The Failure of LinkedIn - Quality
During a discussion over our first grilled meal of the year (brats), J-- and I got into a discussion about LinkedIn. I asked J--, as a recruiter, what was she looking for in a recommendation? As we talked through the ins and outs of a telephone conversation with a listed reference, it become readily apparent that a certain quality in a recommendation was needed to lend credibility. For instance, there was a base assumption that the person to whom she would be talking was hand-picked to deliver the best possible recommendation for a candidate; that's why the person made the candidate's recommendations list. But it took a very special level of interaction to raise that recommendation to a level where it could push a candidate over the top, and much of that interaction started in the initial interview with the candidate.
On the Web, this scenario doesn't play out much. Take LinkedIn for example. If you are hiring someone and find their LinkedIn profile during a Web search on the candidate, recommendations for that candidate really carry little weight with a recruiter unless there's some previous value placed on the recommending person (say, the recruiter happens to know the person giving the recommendation). Otherwise, the recommendation carries varying levels of implied authority from Former Boss Who Really Liked Candidate to Random Co-Worker Who Got a Reciprocal Recommendation. There is no quality behind a recommendation.
Many recommendations succeed or fail on the ability of the person writing the recommendation; someone who is articulate and efficient in their writing can craft a recommendation that sounds fantastic. What it doesn't do is provide what a recruiter should be looking for; corroboration and authority. J-- and I joked that we should just recommend each other since our last names are different, but that raised the very real issue that that is likely happening in a dozen shades all across the LinkedIn network. A recommendation on anyone's profile is without authority on it's face, and the probability of talking to the recommender to put some authority to a recommendation approaches zero.
Now, I pick on LinkedIn, but this isn't unique to that site. The LinkedIn model is ripe for exploitation just as the analog model of resumes and recommendations is. Most, if not all, interactive "Web 2.0" sites have this same problem; Digg, StumbleUpon, and their ilk have been gamed forever. The challenge for these sites isn't the content model, it's the arms race of staying ahead of the spammers and exploiters. For LinkedIn, the challenge isn't spammers, it's trust, a battle they are winning. The anecdotal success stories of someone being found through LinkedIn are the equivalent of winning the lottery: it won't happen to you.
As we chewed through brats and the issues around quality in the LinkedIn recommendations, pretty much every idea we had was open to gaming:
- Allow co-workers from the same company with overlapping employment periods to flag reviews: too easy to say you worked together when you didn't, open to crowd-based reprisal attacks, liability laws.
- Certify recommendations through a third party: who would be trusted by both candidate and potential employer, who'll pay for it.
- Highlight recommendations by ratio of recommenders to number of employees in the recommender's company (10 of 20 is better than 3 of 10,000): could argue that the opposite is true just as well, no strong case that either is true.
- Build algorithms to look for suspicious activity (X or more recommendations from same employer in Y period of time from similar IPs): it's implied suspiciousness, not proof of gaming.
And, the last nail in the coffin, not everyone will have a profile, so it's not a reliable tool for a recruiter anyway. In the end, we didn't come up with any way to provide a tool for a candidate to make a recommendation anything other than the digital equivalent to the analog list of references. In fact, it had a little less value because the candidate can control the recommendations whereas the current model produces some surprising conversations with listed references. (Yes, people list references that give bad references.)
I'm not saying "don't have a LinkedIn profile", I'm just saying that all those recommendations may not count for quite as much as you think. Or, if you have a way to solve the quality issue, drop me a line and we'll be millionaires together.