I was lucky enough this morning to hear Nancy Lublin’s opening keynote at the Data On Purpose conference hosted by the Stanford Social Innovation Review. The conference theme was about using data for social change and sessions were addressed the growing proliferation of data in the nonprofit sector and the sector’s increasing ability to make productive use of it, guard it, and share it. Nancy gave an inspiring, humorous, and insightful talk, sharing her experience using data with DoSomething.Org and Crisis Text Line. She totally blew me away!
Here’s a curated transcript of Tweets from her talk, including the one above that she wanted everyone in the audience to tweet. Here’s what I learned from Nancy’s inspiring talk about data for good:
How It Got Started
Six years ago as the CEO of DoSomething.Org, one of her board members was speaking at conference and talked about that the birth of Web 3.0 was all about data. Data was everywhere and it was important to use.
She wrote a memo to the board telling them that their organization had lots of data but they were not listening to it. She wanted to hire a Data Scientist. The West Coast part of the board said yes. The East Coast part of the board asked how much will it cost?
Her answer, “Less money than if we don’t do it.”
She hired Bob Filbin, Dosomething.Org’s first data scientist. She confessed to the audience she thought data science was about measurement and accountability. But the real words to describe how the organization thinks about data, “It made us better at learning and growing.”
She shared that Bob likes to talk about nonprofit data scientists as being dot organisms. Just like your brain, you have nerves everywhere that tell you to take voluntary actions like “raise your arm.” Or involuntary actions, like breathing. Great data in a nonprofit will work the same way – you won’t even be conscious of it.
Shortly after, they hired a second data scientist, Jeffrey Bladt. Nancy joked that “It is good to have two data scientists like it is good to have two Mormons because less beer gets consumed.” They disagree a lot and that is good.
Hiring data scientists help the organization, DoSomething, collect, store, and analyze data. They also had to think about security and privacy in new ways. And, they had real time data – they could be in the moment and understand what was going on. “It doesn’t mean that we’re going to do a survey once as a fancy report for a Foundation. It is ongoing.”
A Culture of Iteration
Nancy used the analogy of operating like car company that puts out different vehicle models and each year will have a different version with different features.
This mean that they had to think of social change as campaign templates so they could do more systematic iteration and learn from it. “If you are constantly changing, you are not iterating.”
DoSomething is the largest youth organization in the world with 3.8 Million members who are young people. As Nancy likes to say, “Bigger than the Boy Scouts because we’re not homophobic.”
Their primary communications method is text and that lends itself to templates. Because text is limited, they A/B test the messages they put out and learn. They can also analyze the message they receive back and learn. They don’t ask open-ended messages, but asked close-ended questions so they can measure.
They run many campaigns and created 7 different types of campaign templates. For example, one of them is “Collections.” Young people like to collect stuff.
For example, they ran a campaign for teens to collect food for a Food Banks around the country. Her team spoke with Food Banks and learned they didn’t need more soup, but non-perishable protein was needed. So, they launched a campaign with “Team Crunchy vs Team Smooth.” They collected thousands and thousands of peanut butter jars.
Another example was to help Youth Homeless Shelters. They interviewed them and found out that homeless teens want one thing – jeans. So they started a partnership with Aéropostale and campaign called “Teens for Jeans.”
They know from their data that the Shelters used the jeans to help lure the teens into the shelter for the night. Kids learned about homelessness. Moms were happy that kids were cleaning out their closets.
All of their collection campaigns are done on a template – where they have iterated on what works – from the messaging, timing, etc. They iterate and get better. The bottom line is member acquisition – so they know the one metric that matters.
Focusing on KPIs
Nancy mentioned that listening and using data inspired them to kill a bunch of programs, even popular programs. For example, their bootcamps which were sponsored by American Express.
The program was a lot of fun, but when they looked at the data, it was not scaling. Only 150 people came to the event and while it was a great experience, it didn’t recruit new members. Nancy said, “If a program doesn’t send us new members, it is outside our KPIs. If a program is outside our KPIs, we don’t waste time or money on it. “
“When you have clear KPIs, there is less wasted time and everyone on staff is focused on member acquisition. When you are focused on that one thing, you can look at what is working more honestly.”
Nancy says they listen to their data and members more – and funders a lot less.
Text Crisis Line
Nancy shared the story of the how and why Text Crisis Line launched. At DoSomething, they were getting text messages back from teens. This one stopped them in their tracks:
“My dad is raping me. He told me not tell anyone. R u there?”
They sent information about a rape crisis line, but didn’t hear back. They don’t know what happened, but Nancy tells the story in the hopes that the person knows what an amazing project they inspired.
They built the crisis text line to be data driven. The first hire was a Data Scientist and CTO. It was designed and built on the notion of data collection and analysis. In August 2013, they sent an opt-in text message to area codes in El Paso and Chicago. Within a few months, the list expanded to all area codes in the US.
The most reference word in the text responses from users is the word “today.” It is used 6 times more often than other words. It means that the Crisis Text Line is reaching teens at the moment of crisis, an effective way to provide counsel.
It is private and teens texting us about everything from eating disorders to suicide. In fact, 30% of messages are about suicide compared with 3% about bullying.
The Text Crisis Line has an algorithm that analyzes priority requests. The words “I want to die,” will get the teen to the top of cue. Another stellar example of how data is being used for social good. Most of the time, they can get a counselor on the text line right away and have them put the pills in the drawer.
If they get silence, they can track them and usually the next text is from the parent saying their going to the hospital. The number of suicides happens more than you would like to think.
Data Sets: Purpose, Volume, Velocity and Variety
What is starting to happen now is that their data base of 6.6 million messages lets them do predictive analysis. For example, if a text has the phrase “nums and sleeve” it means the teen is doing some cutting. If it has the words “sex, Mormon, and oral,” 99% chance the teen is questioning LBGT.
One of their researchers, someone from India, came into Nancy’s office and said, “The algorithm is broken is just combined having sex with giving up the goods.” Nancy told the researcher that the algorithm understand American slang better than you do!”
The data set has been stripped of personal information and now available on a creative commons license at Crisis Trends.
What this means is that can analyze questions like: What is the worst day of the week for eating disorders? It’s Monday. So, what is the school’s cafeteria serving? Are there guidance counselors? Nancy said this data set could turn on the lights for every issue – journalism, police, schools – information to change everything for the better.
This is what she meant by: Data scientists are not overhead!!!
What’s good? What’s impact?
Nancy said that funding closes their organization off from innovation because funders often held them to what they wrote in the grant proposal 2 years ago. “It makes shoot for the number the funder established and doesn’t let us pivot.” We now do what works for us, not works for our funders.
For example, they partnered with established crisis centers and did a lot of a/b testing with data of “best practices.” For example, “counselors should not use “I” statements.” But their data found that when the counselor says things like, “I hear you,” it inspires more text messages.
In other words, when you test things, it makes them better. It is about what works, not what has always worked.
They are now working with volunteer counselors who go through 30 hours of training and do 4 hour shift once a week. Not everyone is cut out for this, so by measuring their progress they understanding “melt rate” (people who drop out of training) and “churn rate” (people who drop out once they’re trained.) Nancy said, “We know exactly who is a good candidate: twenty something lesbians and vets.
She also noted that funders ask them for their outcomes, what happens the week after they text you. They don’t want to re-establish content after words because it trigger depression again. “The text crisis line is like 911. Do they call a week after your house burnt down?”
Nancy’s talk was inspiring, but more importantly an extraordinary example of how an organization is data informed and how open data can be used for social change. What inspiring examples have you seen where nonprofits are data-informed or open data for social change?