Our research boils down to three elements to matching a circle that is likely to stick together.
- They want to go to a similar destination
- They view each other as peers (or better)
How specific do the common goals of learners need to be? Weight Watchers and Alcoholics Anonymous both have clear measurable goals. Maybe the common goal of a study group in school is for each member to get the highest possible grade. Young Presidents’ Organization (YPO) is less clear – I remember joining with a vague understanding that the forum would help me solve work problems. YPO turned out to cover a much broader range of topics and purposes than that.
My theory is that we should strive for as tight of a goal definition as possible. Let’s start with a group of new data scientists that want to be better at their jobs. But why do they want to be better? For a raise, a promotion, a new job, or more free time? Perhaps we can further refine to tighten our circle on one of these criterion. Perhaps the group shares the goal of wanting to master a particular skill or subject, but hasn’t thought more deeply about where that would lead. Let’s see what kind of data people will give us, and how and if they can articulate their goals.
I think that these goals should be shared publicly amongst the group. Everwise did this recently in their Circles trial, using data about aspirations to form groups. Then showing them and having them discuss the data helped to bond the groups at the kick-off.
One side-bar. This touches on the question of what “pain” will compel someone to subscribe to Circles. I look at this with a framework that marketing guru Richard Currier taught us at M5: “pre-sales” and “post-sales” needs are different. When buying a phone system for their business, people cared about certain advanced features, which they often forgot about after the sale. For customers, service was the most important thing. But that was a weak message when prospects were evaluating competitors who made indistinguishable service claims. This is a tricky game, tempting overselling, broken promises and a confused reputation. A recurring revenue business model, visible satisfaction metrics, and teamwork between sales and service can keep the organization honest. We will test separately what messages get people to subscribe, and what gets them to continue.
The word peer has its roots in the latin “par” meaning equal. Who are your “equals?” This implies some criteria and standards for leveling people. They will depend on whether we are dealing with orchid growers in Southern Florida or managers at big phone companies. So to start, we can anchor the definition of “peer” to a learner’s stated goals. Are they joining to get better at their jobs? Then we should match based on things like role, company size accomplishments, skills, career trajectory and salary.
But perhaps “peer” is a softer assessment. Maybe it is a feeling? More about likability or respectability? The first dictionary definition is actually “a member of the nobility in Britain or Ireland, comprising the ranks of duke, marquess, earl, viscount, and baron.” Are we really looking to hang out with nobility?
Sean Duffy impressed me with his peer group formation rules at Omada Health. They are placing thousands of pre-diabetics each week into peer groups that help make changes to avoid developing diabetes. They match based on three factors: Body Mass Index (BMI), Age and Geography. The groups do not meet physically. Geography and age are a way to build a sense that these other people are like them.
I’ll throw in two anecdotes from the online dating world, which I gleaned from the podcast “Startups” reporting about the company The Dating Ring. Online dating companies have found that people who are close to each other in “hotness” are much more likely to couple. But Dan Abelon, founder of SpeedDate, offered that the online dating industry has not found any magic bullets despite a ton of data. Dan said, “It is not clear that any of eHarmony’s 100+ questions make a difference, but there is evidence that people care that the person they are meeting has been vetted with 100+ questions.”
It is clear to me that all learners must feel that they are amongst peers, so I’m going to monitor their answers to that question as a way of honing our sorting and matching process.
Then there is all the other possible factors. We have to test. Is there a magic bullet in this list?
How do we optimize the commonality needed for a group to jell quickly, with the diversity that will power deeper learnings from the group in the long-term? Do we risk over-indexing for peers or common goals?
Thomas Malone of MIT and Anita Woolley of Carnegie Mellon published a study in late 2010 in Science that concluded that three factors mattered most in team performance, and two were the same. Sharing the communication – equal air time – no wallflowers or conversation hogs – was one. I plan to implement this. High “social sensitivity” and number of women were basically the same. You could try to measure social sensitivity using a test like the “reading the eyes” test online. But maybe it is easier just to try to have more women in each group? As a side note, they argued that group cohesion, motivation and satisfaction mattered far less than these other variables. Let’s see how this plays out in our application.
Shirzad Chamine suggested that the Meyer’s Briggs framework would be instructive. He’s seen a larger percentage of “Feelers” vs. “Thinkers” drawn to work in Forums.
Pragmatically, I think we should match for available meeting times early on. Everyone has an existing weekly rhythm and bucking it would reduce the chances of continuing.
Philipp Schmidt pointed to some Harvard research that entering a study group with one or more pre-existing friends increases the likelihood that the group will jell in the short term.
There are many personality tests. Which should we try? Sebastian speculated that Belbin’s framework of roles within a team might be significant. (Here’s a summary). Roger Schank built a video tool to vet people based on shared values. Enneagrams?
What Do You Think?
We need to be on a mission to come up with few but powerful criteria. What do you think this criteria should be?
Share your comments and feedback about designing the sorting app over on this blog post.Join The Conversation