Will the Revolution be data-driven?
by Milan de Vries, outgoing Director of Analytics, MoveOn
At the end of this year, I will leave MoveOn and my role as Director of Analytics.
The timing is right for me to do that — my kids are young and I’d like to spend more time with them. After eight intense years, I’m also ready to rest up before a next adventure. Reaching that decision, though, wasn’t easy because for me this has been a dream job, with a dream mandate — to harness data and technology in service of excellent grassroots organizing. And I can’t imagine a more special place to do that work than here at MoveOn.
I leave with love in my heart for MoveOn and the colleagues who will carry on this work. As I prepare to set aside the day-to-day for now, I look forward to staying connected as a member and close-in alum.
I also find myself looking back. I’m proud of the analytic advancements we’ve made in this time and proud of a great many of the projects I’ve been part of. I’m reflecting, too, on some things that I’ve seen change in our politics and country in that time — and some things that have held constant.
Like the ongoing debate — reignited every election cycle in particular — over whether organizations and campaigns rely way too much on analytics or not nearly enough.
What my time at MoveOn has taught me is that this debate almost entirely misses the point. The question shouldn’t be how much analytics to do but whether any of it is any good.
Here’s what I’ve learned in my time at MoveOn about what good political analytics looks like.
Good political analytics listens
Good analytics doesn’t just hear, it listens.
If you want to know what five people think, you can just ask them. If you want to know what 500,000 or five million people think, you need data.
In 2011, MoveOn launched a people-powered petition site. Hundreds of people a week create petitions about issues that matter to them and their communities. A key element of this project at MoveOn is that if a petition resonates strongly with MoveOn members, we can share it with members in that area and, in doing so, amplify it from a petition one person thought of to a campaign that a community of activists works together on.
To listen to the input of MoveOn members at scale, we built an automated testing system that sends petitions out to small numbers of members for feedback — measuring both positive and negative opinions — and then iteratively sends the strongest petitions to more and more members to find which petitions were of the most interest to MoveOn members in different areas.
Here analytics lets us sort through hundreds of petitions a week and understand what matters to members. That’s the power of data to listen.
Great political analytics anticipates
Great analytics doesn’t just listen to the world as it is, but to the world as we want it to be.
A dashboard, alone, will give you a snapshot of the world — how many active members you have or where they are located. It won’t tell you how much potential different members might have for becoming more active or what to do to make that so.
In early 2017, there was an explosion of grassroots energy around the mass mobilizations of the Women’s March. We, as an organization, knew that not every weekend would look like that weekend, and we, as an analytics team, asked ourselves how we could bottle some of that energy for later.
We decided to run a free sticker distribution project with the premise that we’d get people to put stickers on their cars in the short term, and they’d still have them as reminders months later when the energy of that weekend had dissipated.
Nobody was asking for a sticker as a follow up to that weekend, and certainly nobody was asking to be reminded months later of their activism earlier in the year, but we felt, coming out of that weekend, that that was a useful goal and knew we could use analytics to achieve it.
We ran a combination of different survey types to test twenty different sticker designs and find the ones that most resonated with people in the moment and which they were most likely to put on their car. We optimized the sticker giveaway through email, SMS, and social media, and we used testing to develop a balance of multi-pack sales and back-end donations to continue to be able to give away most stickers for free.
In the end we sent out over 300,000 RESIST stickers, at break-even for the project. As I drive around my town I still regularly see the stickers on cars, a small reminder to all of us of the collective energy we saw in force that weekend, made possible by a data-driven intervention months earlier.
In my time at MoveOn, some of the best campaigns we’ve run — from supporting diplomacy with Iran, to advocating for ending student loan debt, to encouraging Elizabeth Warren to run for president, to helping the labor movement fight for $15 per hour minimum wage, to combating Islamophobia — were predicated on ideas that hadn’t manifested yet.
Great analytics is driven not by the data you have but by the ideological outcomes you need. That’s the power of analytics to anticipate.
The greatest political analytics has impact
Really great analytics doesn’t just listen for what the world could be, but helps make it so.
All of the analytics we do should be in service of the campaigns we run, but sometimes organizations or campaigns, and even analysts themselves, allow analytics to be relegated to a role of reporting passively instead of helping actively.
The last project I did at MoveOn was a large-scale digital persuasion program for the 2017 special elections and 2018 midterms that we called the Real Voter Voices project. In this project we collected over 2,400 unscripted, self-made videos from activists, talking about why they’re voting and why they’re supporting their local Democratic candidate for Congress or governor.
We then developed an in-house persuasion measurement methodology, native to Facebook, that let us determine how persuasive each video was and which voters it would motivate. Using that data, we could build specific targeting models for each video that allowed each video to reach the voters it was most likely to move. In essence, we could take the digital equivalent of thousands of canvassers (in the form of authentic videos) and use data to let them knock on the (virtual) doors of exactly the voters they could most persuade.
We also measured the real-world impact of our work through randomized-controlled trials and could see that our targeted, authentic videos moved thousands of votes in close elections. That’s the power of analytics to have impact.
A parting thought
The debate about how much to rely on analytics in politics was always bound to become a debate about quality. Do bad analytics and people won’t want more. Do good analytics and it’ll make itself wanted.
As an analytics community, we should embrace that debate. The debate about what makes political analytics good.
As I prepare to leave MoveOn, I look back at the work we’ve done and feel proud. But I also see that there’s so much more work to be done and so many challenges our country still faces.
Rising to those will take organizations and candidates that listen genuinely to people, that anticipate ideological possibilities, and that drive relentlessly towards real, impactful change. Supporting that with data will require analytics that grounds itself in the same.