were trying to do the same thing.
I don’t think blame is useful. Unfortunately in cases like this sometimes good people get blamed for something that should just be a big red flag that something about the system needs to change.
So they got a lot more visitors than they expected. Why didn’t they expect that many? How could they have dealt with it differently? Here are some questions I’ll throw out. Ideas for investigation and continuous improvement to produce more successful launches in the future, for the Census and all organizations.
Throw in your own ideas in the comments and share your thoughts.
- How could they have done a better job at gauging the demand, the scope of the project?
- Could they have distributed demand over time? What if they rolled it out gradually, state by state?
- Could more virtualization have played a role in dealing with a sharp spike of demand initially, which will inevitably simmer down shortly afterward?
- How could an implementation plan be structured to expect peak capacity initially with the subsequent decrease in traffic, without buying tons of expensive new hardware that will mostly be wasted after the spike?
As Eric Ries said in the lean startup, “Mistakes are usually caused by flawed systems, not bad people.” There are some important lessons to be learned from today, lessons we can apply systematically in all our projects, very likely.
Please share your thoughts.
Leave a Reply
You must be logged in to post a comment.