Sunday, July 11, 2010

Unvarnished Unwrapped


A few months ago I mentioned a website called Unvarnished, which was getting a lot of mixed press. The basic concept is (as they describe it) Yelp mixed with LinkedIn. You provide anonymous reviews of people you've worked with. Call it a social resume, call it a web-based reference, I call it fascinating. And something that anyone interested in recruitment and assessment should pay attention to.

I had a chance recently to test out the site, and then I met with the co-founder, Peter Kazanjy. I'm still not sure which direction this will go, but I think you'll agree after reading what follows that the concept merits our attention.

The Test Drive
First, the test drive, which started with an invitation through Facebook from Peter. After spending some time on the site, I'm more optimistic about Unvarnished in some respects, more cautious in others.

Why optimism? The site hits it out of the ballpark on two accounts: it's simple and fast (at least someone learns from Google). Simple and fast is good, because one of the biggest challenges will be building a large community and making reviews easy helps immensely.

Even better, the ratings are relevant. This isn't a popularity contest, it's an honest attempt to provide a useful description of someone's performance. While it's unlikely the rating scales were developed after reading a Personnel Psychology meta-analysis, I was pleased to discover that they pass the smell test and some even have benchmarks.

The ratings consist of an overall performance rating (5-point, anchors described), what job you are rating the person in, four 10-point scales that are described but not anchored (skill, relationships, productivity, integrity), and an open-ended strengths/areas for improvement box. That's it. It's like a super-basic reference check form that takes all of about a minute. You can see what this process looks like below.

Your Unvarnished homepage is fed by your network and generates suggestions for your review (PMYK--People You Might Know). You can review people at any time (even people who haven't claimed a profile) or request reviews using your Facebook contacts. The open comments section is limited to 500 characters to encourage people to review and move on.

The site was developed to be heavily reliant on algorithms. A reviewer's reputation is based in part on the pattern of reviews they have generated as well as how their reviews have been rated. Recommendations for reviews are backed by similar math. A smaller (but important) feature is a profanity filter, which may allay some concerns regarding people looking to settle a score.

Speaking of Facebook, one of my major concerns is it relies HEAVILY on Facebook (not unlike Quora). At least in its current iteration, your identity is verified through having a Facebook account, you invite people through your Facebook contacts, and invitations are posted to Facebook. This is both a potentially good thing (e.g., cuts down on spammers) as well as a bad thing (e.g., not everyone wants to use their Facebook profile in the service of another website). It also begs the question of what would happen if Facebook went belly up. Read on to see what Peter had to say about this.

The interview

I happened to be in San Francisco (interestingly enough to go the Exploratorium) and had an opportunity to talk with Peter about the product and his company. After talking to him for about 10 minutes, one thing became abundantly clear: this guy has thought a great deal about online reputation management. I'd love to get him and Bob Hogan together.

Background
The impetus for the site came from a couple directions. One was a previous job where he was continually surprised that competitors (e.g., a certain company we'll call Bicrosoft) would recruit away the lesser talented employees. Why? His theory is they lacked important information--namely the reputation people had within the organization. (One could argue that they should have done better reference checks, but we all know how easy/productive those can be)

On the other end of the spectrum, he saw excellence not being properly recognized and questioned whether upper management really knew what their talent looked like (a point not lost, btw, on purveyors of performance management software with their 9-box grids and helicopter views).

His ah-ha moment (or one of them) came when he realized that if social ratings can work for things like books and software, couldn't they work for people? If he could develop a site that aggregated high quality information about people's performance, talent decisions would be higher quality as well as more fair (it's hard to argue with that goal).

Peter also believes there is an important employer segment not being served by existing background/reference checking processes. Employers that hire hourly workers rely largely on criminal/credit checks. Those hiring for executive-level positions often rely on high cost search firms. But for employers hiring large number of employees in the middle, there isn't really a good option.

Such a site would have three primary users: those being reviewed, those providing reviews, and those using the information (e.g., employers checking out candidates and vice-versa). And quite commonly a single individual could be in all three roles at various times. The site would have to accommodate all three perspectives.

Concerns/criticism
So what about my concerns? When it comes to the reliance on Facebook, Peter pointed out that it's a good bet Facebook will be around for a while, but the site is not being built to rely solely upon it. It has--or will have--the ability to use contact information from sites like Gmail and LinkedIn. I still have a concern about forcing people through Facebook, so it will be interesting to see whether this impacts the ability to generate reviews.

Concern from others has focused on the potential for abuse. But Peter made several important points. First, this isn't like an online newspaper comment space--those are anonymous with no repercussions for inaccuracy. In Unvarnished your reputation would suffer and your reviews become less valid (assuming your reviews are themselves reviewed). Second, this information is largely out there (or potentially at least) in the form of places like Twitter. But it's not in a central location that can be easily managed, and it's not objective. By guarding invites and relying on anonymity, the goal is to make legitimate reviewers feel safe in leaving honest feedback (whether it's an A+ or a D-) without worrying about the interpersonal implications.

What about stickiness--why would someone want to keep coming back? They're working on several (re)engagement initiatives. One idea is to provide people with periodic updates letting them know when people in their network have been updated (let's hope it doesn't turn into those updates from LinkedIn that are easy deleted). They're also working on the ability to follow an individual based on your "gestures" (e.g., you've reviewed them).

Future directions
The team is considering adding several features. One is the ability for reviewers to better define their relationship--e.g., describing not only the organization they worked together in but their relationship. This would factor into behind-the-scenes algorithms but would not be published.

They're also discussing allowing people to identify themselves, but this raises all of the associated issues such as accuracy, thoroughness, and feeling a need for reciprocity.

Another is allowing access to premium features (e.g., at a cost) such as making trusted reviews more obvious, something that "super users" such as recruiters would likely be willing to pay for.

In terms of opening up access, they're in no big rush to expand access beyond Facebook invites. While this may hinder their growth, it helps keep the data quality high, and they're willing (smartly, I think) to make this trade off.

Conclusion
Currently the company is focused on acquiring the talent it needs to succeed (check out the way they recently advertised for new engineers). One of Peter's primary concerns is that the community evolve in the right direction. Right now it's somewhat of a "love fest" with lots of positive reviews. The site will gain in usefulness when reviews are a combination of pros and cons.

Given what we know about performance ratings, it will be interesting to see if the existing invitation and rating process is sufficient to generate that depth. It also remains to be seen whether some type of incentive will need to be given to generate reviewers.

My overriding concern continues to be the size and diversity of the user group (right now primarily filled with Silicon Valley IT folk). Accuracy, something other writers have been obsessed with, is less of a concern for me after kicking the tires and talking with Peter. But we'll see how the promise and concerns ebb and flow with the user base as well as changes to the service.

At the very least, I hope you'll agree with me that this website is a fascinating development and one we should watch. After all, you might want to use Unvarnished to provide feedback on someone. Or research a potential boss. Or research an applicant. Or...you may be the applicant.

2 comments:

Anonymous said...

Great article on Unvarnished. Full of thoughtful comments, keep them coming. The UV team relies on community feedback; pro or con. (btw, I have the same concerns about the linkage to FaceBook)

Amandapanda801 said...

Oh wow... I think is site is scary. I don't think companies should be able to read what other people say about a co-worker because, to be honest, not everyone likes their co-workers and different people are going to think different things about one person. Also people might just say bad things about someone in order to spite them so I really think this site is a bad idea. I mean if you want to try to see the potential of a cadidate you should just try psychometric tests to assess their personality and stress levels. Central Test is a psychometric testing site and at least only the person giving out the test can see the results. http://us.centraltest.com/