My favorite talk from Social Dev Camp Chicago was Michael Slaby (@michaelslaby), former CTO of the Obama campaign.

The full notes follow, but some of my favorite tidbits are:

  • Obama had over 100 paid staffers doing new media - McCain campaign had 10
  • $500M of the $700M raised by the campaign was raised online
  • Until they announced Palin, we thought we could win alaska.
  • In the first six months of the campaign, Michael was the only guy who knew HTML/Javascript
  • Worked 600ish days - had five days off (?!)
presidential campaign is not like a normal company
presidential campaign is a temporary enterprise
billion dollar 4000 person enterprise
then we tore it down

most important thing about data is not about the tech
its about who you are/what you are trying to accomplish
pres campaigns produce so much data and information
if you don't understand yourself you'll never be able to wade through the data
real easy to get off track

pres campaign is not a tech company
"how does this help us win?"
"we don't do novel tech things because they are novel"
knowing who you are eliminates some of your decisions which makes it easier
the center point of how we thought of everything was getting obama elected

The campaign narative
"america needs change"
"real change comes from the bottom up, not the top down"
"openess and transparency"

election timeline
think in long time-frames
knew that this would be a long campaign
primary season had a lot of elections
think about how you engage people on a lifetime value scale
think about how you are going to have a relationship with people that keeps them engaged for two years

New Media and Technology in the campaign
Was about:
1. message
2. money
3. mobilization
digital strategy people reported directly to campaign manager instead of an opps guy
they were a peer of the communications department - not under their rule

field data drives campaigns
how many people did you talk to? what are people saying?
how do you deploy resources? where are you going to focus?
what is polling telling you about different territories?
how many people do you need to reach to win that area?
if you're gonna lose, how do you lose effectively?
how many people do you want to get out in a caucus state so you don't lose so bad that you lose a delegate?

we raised 700 million dollars in the campaign
500 million was online
2.4 M facebook friends
90m youtube views
500 mil online donations
4m individual donors
3m volunteers
Online program did not win the election - the field organizers is what wins the election.

Multivariate Testing and Conversations
What are the conversion points that matter in your organization?
/shows a webpage with two elements: A picture and a button
Tried variations the button:
1. "sign up"
2. "learn more"
3. "sign up now"
4. "join us now"
1. get involved image
2. family image
3. change image
Used google site optimizer to see what works
combos of "learn more" and the family image worked 40% better than what they were doing before
40% improvement is worth $57million dollars, 288k volunteers, 2.8m email subscriptions
knew what a lifetime value on someone on the email list was worth
Knew a cost per acquisition
Could pay for a acquisition campaign that paid for itself in three days - before the check had to be sent to the CPC vendor

how fast can you make things talk to each other
-field data
-polling data
Never got all the way there... trying to do that for 2012
Web data was most active data... but of the poorest quality
On web we only knew email / zipcode.  Didn't know name/gender
Had to do more to get their info.... so we give away bumperstickers
didn't give the bumperstickers just to give em... gave 'em to get the information
people got another step more engaged by displaying their affiliation
now we can match the field DB with the web DB
integration for the sake of integration is nice, but if its not useful for the part of the org that needs the data then it doesn't matter

make sure these platforms never go down
times when we were taking in millions of dollars per hour on the web
when he started he was the only guy doing html/javascript for like 6 months on the campaign
the stack did not look polished... it was something we build as we went along
"we were the senator from Illinois with the funny name that no body knew"
"we didn't have any money, that $500m came later"
undrestand the way you need to scale your infrastructure later

had staff of PhD in econometrics
never have as many resources as we want
how many staff do you need in each state
what is the vote count you need in election day?
how many people do you need on the field to get that many votes?

Key tech metrics
1.4 new voter registrations
10m polling location emails
passed $10m/day in donations 5 times
98% accurate turnout prediction

know yourself - you only get one story
understand engagement ladder
envision bottlenecks and scale
This isn't about how to scale a website - but how to scale your website

Q: how did data support a 50 state strategy?
when knew based on political experience that there was an opportunity to win places that were not traditional
this was high level knowledge, not granular data
Until they announced Palin we thought we could win alaska.
People thought we were nuts, but we're pretty sure we would have.
Youtube didn't even exist in 04 during the last campaign cycle
We were breaking facebook every day - before fan pages.  We were part of the reason they made fan pages
twitter launched at the beginning of our campaign
A lot of our social media data is very anecdotal
voter files don't have email addresses typically - they're not well maintained
did not buy email addresses
had a massively active email list
we understand voting habits of people we are talking to
are you talking to undecided voters? people who stumbled into your site an are curious?
had an interesting mix - a lot of die hard who were really interested in being on the email list.
Also huge group of people hanging out on the email list who were spectating, not getting involved.
we had over 100 paid staffers doing new media
mccain campaign had 2 in primaries, 10 in general campaign

Q: How do you bring a cause focused tech team together?
A: Have a really good cause.
Worked 600ish days - i had five days off.  I survived off of caffeine and cigarettes.
we introduced process as time went on.  You finish your work at 2am and you work on the deploy process until 4.
its important to walk and chew gum at the same time.
37S guys talk about scale later. I think they may take the concept too far.
We're going to buy a block of 30 minutes on all four networks at the same time.
I found out three days before that the URL was going to be on all the time.
I was flipping over tables, freaking out.
Media guys were like "What?" Didn't understand why it was an issue.

Q: Why does the whitehouse have problems doing the same things the campaign did?
political process is hard. don't have control over the process
have to be more careful.
bad decision on the campaign = embarrassing
bad decisions in the whitehouse = people die
we all want the decisions in the whitehouse to be more careful than the campaign
It took a team of lawyers sitting around for months during the transition period to decide that "yes, we can have a twitter account"
9 ways increasing page speed can help your business