Save the Ugly Baby: the Pixar Success Story

A recent article from Fast Company titled “Your Ideas Are “Ugly Babies,” You Are Their Champion” talks about how Pixar nurtures creativity in the production of their many successful animation movies. For many folks in the creative fields, Pixar is an example of a company that has it all figured out. The innovative digital animation giant has demonstrated an unprecedented knack for creating successful movies, having created 14 No. 1 movies in a row (e.g., Toy Story, Cars, The Incredibles, A Bug’s Life, Monsters Inc., WALL-E, Up, etc.). What can we in the data science space learn from Pixar about nurturing creativity and supporting innovation?

Let your Ideas Suck

One of the important lessons that Pixar president Ed Catmull credits for Pixar’s successful track record is the importance to “Let your Ideas Suck.” Catmull insists that every movie the company has created starts out “ugly” and that early, ill-defined ideas need the most protection lest they die too young. For example, he shared how the first version of the movie “Up” included a king in a castle in the clouds. The creative team threw everything out from that first idea except a bird and the word “up.” From there the film’s concept  went through several other iterations with a little more of the final story emerging each time. Pixar had to make a lot of mistakes and have a lot of failures along the way to achieve the final product, he said.

Let me reiterate a couple of key points in the preceding paragraph:

• It [the idea or concept] went through several iterations with a little more of the final story emerging each time
• They made lots of mistakes and had lots of failures along the way to get to the final product

The story concept begins to take on more relevance when you contemplate that the Pixar movies start out like a newborn baby—ugly. Every one of Pixar’s stories started out ugly. A new idea or thought  is hard to define initially, so it’s likely not very attractive upon conception. Consequently, the idea or concept requires lots of protection and nurturing in its formative days. Every new idea and new concept in any field that involves creativity and innovation needs protection in its “baby” stage. Pixar is set up to protect the director’s ugly baby and their success speaks for itself (see Figure 1).

Figure 1: Box Office Mojo

Let Your Analytic Models Suck

Much like the innovative approach taken by Pixar directors in developing their industry-leading movies, folks in the data science space need to understand that their analytics baby is going to start out “ugly” as well. The first iteration of the analytic model probably won’t look pretty and, in fact, probably won’t yield much in the form of any new or actionable insights. But it’s critically important to all involved parties—data scientists, business stakeholders and senior management alike—that the data science team protect the ugly baby and let it grow during the formative stages of the model development.

For example, in a current vision workshop that we are running in the area of improving product quality and testing effectiveness, one of our data scientist grabbed industry production data from the data.gov website and wanted to explore whether there was any correlation between industry growth and the potential demand spikes for our customer’s products (see Figure 2).

Figure 2: Industry Growth Trends

This analysis doesn’t really tell us anything on first glance. Pretty rudimentary analysis, but it’s a good starting point (yep, the baby is sort of ugly at this point). But then we start to ask additional questions, such as:

• How are we planning on using the information on demand trends by industry to enhance our ability to improve demand spikes that may lead to quality problems?
• Can we integrate this data with our sales data (e.g., SalesForce.com) to better predict/validate the sales demand forecast and flag industry trends that might indicate better or worse performance than indicated in the sales forecast?

In another example, we’re trying to predict line stops, so we are scouring the data looking for any correlations and anomalies that might point to some predictors or lead indicators for line stops (see Figure 3).

Figure 3: Analyzing Line Stop Data

This analysis provides a great starting point, but the baby is ugly because there is nothing here that is yet actionable. The analysis begs the next level of questions or inquiry, such as:

• If root cause and containment are the two major outcomes (lag indicators), what are the variables buried in the line stop data that might predict these two outcomes (lead indicators)?
• Can you uncover any rules that are indicative of a root cause or containment outcome?
• What are the differences in the variables that might lead to a root cause outcome versus a containment outcome?

Manage Expectations: The Baby Will Become Beautiful

Organizations that may not be familiar with the data science “exploration, develop, test, explore, refine, test, explore…” process should take some level of comfort in the fact that this is not new and that there exists a well-defined process, called the Cross Industry Standard Process (CRISP) for Data Mining. This process defines how data scientists are going to convert the ugly baby into something beautiful (see Figure 4).

Figure 4: Cross Industry Standard Process (CRISP) for Data Mining

It takes time, but it works as a natural process based on curiosity (get some results, as some more questions, get some data, and get some more results).

Stay Curious My Friend

Let me reiterate a couple of key points from the Pixar approach:

• It [the idea or concept] went through several iterations with a little more of the final story emerging each time
• Lots of mistakes were made and failures were experienced along the way to get the final product

It’s critically important that the analytic model is allowed to start off ugly and that it goes through its natural failure and growth pattern. And there will be many, many failures along the discovery path. And that’s good, because acknowledging and dealing with initial failures helps you move that much closer to a successful end result.

To quote two famous innovators:

“Failure is success in progress”

― Albert Einstein

“I have not failed. I’ve just found 10,000 ways that won’t work.”
― Thomas Edison

And to quote a third innovator, “Be patient and stay curious, my friend.”