Saturday, July 14, 2012

Do you know anyone who...?

In the business world, the question "do you know anyone who..." must get asked thousands of times a day. Do you know anyone who... can program? Would be interested in investing? Would be a good fit for this position? Is looking for a job and has this particular skill and background?

I personally get asked this question several times a week. Most of the time it's by a business contact that I'm either friends with or have recently met. Sometimes the question is code for "Are you interested?" But most of the time, when it's used in that way, the sentiment comes bundled with "well if you aren't, do you know someone who might be that you'd recommend?"

Whenever I get asked this question, I struggle to think of names on the spot and often promise that I'll "keep them in mind" afterwards and send anyone I know their way. Although I try to follow-through on this commitment I almost always feel like I'm letting the person down afterwards. In my address book are 728 entries and I probably "know" several more thousand people in Silicon Valley alone.

Not only can I not remember everyone I know, an equally hard if not harder problem is that I don't know what they're all up to. Each person is in a different state of their career progression and while I might know someone with the right skill set to answer the "do you know someone" question getting that person at the right time is difficult.

In a sense, answering this question is an information retrieval problem. The field of information retrieval has the notion of precision versus recall. When people struggle to answer this question they're primarily facing a recall not a precision deficit. In other words, the people they're able to remember are probably pretty good fits (no false positives) but they are likely missing many individuals who would make good candidates (many false negatives).

Why are so many missing? For two big reasons: they can't remember people or they don't know if a person would be a good fit. They're both large components of why the question is tough to answer and I'm not sure which is the larger factor. The first reason is a consequence of how the human brain works. It just isn't made like a computer database, able to lookup thousands of entries in milliseconds by a particular criteria. The second is the lack of perfect information. You might remember that Jane would be a good fit for a particular position, but is she looking for a job right now? Or is she happy as a product manager at Facebook.

Finally, even if you have remembered someone and have recently spoken to them (thus having perfect information) the actual assessment of whether they match the criteria asked in the question is a difficult and fuzzy one. Would an engineer who became a product manager be a good fit for a role at a venture capital firm? It depends. Notice how this assessment of relevance is more complex than picking a restaurant, where the most important relevant factors are likely only location, price, food type, and star rating.

To summarize, why is answering the question "do you know anyone who..." difficult:

  1. Recall - it's impossible to remember everyone you know, especially when trying to filter according to a particular criteria. Your brain isn't designed to sort through thousands of entries and find the correct ones.
  2. Missing or outdated information - even if you remember a person, you might not know that they're looking for a job, unhappy in their current position, or would be enticed by a new role. What's worse, this information is constantly in flux. You might have checked in with a person a two months ago but now they've found a new job.
  3. Complex criteria matching - assuming you've remembered a person and have perfect information about their situation, making an assessment about how well they satisfy the criteria of the question is a difficult one. 
I personally don't like the feeling of not living up to my commitments. I also like the feeling of genuinely having helped someone else out. The solid referrals I've made in the past are some of the more emotionally rewarding "favors" I've been able to perform.

Some of my recent brainstorming has been around building a tool that would better help me (and hopefully many others) better answer this common question. And I believe before thinking about solutions it's best to understand deeply the problems associated with answering it.