In late 2016, Kalyan Veeramachaneni, an MIT researcher on Big Data and AI, highlighted some of the flaws of using data science to drive business growth. The conclusion of his study was that data analysts tend to view their jobs as an intellectual challenge. They are therefore disconnected from the economic questions they should strive to answer. Let’s find out the other reasons behind less than stellar results for companies that spend big money on big data without getting proportional returns on their investments.
The first and most important aspect any organization needs to grasp is that data science is not a miracle cure for your company’s underlying problems, just a gauge.
Data science has a reputation for having saved some organizations thousands of dollars and for rocketing start-ups to worldwide recognition. But that only happens when it helps answer some well defined questions.
Hiring the best data scientists without giving them direction is akin to acquiring a fancy sports car without having a driver’s license.
RELATED ARTICLE: SMALL BUSINESS BUDGETS: HOW TO SAVE BIG BUCKS ON THE BEST TECHNOLOGY
Answers are usually no better than the questions asked. Therefore, data science should be part of a coherent business strategy.
Top management should define the company’s issues they hope to find solutions for by using big data. Furthermore, the variables should be chosen not by the scientists, but by those with knowledge of the processes.
A data science project needs to have clearly defined goals, metrics, and milestones. This should be no different from opening a new branch or buying new equipment.
You will need to assess opportunity costs, decide why are you making such an investment and how soon you expect to see results. Have smart goals for your big data project and instruct the data scientists about those goals from day one.
Especially for start-ups, it can be smart to hire a data science company to conduct the project instead of creating a new department. In this case, it is even more imperative to know what to ask for in terms of your expected results and business goals.
Since most data science projects are in the experimental stage, they lack explicit control. The positive aspect of this is that data scientists have opportunities to try and develop a wide array of models. However, the downside is that without a minimal viable product (MVP) requirement, their work can become too academic and lack real-world applications. A possible solution includes using an agile approach with weekly and monthly progress reports related to the initial goal.
Data scientists and analysts usually come from an academic background. They’re focused on mathematics, statistics and science. Managers, on the other hand, are involved in the economic aspects of the business. They want to know how to deal with clients, save costs and avoid legal problems.
Misunderstandings and other communication difficulties can result in a lack of trust between the two groups.
A possible solution to this problem is to create cross-functional teams. The first rule is that the teams should establish a common language with clearly defined terms.
Additionally, invest in helping the data scientists develop some business acumen.
Another solution involves appointing, as head of the data analysis team, a leader who makes the two worlds come together. This person needs to have relevant experience both as a business consultant and as a data scientist.
It is not the most beautiful or interesting model that matters, but the one that solves the problems with the fewest resources.
Data scientists are highly intelligent individuals who are not all that interested in climbing the hierarchical ladder. However, models they can refine into thought diamonds fascinate them.
Tap into their natural inclination to solve problems by defining the struggle for them and letting them solve it by whatever means they come up with. Resist the temptation to micro-manage them. Design opportunities to make data scientists feel valued. Reward their innovative ways with industry-recognized competitions and award galas.
Make them care about the projects they are working on by taking them out of the lab and getting them on site if possible. By looking at the phenomenon unfolding in front of them, they will become more receptive. They will likely even identify interactions or ask for new sets of data that are more appropriate for the job.
Analysts should focus on speed and pick big data’s low-hanging fruit. You want them to develop simpler models, test those models against diverse data sets and pick only the most relevant sets.
However, this approach requires combing through enormous quantities of data and readying that data to be used in algorithms. This work, although necessary, is not all that enjoyable or intellectually stimulating. Therefore, it is a perfect candidate for automation.
Keep in mind that the models don’t have to be perfect. They just need to be good enough to predict what they are built to predict. They also need to have a simple interface that non-technical staff, such as sales agents or shipment personnel, can easily use.
In the absence of clear management directives, human nature will take over, and data scientists will focus on what they’re good at. They will create and calibrate models. This value-added activity is important, but it should represent only about 10% of the entire work.
The teams should direct half of their efforts at interpreting the results and creating actionable insights. The rest of the project involves the hard work of selecting, preparing and cleaning the data.
Successful companies are helping their data scientist see beyond the numbers. And they’re helping members from other departments make sense of data science’s results.
The post 5 Easy Ways to Get the Most Value from Data Science appeared first on Business Opportunities.