Find out how most companies get operational analytics wrong, and how data science can play an important role in steering these efforts in the right direction.
I am constantly amazed at how smart cars are these days. When I was growing up you had your basics: Speedometer, tachometer, temperature, and fuel tank indicator. Now, even the most basic car comes with a small cockpit--with everything from fuel consumption efficiency to range (i.e., how far you can go before you run out of fuel).
What's more impressive is the intelligence built into helping us drive better. Rear view cameras, collision warning systems, and adaptive cruise control make it very easy for us to make the right decisions while we are behind the wheel. Extending this concept into the operations of your own organization can have a powerful impact.
SEE: Big data policy (Tech Pro Research)
The most competitive organizations are using data scientists to bring analytics to their day-to-day operations. Many people believe data science belongs behind the scenes; they envision large amounts of operational data funneling into a massive data lake where it waits for superintelligent analysts using high-powered algorithms to work their magic.
Although this is a very plausible scenario, this is not the only place where data science provides great value. There are many situations where operational data science makes sense. One such area is operational analytics--bringing more intelligence to your corporate machine.
Two key ways data science helps with operations
A typical company has an organization that develops and an organization that operates. When I was consulting with PayPal, we had a group of talented professionals that constantly improved the functionality of the PayPal website.
There was an equally talented group of professionals responsible for handling the operations of the production site. This operations group had a very different environment within which to succeed. That is why they had the best tools available to analyze what was happening at any point in time, and the best practices for troubleshooting problems in the moment.
SEE: 60 ways to get the most value from your big data initiatives (free PDF) (TechRepublic)
Data science can help tremendously with monitoring and troubleshooting. A key difference between operations and development is in their perspective of the status quo. For operations, stability is the goal--preserve the status quo; therefore, data science must be used to alert operators when the situation is not normal. For instance, classification algorithms are very useful for identifying when conditions are moving away from normal.
Once abnormal conditions are detected and verified, the operator must take immediate action to bring things back to normal. This is where an expert system can be a valuable friend. An expert system can learn all the ways your machine can go sideways and, more importantly, how to get your operation back on track.
Predictive analytics and leading indicators
The key with operational analytics is to give operators a concise set of leading indicators with which to work. Whenever you develop a dashboard for operators, it must be as simple and concise as possible. This is where most companies get it wrong--they feel data science and analytics implies sophisticated analyses and visualizations. Although this may be appropriate for development or strategic functions, it is absolutely the wrong approach for operations.
Consider a scoreboard for any professional sports event: Players want to know if they are winning or losing and how much time is left. The same goes for your company's operations--keep the dashboard simple with a select set of leading indicators.
Leading indicators predict lagging indicators. The amount of gas in your car predicts how far it will travel before it stops. It does you no good to be alerted of the fact that your car cannot travel anymore because it is out of gas--it is better to watch the gas level and refuel when it gets too low. In the same way, your company has operational goals, but that is not what your operators should focus on.
Instead, operators should focus on the metrics that predict your operational goals; this is another area where data science can help. Oftentimes, companies choose leading indicators on the assumption they will predict lagging indicators (i.e., operational goals). Data scientists have the ability to confirm predictability and prediction confidence.
Data science can play an important role in the operations of your company. In some cases, you may want to put data scientists on the front line to do real-time analyses for on-the-spot troubleshooting; however, they don't need to be on the front line to add value. Your data scientists can bring more intelligence to the operators that are running your machine.
Classification systems can be applied to your monitoring system to make sure everything stays in control, and expert systems can be useful when the operation needs immediate attention. Remember to keep your operational analytics simple and use leading indicators so your operators can address small issues before they become big issues.
If you make your own operation as smart as your car, you will be on the road to success.
- Cheat sheet: How to become a data scientist (TechRepublic)
- 10 signs you may not be cut out for a data scientist job (TechRepublic)
- Predictive analytics: A cheat sheet (TechRepublic)
- Why data scientists need to understand the business (TechRepublic)
- Why 80% of UK businesses plan to hire a data scientist in 2019 (TechRepublic)
- The future of customer service needs a new model of operational excellence (ZDNet)