Lights! Camera! Action!
The glamour. The glitter. The pageantry. Joan Rivers on the red carpet.
Of course I'm talking about building a predictive model and putting it into production.
As we approach Sunday's Academy Award extravaganza, it occurred to me that implementing a predictive analytics solution is like creating an academy award-winning film.
So, sit back, relax, grab some popcorn, and on with the show...
1. Developing a story. The story is essential to a great movie. It keeps the audience interested, entertained and on the edge of their seat. For organizations, the story writes itself almost daily - markets are changing, competitors are catching up, and customer expectations are rising. If you're the protagonist (CMO, CFO, line of business manager) in this melodrama, then it's about identifying the main organizational conflict (e.g., reducing churn, increasing up-sell, reducing fraud, etc.) and devising a plan for fast resolution. To borrow a line from the movie Speed, "Pop quiz, hot shot. You're losing customers at a faster rate than you're acquiring them? What do you do? What...do...you do?"
2. Pitch to studio. If the story is promising and has broad appeal, it'll be a good investment opportunity for a studio. Every filmmaker hopes to get their project "greenlighted." For any line of business manager, pitching a predictive analytics project to the CEO/CFO can feel the same way. But, it's been proven that predictive analytics is a worthy and sound investment. Survey results from a recent Predictive Analytics World event revealed that "those who deployed the technology attained a positive ROI, even for their least successful initiative." And, if one project is successful, then pushing an analytics project across other areas becomes that much easier. Organizations also might find that using predictive analytics in one area (such as marketing) is interconnected with another area (such as risk), just like a Tarantino movie.
3. Pre-Production. Preparations are made for filming, such as selecting locations, determining what sets to build, hiring a crew and creating a budget. For organizations, pre-production is centered on two areas: acquiring the data (especially if it's unstructured) and preparing it for analysis. Manual data preparation can account for as much as 40 to 90 percent of an analyst’s time on a given project. Today, this is a thing of the past – like the VHS tape or 2D. Predictive analytics tools now streamline this step and automate the process by detecting and correcting quality errors and imputing missing values in one efficient step.
4. Set Design. Building the right sets is imperative to creating a visually appealing and believable concept for the audience. For organizations, building models no longer requires mad scientist types who work tirelessly day and night to create the perfect, living, breathing model and then scream "It's Alive!" when it's completed. Today, business users can make three simple clicks and build a model. Predictive analytics has turned Dr. Frankenstein into Gary and Wyatt from Weird Science.
5. Casting. The story might be compelling, but if the movie doesn't have the right actor for a particular role, the audience will be immediately turned off (Kevin Costner as Robinhood?). For organizations, predictive analytics now selects the best actor...errr, predictive model...that will produce the best results for a specific problem. And, if a particular model becomes old or outdated, a new model will automatically be selected and used (sort of like when Daniel Craig was refreshed as James Bond, or Christian Bale as Batman).
6. Editing. This entails putting all the pieces together, such as mixing sound, adding music and sound effects, or correcting a mistake (like a boom mike in a shot). For organizations, the editing process takes the form of "what-if analysis." Multiple simulations can be run on the models so they can be “tweaked’ before being deployed. For instance, a retailer who runs many multi-million dollar marketing campaigns can run tests and simulate results before spending a dime and putting the campaign into action. This will ensure that the campaign is targeting the right customers and not wasting time and resources.
7. Distribution. The film is ready for prime time and its premiere. For organizations, this is when the winning predictive model is deployed – sometimes directly into an operational system. Once deployed, it’s just matter of time before seeing the results. Marketing campaigns now yield greater results, churn is reduced, crime rates are decreased, and insurance claims are more streamlined to better identify fraud.
8. Accolades & Awards. With the final count-down to Oscar Night, the odds-makers have been calling their winners among the host of categories. If winners were picked by critic reviews, audience appreciation and box office receipts – the prediction would be easy. But, the real outcome is often buried under other data points including sentiment – a well-liked actor overlooked for too long who puts in a good but not his greatest performance, say Jeff Bridges for Crazy Heart. And, it’s those buried gems of data that predictive analytics uncovers to provide a more accurate expectation of outcome.
That’s why organizations with an award-winning predictive analytics deployment can better identify star customers, or a plot for fraud or a possible crime-scenario with better accuracy and ease.
And when you measure the returns of deploying predictive analytics vs. not, well that’s the palpable difference between making a flop – remember Ishtar or Waterworld – and this year’s winner (insert drum roll) … the envelope please … and the Oscar goes to … The Fighter.
That’s my prediction.