But the reality is far different: hits are not only few and far between, but they are devilishly difficult to predict.
Movie studios, music labels, talent agencies and sports franchises have for years attempted to identify, isolate and manage the variables they believe may enable them to pick hits without error. As the record shows, however, that magic formula continues to elude just about everyone.
It is probable that there is just too much statistical 'noise' in the systems that generate hits for anyone to be able to accurately select them without error. There are, however, processes one may use to improve the hit-to-heckle ratio.
Pixar is often mentioned in discussions about how to churn out consistent hits. As the following article explains, however, the company's process focuses more on eliminating flawed or less exceptional fare than in identifying the best. A big part of the process is insisting upon commentary from outsiders whose voice and opinion is not as subject to the internal relationships and politics within the organization. It also recognizes that figuring out what might not work or what is wrong is a lot easier than 'knowing' what is right.
Once the choice has been made, chances for success can be improved by using a variety of channels and platforms to reinforce the notion that whatever the vehicle - film, song, star - should not be missed. This strategy acknowledges that opinion leaders and the general public get their references from a variety of sources, so assuring that the message replays and echoes all over the media and personal firmament is crucial. The larger point is that while there is no specific system for predicting hits, there are many ways of reducing the likelihood of failure. JL
Kartik Hosanagar reports in Wired:
One of the harder things for people to understand is why top actors and musicians get paid so much more money than top teachers and nurses. It isn’t because we don’t value these professions. It’s because they don’t have scalable distribution, or the ability to reach many more potential users in a short period of time.
This same factor is at play in “hits-driven” industries ranging from movies, music, and book publishing to venture capital and mobile apps. Our analysis of a top-tier VC fund — one of the top performers of its vintage year — showed that over three-fourths of its tech startup investments did not break even, while the top investments generated outsized returns. Looking at over 3,300 Facebook games, we found that the top 20 percent of games accounted for roughly 95 percent of the monthly active users (MAUs) in our dataset. And we already know that most movies made by studios and indie producers lose money or barely break even; the industry follows the 80-20 rule, where the top 20 percent of movies generate about 80 percent of box-office revenues.
So we know that scalable distribution results in a winner-takes-all phenomenon where a few players take an overwhelming share of the rewards in their market, leaving very little for the median players. The question is, can we predict or create these hits with any consistency?
No One Can Really Predict a Hit…
Given the economic value of spotting would-be hits, you’d think that hit prediction would be a science by now. It isn’t. There are countless stories of experienced movie studios missing big hits, established publishing houses missing bestsellers, leading VCs missing the next big thing.
Why? Because, scalability. It not only contributes to the winner-takes-all dynamic, but also increases unpredictability in markets in two ways: (1) small differences in quality translate into big differences in market share; (2) social media and other powerful forms of word of mouth create social influence. The former is hard to spot, the latter is hard to predict.
Quality has always mattered. But it matters differently in scalable markets.
In the past, being better than everyone else created a local bestseller like “the best restaurant on Main Street.” It was hard to expand to new markets without incurring significant costs until mass production, cheaper transportation, and more recently the internet made distribution more scalable. But this scalability amplifies small differences in quality (or rather, the impact of them), because being slightly better than others is enough to access larger markets.
The problem when it comes to predicting hits is that while big differences in success are easy to observe, small differences in quality are harder to spot. Even for experts who have accumulated a ton of market information.
Yet quality alone doesn’t explain why movies like Grown Ups 2 or bands like One Direction do really well. This brings us to the second reason experts are unable to predict which products will be hits: social influence. The role of social influence is perhaps best illustrated through an experiment run by Princeton University professor Matt Salganik and his coauthors.
Intrigued by the tremendous success of the first Harry Potter book — which was rejected by eight publishers before finally being picked up — Salganik set out to design an experiment that separates out the role of luck from quality. “[Imagine] if you could rewind to the day that Harry Potter was released and make eight copies of the world, and have them all evolve independently,” he described his experimental design to me. “Then you could tell more about the role of luck by observing whether success in one world replicates in another.”
To test this idea, Salganik and his team created an artificial music market with over 14,000 participants listening to 48 previously unknown songs from 18 unknown bands. Participants could listen to the randomly ordered songs, rate, and download them — the number of downloads served as a measure for quality. The difference between the social (vs. independent) condition of the experiment was that participants could also see how many times each song had been downloaded by previous participants. Then Salganik and colleagues added another brilliant twist to this social condition: The participants were randomly assigned to one of eight parallel universes, and could only observe the download counts within their own “universe”.
The impact of social cues was immediate. The top downloaded songs in the social condition had a greater share of the market than those in the independent condition. Not only did later participants increasingly gravitate towards the songs chosen by prior participants, but the team found that even small variations in the choices of the initial few participants could introduce large variations in the final outcomes.
Early momentum created by seemingly random factors can have dramatic impact on final outcomes. One song, Lock-Down by 52 Metro [song ranked #27 in the independent condition with x, y coordinates at (1, 40)], came in 1st in one universe and 40th out of 48 songs in another universe. Graph based on data shared by Matt Salganik.
It is striking to see the variation in a song’s rank across the different social universes: A song in the top quartile in the independent condition can end up in the bottom quartile in the social condition simply because of how the social cues played out. Our research confirms that social cues in product reviews and automated recommendations play out in a similar fashion. In short, the random choices of early users combined with social cues makes the market highly unpredictable.
…But You Can Help Make Hits Happen
Despite the above uncertainties, some companies have been surprisingly consistent in producing hits. Pixar, for example, has made 14 major films to date and all 14 have been blockbusters, each generating well over $100 million in box office returns.
The probability of Pixar having 14 out of 14 hits is actually 1 in 100 trillion, given the historic 1-in-10 odds of studios producing a blockbuster success. Clearly, this kind of success is not due to pure chance. It has to do with how Pixar — and other successful organizations — manage uncertainty, particularly around the factors of quality and social influence.
Addressing Quality: You Can’t Predict Hits, But You Can Fail Better
While many organizations embark on high-risk, high-return projects by relying on a small group to generate and select ideas and then fully commit to them, Pixar brings in enough stakeholders into the idea-generation process to maximize the number of ideas.
Pixar starts with close to 500 one-line movie pitches, such as “A hot-shot race car named Lighting McQueen gets waylaid in Radiator Springs, where he finds the meaning of friendship and family” (Cars). These are then shortlisted, further developed, and iteratively narrowed down until they end up with one movie worth pursuing. As Toy Story 3 director Lee Unkrich shares, “We fail a lot. We just don’t fail by the time the movie comes out.”
It’s not enough for an iterative design process to go through several rounds of testing. The way the best idea or design is ultimately selected is equally important.Firms such as Pixar produce one blockbuster after another not because they consistently come up with the best ideas — but because they consistently weed out the poor ones by failing often and early. But it’s not enough for an iterative design process to go through several rounds of testing. The way the best idea or design is ultimately selected is equally important.
Hybrid group structures — where individuals independently come up with ideas and then come together as a team to evaluate them — result in better ideas and a higher ability to discern the best idea, according to recent findings by my Wharton colleagues. Unlike in traditional brainstorming processes, where people are likely to respond based on the trajectory and history of discussion rather than the ideas themselves, hybrid structures ensure more independent evaluation … and therefore higher quality ideas.
Addressing Social Influence: You Can’t Make Something Go Viral But You Can Help Move It Along
Perplexed by how eyeglasses cost more than iPhones, a couple of my MBA students a few years ago, Neil Blumenthal and Jeffrey Raider, started working on a business plan for an online eyewear retail business. It was Warby Parker. When they first shared the idea with me, I remember thinking it would be hard to convince people to buy eyeglasses online without first trying them and getting feedback on fit from the optician or friends.
Several months went by before I heard about them again. Suddenly, a lot of people in my Facebook feed started uploading photos of themselves seeking input on different Warby Parker frames. Sales had taken off, so I visited their office to better understand what had transpired in the year since they had graduated.
You’d think that hit prediction would be a science by now.Warby Parker had not only launched a try-on-at-home program, but they actively encouraged users to upload photos on Facebook and get feedback from their friends — sometimes even posting those photos on behalf of consumers. This approach helped generate social cues from users; the key, as per the findings of the Salganik Musiclab experiment, was that these cues were early.
Now, social media strategy is nothing new; how many firms spend millions on social media campaigns and agencies? But unlike the contrived “share on Facebook and get $20 off” approach, Warby Parker took people’s natural social need — to get second opinions on “how does this look on me”? — and converted that into a marketing tool for the company. It worked: the company is estimated to be worth over $300 million today.
We can’t contrive virality, but we can organically encourage early social cues — and it’s especially effective when it’s part of the actual product experience.
***
More and more industries are becoming hit-driven, especially as technology brings scalability to traditional markets (the popular example du jour is MOOCs and education). This means predicting and making hits will require us to rethink how we judge quality and harness social influence.
But just as scale giveth, it also takes away: The median player will no longer survive in such a hit-driven market. Still, at least scalability means the best teachers will finally be accessible to millions of students. It won’t just be actors and musicians who have scalable distribution anymore.
0 comments:
Post a Comment