Becoming “Netflix Intelligent”: Something Every Company Can Do!

Nobody does it better, 

Makes me feel sad for the rest.

Nobody does it half as good as you,

Baby, you’re the best.

–Sung by Carly Simon in the James Bond movie, “The Spy Who Loved Me”

Actually, that’s a movie (“The Spy that Loved Me”) that Netflix recommends for me since I’m a James Bond junkie and Netflix knows that. In fact, Netflix knows a lot about me as it knows a lot about all of its viewers, which is one reason why Netflix is a Wall Street darling and has rewarded its stockholders very well over the past couple of years (see Figure 1).

Figure 1: Netflix Stock Price

But Netflix isn’t doing anything that other organizations cannot do. To replicate Netflix’s business success starts with thinking differently about the role of data and analytics in powering the organization’s business. And while I do not have any “insider secrets” to what Netflix does, it doesn’t take James Bond to piece together the secrets to their success.

The Secret Ingredients to Netflix’s Success

The secret to Netflix’s success comes from the power of combining detailed viewer behavioral data and detailed show/program characteristics data with machine learning to make predictions about what shows what viewers might want to watch. This is a recipe that every company can and should follow! And it’s really not a hard recipe, if you have the right ingredients:

Ingredient #1: Targeted Use Case

You can’t get to the “What” and the “Why” by focusing on the “How”

They key to any organization’s big data and data science success starts by identifying the “What” and the “Why; that is, “what” are the business use cases upon which your big data and data science initiative should focus and “why” are these use cases important to the business.

Netflix’s market domination and outstanding business performance rests on their ability to improve customer retention and drive customer loyalty through increased viewer engagement. And the heart of Netflix’s increasing viewer engagement is their movie recommendation engine (see Figure 2).

Figure 2: Netflix Recommendation Engine

The movie recommendation engine has been incredibly successful in increasing viewer engagement. The number of hours the average viewer watches Netflix has increased an average of 16.4% per year since 2011. The average Netflix viewer watched almost 12 days more Netflix in 2016 than they did in 2011 (see Figure 3)[1]!

Figure 3: Average Netflix Viewing Hours

Ingredient #2: Create Analytic (Viewer) Profiles

Netflix has over 250 million active viewers worldwide. From their individual profiles, Netflix is able to determine what people watch, what they watch after, what they watch before, what they watched a year ago, what they’ve watched recently and what time of day[2]. This viewer viewing behavior data forms the foundation for their business model.

But it isn’t just the data. It’s critical to be constantly mining the viewers’ viewing behaviors to uncover insights about that individual viewer – their viewing behaviors, preferences, tendencies, inclinations, trends, biases and patterns (see Figure 4).

Figure 4: Big Data is Learning about Individual Behaviors

Being able to integrate viewers’ viewing behaviors with external data sources makes the customer data even more valuable…and actionable from a recommendations perspective. For example, we could learn more about viewers’ interests, passions, affiliations and associations from their social media engagements on Facebook, Twitter, Snapchat and LinkedIn. We can measure their strength of advocacy for certain programs and types (genre) of programs based upon what they are saying publicly on social media sites.

We can capture all these analytic and behavioral insights in a customer Analytic Profile. Analytic Profiles capture the organization’s analytic assets in a way that facilities the refinement and sharing of those analytic assets across multiple use cases. An Analytic Profile consists of metrics, predictive indicators, segments, scores, and business rules that codify the behaviors, preferences, propensities, inclinations, tendencies, interests, associations and affiliations for the organization’s key business entities such as customers, patients, students, athletes, jet engines, cars, locomotives, CAT scanners, and wind turbines (see Figure 5).

Figure 5: Customer Analytic Profile

Analytic Profiles enforce a discipline in the capture and re-use of analytics insights at the level of the individual key business entity. See the following blogs for more details on the concept and application of Analytic Profiles:

• How to Avoid “Orphaned Analytics”
• Analytic Profiles: Key to Data Monetization

Ingredient #3: Capture Show Characteristics and Viewing Patterns

In much the same way that organization can build detailed analytic profiles on all of their customers (viewers), organizations can also build detailed analytic profiles on all their products …at the level of the individual product. Individual products (e.g., cars, wind turbines, medical devices, jet engines, CAT machines) develop unique “fingerprints” that can be used to distinguish between the different operational behaviors of the different products.

I have no insider’s knowledge into how Netflix classifies their programs and shows, but I did find an article titled “So How Does Netflix Categorize Titles in Its Film & TV Libraries?” that talks about how Netflix classifies shows:

“[Viewing Analysts use] a spreadsheet to take notes of all the info that goes into the 100+ data points used by Netflix to tag each title.”

Now to a certain extent, Netflix has it easy because their “products” are all unique and there is a bounty of external data about each of their “products” from sites such as Rotten Tomatoes, IMDb, Common Sense Media and Flixster. But having the data is very different from knowing what you want to do with that data.

Ingredient #4: Mastering Machine Learning

As we discussed above, Netflix has access to a broad set of viewer viewing data—what each member watches, when they watch, the place on the Netflix screen the customer found the video, recommendations the customer didn’t choose, the popularity of videos in the catalog, etc. All of this data gets fed into supervised (classification, regression) and unsupervised (dimensionality reduction through clustering or compression) Machine Learning algorithms to create critical scores such as:

• A video-to-video Similarity algorithm makes recommendations in the “Because You Watched” row.
• A Video Ranker algorithm then selects the order of videos in genre rows for each individual viewer[3].

Figure 6: The training and evaluation of algorithms takes place offline, while A/B testing takes place online. Images adopted from ACM.

Figure 6: The training and evaluation of algorithms takes place offline, while A/B testing takes place online. Images adopted from ACM.

Understanding how Machine Learning works isn’t hard (See “What tomorrow’s business leaders need to know about Machine Learning?” for more details.). And while there certainly is an art and a perseverance necessary to explore the different types of structured and unstructured machine learning algorithms to find the ones that work best, the key to the success is to have a data science exploration and learning process that allows the data science teams to explore the bounty of viewer and show data in search of those variables and metrics that might be better predictors of viewer’s behaviors (see Figure 7).

Figure 7: Data Science and Exploration and Learning Process

See the blog “Dynamic Duo of Analytic Power” for more details on the Data Science exploration and learning process.

Ingredient #5: Management Fortitude to Become “Netflix Intelligent”

Ingredient #5 is probably the one ingredient that trips up most organizations because none of the other ingredients mean a thing if you don’t have a management team that knows how to properly value and use the organization’s data and analytics. Netflix is a digital-first company, but they weren’t always that way. Netflix started as a company that shipped movies to their customers using physical DVD’s. But as technology capabilities (increased bandwidth and availability of internet access) and market demands changed (people wanted to watch their movies now, not two days later), Netflix made the hard digital transformation.

Becoming “Netflix Intelligent” is something that any organization can do today. There is nothing magical about the data or the technologies. It just takes an approach and a willingness to act. I’ve just given you the approach. Now the key question is “Do you have the willingness?”

Sources:

[1] “Average Netflix Subscriber Watches Almost Twice as Much Netflix as They Did 5 Years Ago” 

[2] “This is How Netflix’s Top-Secret Recommendation System Works” 

[3] “How Machine Learning Fuels Your Netflix Addiction”

--------------------

Thanks for taking the time to read my post. I’m fortunate that I spend most of my time with very interesting clients which fuel many of my topics. I hope that you are able to leave a comment or some thoughts about the blog. If you would like to read my regular blogs, please follow me on LinkedIn and/or Twitter.

In case you are interested, here are some of my favorite posts:

·     Determining the Economic Value of Data

·     The Big Data Intellectual Capital Rubik’s Cube

·     How to Avoid “Orphaned Analytics” 

·     To Achieve Big Data’s Potential, Get It Into The Boardroom

·     Vision Workshop

·     Big Data Business Model Maturity Index (animation)

·     How I’ve Learned To Stop Worrying And Love The Data Lake

I am the author of two Big Data books: “Big Data: Understanding How Data Powers Big Business” and “Big Data MBA: Driving Business Strategies with Data Science”.   I also teach the "Big Data MBA" at the University of San Francisco (USF) School of Management, where I was named the School of Management’s first Executive Fellow. The opportunity to teach at USF gives me the perfect petri dish to test new ideas and concepts both in the classroom and in the field with clients.