It is tantalizing to envision what the nonprofit sector could achieve with the vast amounts of data and sophisticated analytical tools that already exist in the commercial realm. After all, if Target is able to discern a woman’s first trimester pregnancy from changes in her purchasing habits in order to send her specific promotional offerings, just think what we could do to meet the individualized needs of at-risk students, or find the right opportunities for those who are unemployed.
Already, the nonprofit sector is being flooded with data. The 990 tax returns will soon be searchable online. Guidestar, Google, and Charity Navigator are among the many organizations vying to help donors find and assess charities. Evaluations, scientific studies and research reports are piling up faster than anyone can possibly read them.
But what if the data we have is not the data we need? Suppose solving society’s problems only requires a little bit of data that is simple and easily understood. Considering three purposes for which data can be used:
- Understanding (e.g., What teaching techniques are most effective?)
- Selection (e.g., Which program or grantee should I fund?)
- Behavior change (e.g., How does my performance compare to others?)
All three are useful in the social sector, but being useful is not the same as being used. It is estimated that less than 1 in 100 dollars of government funding is based on any evidence of effectiveness, and many billions are still spent annually on programs known to be ineffective or even harmful. Without the motivation of profit and competition, it seems that our understanding of what works is often left on the sidelines, overtaken by politics, false assumptions, forceful personalities or simply inertia.
Using data for selection seems to offer great potential. With all the information at our Internet fingertips, we ought to be able to identify the most effective charities and target specific programs to those who would benefit most. Unfortunately, this doesn’t happen either. Programs remain generic and only 3% of donors today use any research to decide where to give their money. This may change over time, but we have a long way to go.
The third use of data – to change the behavior of people and systems – is the simplest and the least discussed, yet potentially the most powerful because it is not easily ignored.
For data to change behavior, it must be shared – that is, everyone involved must be looking at the same information. It must be timely and linked to actions, so that changes in behavior show up promptly in the results.
This type of data changes behavior in two ways. First, it helps different players align within a system. Most of our major social problems are rooted in poorly functioning systems: the education system, the health system, the prison system, and so on. Yet most of these “systems” are composed of independent players who have little awareness of how their actions affect the overall result. Focusing all participants’ attention on the same performance data improves the overall system.
The juvenile justice system in New York State, for example, involves countless different players: police, judges, probation officers, social workers, community advocates, etc. Within three years after developing shared data about all aspects of youth arrests and detentions, the rates of youth arrest and incarcerations dropped 22% and 49% respectively.
Second, competition is a powerful motivator. School principals in Seattle didn’t pay much attention to whether their students enrolled in a state scholarship program, until the school districts began to track and compare the number of students who signed up. Within a year the percentage jumped from 51% to 94%, simply because no school principal wanted to be at the bottom of the list.
The kind of data that changes behavior is unfortunately the one type of data we do not have and rarely discuss. It is being eclipsed by the excitement around data mining, randomized control trials, and advanced analytics.
All data has value. We will always need to build our understanding and improve our selection. But the very simplest uses of data – to inform participants across a complex system and to awaken the competitive instinct — may yet be the data we need.
