socialdna, power laws, information cascades, winner take all, network effects

Superstars!

This second lecture is where we pick up the thread from the MusicLab experiment and finish the rest of the story. There is more than one way for a Superstar to emerge than just social influence. In fact it has not always been thought that such an 'irrational' process could drive success (we are Homo economicus after all).

So the first thing that we should clear up is why social learning isn't such an irrational process, then we'll discuss other processes that will result in a Superstar occuring.

Social learning and information cascades

Let's pretend we're in a pretty normal situation: you're on vacation and you want the very best X type of local food. Like in many cities, there are multiple restaurants that all serve X right next to each other (think about getting Pho on Argyle street in Chicago). Your phone has run out of battery, how do you decide which restaurant to get X at?

If you're like me, you would probably walk around and look in all of the restaurants and the one thing that would really tip you to choosing one restaurant is if it had a big line of locals out the door. This process is, in effect, social learning just like in the MusicLab experiment. When we're confronted with a decision that we have minimal information about we essentially rely on the previous decisions of others to inform ours. Said another way, we're integrating their information rationally to inform our decision.

However, just because we are rationally integrating data because we are information poor doesn't mean that the previous decisions we see are rational. You could easily envision a situation where the first person that chose a restuarant when they first opened made a random decision. The next person after that would have several other empty restuarants to choose from or the only one that has a diner so far. Even if they think another restaurant is marginally better they're getting a signal (by this one restaurant already having a diner) that this is a good place to eat. Their decision becomes a coin flip then, follow their private information or follow the lead of the first diner.

At some point, there would be one restaurant still with no diners and another that has multiple tables full and it becomes an easy situation for all the other newcomers. Even if they think that their preferred, empty restaurant is better they may think that the quality has gone down and this other restaurant across the street is better. That leads them to switch and follow the casacade---and once that cascade starts a real winner emerges in popularity!

But what about a purely rational process

In 1981, Rosen wrote a paper on "The economics of superstars" that posited a perfectly rational model. In it he described a world where the was a convex relationship between quality and success. Put another way that a little bit more quality means a lot more success (also refer to the curve shown in the lecture slides).

In his paper he uses an example of music performance, where seeing the best orchestra is worth much more than seeing the second best and so on (i.e. experience versus finances). This ends up working because these types of performances are easily able to accomodate multiple people (there isn't only one performance to be seen, that would drive the cost up enormously) that are easily purchased. Basically, if a Rolling Stones ticket costs \$200, why pay \$25 to see a Rolling Stones cover band goes the logic. And we all know that selling out Soldier Field will pay a lot more than selling out your neighbor's backyard (a Superstar is born!).

However, the problem with this is that ignores the social learning that takes place in the purchasing habits of consumers (this theory would be in direct competition with social learning in many situations). You can only have one theory being the predominant force for a single product's ascent, so it's up to you to try and discern which theory is behind a product's meteoric rise.

In my mind, the situations where the model of a convex relationship between quality and success makes the most sense is in a tournament type situation where pure skill is evaluated. This could be sports (better teams beat lesser teams) or blind auditions (like for orchestras), where the raw skill of the entrant determines success.

So I've heard of this network effect thing?

Ah yes, the network effect. What does that really mean?

It means that one user adopting a product increases the utility of the product to others. The product hasn't functionally changed, but it's utility has. We can see this in many areas, but as a modern example think of Facebook. Facebook, as a service and core product, doesn't change if 1 or 100 people use it. However, a social network is pretty useless without someone to network with---so increasing users increases the utility of the product. This same example holds for older technologies, like VCRs or landline telephones too, but I thought I would try to skew a little younger in the example.

At some point so many people have started to use one product, that its utility is much higher than the alternative. This causes a bandwagon effect. So many more people were on Facebook that it made MySpace useless and users switched in droves. Once enough people chose VCRs or Bluray discs that new movies only came out in those formats, it was silly stick with Betamax or HD-DVD. Network externalities result in a winner-take-all situation similar to social learning, except this is a rational process where users decide based on the lack of utility of the loser.

However, this process is fragile and unpredictable, just like social learning. Until one competitor reaches the tipping point where the bandwagon effect begins, the process could easily switch due to an influx of users from one side. This is because the decision-making process up until that point is largely driven by personal preferences and it is not guaranteed that everyone will decide simultaneously (many are laggards and wait for a winner to begin to emerge before adopting!).

Three models that drive superstars

We now have the three models that produce superstars and the heavy-tailed distribution of success that we started examining in the first lecture. Understanding which theory is at play is important when introducing a product, since the methods you can use to improve your products standing will change.

If quality is the main driver, then it pays to invest in making a far superior product or performance than your competitors.

If social learning is the driver, then you should figure out how to turn out your fans to improve your standing.

If network effects drive the phenomenon, then you should figure out how to increase the installed user base as rapidly as possible.

These are the dead simple take-aways, but they can be effective in helping you hone a strategy for your product if you expect it to compete in a superstar market.