Insider View: How President Obama’s Campaign Used Data and Analytics to Rally Individual Voters
Today, communications professionals and public affairs officers must be masters of data and analytics. This capability goes beyond measuring the impact of an organization’s communications and public relations—as useful as that is.
Communications professionals need to build expertise and capabilities in data analytics into communications planning and programs as well as capabilities to understand data in order to take effective action—and do so in real time.
These excerpts are from Rayid Ghani’s presentation to the 4th National Summit on Strategic Communications (www.strategicsummit.com) on April 23, 2013 in Washington DC. Ghani was “chief scientist” for data analytics in the Obama for America 2012 campaign.
– – – – – – – – – – – – – – – –
One of the data analysis tools that we built was an idea from somebody in my team who knew nothing about communications. He was a computer scientist who thought it would be useful if we took all of our speeches to see how they were being used; the earned media was around the messages.
So we wrote a tool where the interface was pretty simple. We took a speech, the text or the transcript of the speech and made the font size of the words proportional to how many times the text was repeated in the news. You could break it down by battleground state versus non-battleground states. You could color the words by how often it was quoted in a positive context or a negative context. You could take a speech that you wrote just yesterday, and it was delivered the night before and now you could see based on all of the newspaper coverage how much earned media we got out of the speech and which parts we covered.
So Florida was really covering healthcare and Medicaid but North Carolina was not covering that at all. You could get a very quick glance at your messaging in the news. When you’re writing a speech, you set goals. You want to communicate these certain things. Then we looked at the earned media report to see which message goals were accomplished.
So we were getting data from every little newspaper that we could find in any battleground state, and we were breaking down everything into what issues were about. Is it positive? Is it negative? What are the candidates talking about? One of the lessons that we learned was there are a lot of these things that are very easy to do technically but very hard to communicate to the internal users in the campaign.
It’s easier in a company when you have time to sit down with people and spend time to talk about use of a tool like that. It’s much harder when you’re doing things at the same time as you’re building things and launching tools. You can’t take a break and say, “Let’s just now sit down and figure out what the best way of doing this is.” It’s really critical to get people to use the tools they need and really embed those tools into their everyday processes.
Now, to get to that point we did have to first explain to people how the tools work and get trust and buy-in, but once we had that, it was part of their everyday process when they wrote and sent email.
We also designed a tool around email. When an email was written, we had a system that enabled our email team to target the top 20% of people who should receive the email. Our system would automatically score everybody in our email list, pick the top 20% of people who were likely to act on the email message, and people could just click and send and they didn’t have to know anything about how the system worked in the background.
We were able to completely automate the work so that even we as an analytics team didn’t need to be directly involved in the execution.
We did lots and lots of number crunching, and every night we would update our models to predict who the right people to receive an email were? We did a lot of email because email is, for a campaign, the primary mode of fundraising. That’s what a campaign does with email: fundraising.
We probably sent about five billion emails over a year and a half; maybe a little bit less or more, but about five billion. There were links in the email to contribute. Those five billion emails resulted in people are clicking on links and contributing between $500 and $600 million through email to the campaign. People clicked on the email because it was in their mailbox anytime they opened their email.
We ran a lot of experiments to figure out how to personalize email. In beginning, we did a lot of experiments with demographic data, sending email to people of different ages, gender, ethnicity, and location. It turned out that it didn’t matter that much if you tailored email to demographics. The email that worked well for older people in the south also worked well for younger people in the north and the Midwest. A good email was good for everyone.
Now, where it did matter was if you could personalize messages to people. That mattered quite a bit. We created segments of our lists around people who, for example, never opened our emails. (I don’t blame them but they didn’t.) We wanted to distinguish people who weren’t opening our emails versus people who were receiving email in a spam folder or had different email addresses would never check some of the email.
So we ran experiments around having subject lines that just made people open emails. We didn’t really care if you read the email or clicked on the embedded links. We just wanted to confirm their location or contact information.
We also die experiments to maximize the open rate. Our goal wasn’t to get money. Our goal was to move people to the next level—people who were often opening email but not clicking on links in the email. So we would send them a video with a first frame visible in the email. So they had to click on it to see the rest of the video or we’d send them a short email with a link to click to see what happens next.
Then for people who were clicking on links, we’d try to get them to donate. We tested the right amount of money to ask for? Who sends the email? We did a lot of email frequency tests. At first we thought we were sending too many emails so we started reducing the volume. It turns out we are losing money because we did that. So then increased the number of emails and we made more money, so we started sending more email. The tests helped us optimize our efforts.
This experimentation especially with digital communication is extremely important, but you have to be very careful to do the right experiments. So trying to prioritize what I test and what I don’t test and what’s the impact of the text is important not just in my online world but also the rest of the world that I live in. That’s extremely important.
Hopefully what I’ve shown is that in campaigns, there are a lot of different functions where analytics is helpful. The key is doing two things really well. One is having an integrated data source where you can do all of these things, because without having a single place where you’re getting your online data and your offline data and your social data, you can’t really take actions that are consistent across your different communication channels.
The second important thing is we had access to people who could take actions based on the data and analytics were provided—having the right data in place, the right analytical tools and having people to take action.
The analytics is what’s in the middle. It helps you use the data to really inform decisions in the campaign, so people can take some action whether it’s automated or human
That is the big picture to think about.