Trada recently celebrated its second birthday. It’s been an amazing ride helping our company grow and learning – in real time – about the product that we’re making. Any good organization these days is a learning organization, and I think in general we have a pretty humble attitude about how far we’ve come. While we think we’ve innovated dramatically in the paid search space, we have many things to refine in the subtleties of our marketplace, advertiser onboarding, optimizer engagement, and service delivery.
One of the things that makes Trada both beautiful and complex is that it is multifaceted: it’s a marketplace, a crowdsourcing platform, a collaboration system, and a community.
And each of these elements has at least two sides: buyers and sellers in a market, the crowd versus the consumer, etc. This means that the interactions and incentives between each party must be perfected. The more types of parties or diversity of desires of each party you have, the more complex it becomes.
Since day one we’ve fundamentally believed that we could align paid search experts’ goals and advertisers’ goals to create a positive incentive system. I think we’ve done a good job of pointing people in the same general direction. We‘ve also had to invent – literally – mechanisms to overlay an incentive system on a complex paid search ecosystem – for example, how do you deal with differing bid prices in AdWords auctions? How do you deal with shared keywords or organize ad groups in a collaborative campaign? But while we’ve innovated a great deal, we have been learning. And today I want to announce the second generation of Trada and a concept we call Crowd Mechanics.
By now nearly everyone in the tech space is familiar with the term game mechanics. While it’s existed for a long time in various forms (video games, education systems, etc.) it has re-emerged in technology through location-based check-in services like Foursquare and SCVNGR. The basic concept of game mechanics is that human beings enjoy and are incentivized to keep engaging in a known system of achievements, rewards, levels, and other statuses. I call this technology dopamine – the constant small infusion of adrenaline into an experience that becomes addictive and behavior-changing. At the same time, crowdsourcing has emerged as a new and powerful way of getting things done and the industry has matured at a lightning pace. As an industry, we’re about five years old now (setting aside early outliers like Wikipedia, etc.) and we’re growing hugely. Trada and about 35 other crowdsourcing companies launched the Crowdsortium last month and I believe 2011 will see the first (if not many) crowdsourcing IPOs with LiveOps. An amazing run in just 5 years. But as an industry we’re learning a lot. How do you get crowds to work together? What incentivizes them? What is the right payment system for them? Do they need their own form of game mechanics? What happens when you introduce real money into an incentive system?
Yes, things are different when you’re dealing with real money and when you have a crowd. This is what I call crowd mechanics. The brief definition of crowd mechanics: the incentive and engagement system designed to drive outcomes in a crowd through individual and group incentives that include both monetary and non-monetary rewards, levels, and achievements.
I know this sounds like I’m throwing the kitchen sink into my definition, but its very important to understand how each element of crowd dynamics makes it very different from game mechanics. I’m not suggesting that one is more difficult than another to do well, but there are different variables in the mix that have to be considered. To start, let me explain what I think is the same between the theories:
I fundamentally believe that game mechanics and crowd mechanics share the same basic underlying DNA: they should understand and work with human behaviors. Humans are not one-dimensional, and thus motivation systems (just like the workplace) shouldn’t be one-dimensional either. I wrote a longer blog post about this, or you can watch a video of me talking about crowd motivation.
Now let me outline some differences that need to be considered:
Crowd Mechanics: Money
There is a lot of research that says, “people act differently when money is included in the incentive system”. What’s interesting is that the answer is not always “they work better”, nor is the answer “they work more poorly”. There’s a great TED video where Tad Pink talks about this, and Clay Shirky addresses this in his book Cognitive Surplus. Any way you look at it, money changes behavior. Crowd mechanics systems must contemplate what behaviors they may expect and think through how their crowd is compensated. When money is involved you broaden the general spectrum of behaviors you can expect to see.
On the positive end, you’ll get some people who live and die by working in your system. On the other end, you’ll get some abuses where people try to ‘game’ any incentive system you create. This isn’t any different from designing a sales comp plan or any other traditional comp plan. The comp plan must be designed to make it easier for someone to do what you want them to do (and make what they expect to make financially) than to skirt the system or abuse it. This dynamic doesn’t exist in game mechanics, so the spectrum of uses is much more constrained.
Crowd Mechanics: The Crowd
Depending on the type of crowdsourcing model a business uses, the final product is either the combination of work of other people (e.g. uTest, Trada, Wikipedia) or the best individual contribution from someone in the crowd (like Crowdspring). In our model, we want the crowd to work together. This is something we spend a lot of time on, and we’ll be introducing new features around this soon.
The best way to understand the dynamics of incentivizing the crowd over (or in combination with) the individual is to understand the ‘Tragedy of the Commons’ problem. For a survey on this topic and some suggestions about how governance systems are evolving to handle these situations, read the fantastic Governing the Commons by Elinor Ostrom.
What we’re learning about the crowd:
1) The crowd needs information about itself. Game mechanics has included this mechanism publicly, in the form of leaderboards, because it encourages people to compete with each other.
2) The crowd needs information about its goals. These goals are applicable at both at the individual level and the group level. This is a very subtle point because crowd mechanics gets interesting when some individuals in a crowd are hitting the goal – but some are not.
3) The goals need to be realistic. At Trada, the goal is an advertiser’s CPA. If this CPA is simply unattainable (you can’t get a 50% conversion rate to sales for visitors are your website on a $1000 product) then everyone loses. We’re learning a lot about making sure the advertisers’ goals are achievable as part of the “social contract” that exists between the crowd and its patron.
4) There need to be known group incentives that are substantive compared to individual incentives. For example, a “group win” should not pay someone 1/100th of what they make when they win individually. As much as possible, the group win should be more lucrative than an individual win.
5) Group wins, like individual wins, must reinforce a very small set of core incentive principles. In Trada, the CPA is king and almost all the rewards, achievements and levels are a reflection of this. Group rewards must be based on and reinforce the same core incentive structure.
6) Groups must be able to anonymously socially regulate themselves. We call this the “shoulder tap” – a mechanism where someone in a group can effectively say to someone anonymously “please check your work, it’s way above the goal”. This form of social regulation goes on all the time around us. As a matter of fact, I’m writing this from the ‘quiet car’ on an Amtrak train to NYC. A “shhh” on the quiet car is an example of social regulation and in most cases is anonymous enough that someone in the group is willing to do it.
7) There must be a rules-based regulator that can be called to enforce group behavior. Any group must know that there is a 3rd party regulator (e.g. the SEC, Wikipedia administrators, CJ’s network quality group) that has the power to enforce, in a non-subjective and rules based way, final arbitration policy when someone’s behaving badly in the group (including the patron – e.g. the advertiser – in our model).
There’s a lot going on here and part of the trick is to make the experience relatively seamless. One should be able to perform their work, expert or otherwise, relatively unencumbered by this infrastructure of crowd mechanics but also aware that it’s going on. This is one of the most difficult elements of any game mechanics or crowd mechanics system: that it should be a passive interface underneath the experience, not an interactive part of the experience. Part of what makes Foursquare work so well is the discovery of badges. This discovery element keeps you engaged and exploring the virtual landscape they have created. This applies to anything the crowd engages in.
I think we’ve come a long way at Trada and in the crowdsourcing industry. Our crowd mechanics release is just one of many steps we’re going to take to conquer a massive challenge and opportunity, and we’ll keep learning how to make it better for everyone. We hope everyone that interacts with Trada will give us feedback. We’re not standing on ceremony – and we’re definitely not standing still.