The Research Behind the Making of MeHive

Social Science and Relationship Management

First pitched as Woo Who at one of our earliest StartUp Weekends, SEP began a project, MeHive, which was a relationship management tool. This project gave us an opportunity to work on a product, and to develop in iOS and submit to the Apple app store. It was also a great opportunity to provide a sizable number of interns with an interesting and valuable experience at SEP.

I did some of the social science research on managing relationships as background for the MeHive app. One of the posts I published around that time was about the Dunbar number – the number of people an individual can reasonably care about and keep in touch with – about 150 people. What if we could build a tool to help people manage those relationships, or even expand that number? Could the tool figure out what relationships are getting stale, which ones to rekindle, which ones could withstand long times between connections? The idea was to improve on a basic CRM app by bringing some smart selectivity to the list of people a C-level person might want to maintain relationships with. What are the attributes of a relationship one could use to construct an algorithm for a smart relationship management app?

Attributes of Relationships

Digging into the social science research, one discovers an entirely new vocabulary for often familiar concepts. I won’t cover all of my findings here, but to give a rough idea, it covers concepts such as symmetry (whether two people are on an equal level socially (friends, coworkers) or unequal (father/child, boss/employee). Another concept is homophily (the extent to which people form ties with people similar to themselves in terms of age, gender, race, occupation, educational level, etc). One of the most interesting concepts was relationship-as-story: every relationship has a beginning, middle and end; a relationship that fades has essentially ended a story; another story begins when a relationship is renewed.

After learning as much as possible about the social science research behind relationships, I came up with a fairly limited number of attributes describing relationships in a work setting, and a set of attributes describing the nature of contacts between people. They are:

• Story – We just met and have no story yet; we have an ongoing story; a story began, but is sputtering.
• Status – I am the weaker; we are equals; I am the stronger.
• Strength – Have we worked well together or not?
• Duration – How long has the present relationship lasted?
• Memory – How well do I remember my interactions with this person?
• Kickers – something unusual about this person that I should capture, such as a job change or family life event.

Attributes of contact

• Direction – Did I last contact this person, did they contact me, or were we part of a group together?
• Direct or Indirect – did I speak to them directly, or did someone else from my company?
• Media – email, phone, in person, small group or large group
• Type – This is a little nuanced: was my last contact a simple “thinking of you”? Was I thanking you for something? Apologizing? Requesting? Replying? Sending a quote or contract?

By assigning values to each of the different attributes, we can then assign numerical values and construct a simple algorithm, much the way the Apgar score reflects the health of a newborn approximately but quickly. If we assign large numbers to aspects of a good relationship, not currently needing attention, such as:

• Story – We have an ongoing story; gets a 4 out of 5.
• Duration – I have known this person five years or more; gets a 3 out of 3.
• Strength – We work well together; gets a 3 out of 3.

and small numbers to aspects of a relationship in need of attention:

• Memory – I don’t remember much about this person; gets a 1 out of 3.
• Direct or Indirect – I have not spoken to them; someone else at my company made last contact; gets a 1 out of 2.
• Media – We were in a large group together, such as a conference; gets a 1 out of 5.

Making the Algorithm

The algorithm can assign aggregate scores to our different relationships. The app tracks the time since last contact with each person. It can compare aggregate scores to an average score.  The software can check for low scores and alert the app user of which people s/he needs to get back to, to strengthen the relationship. When we do connect with a person, and record the nature of the most recent contact in the app, the app can find new people in our contacts with whom we should re-establish relationships.

By keeping data entry down to a few variables with choices 1-3 or 1-5, we hoped to make the app both more friendly and more useful. By adding some social science value, and keeping it simple, we hoped to make managing relationships a little easier.