Do you want to map out your implementation of Big Data analytics? What benefits will this bring to your organization? And what potential challenges should you be aware of? In this series, we’ll show you how to plan your own Big Data journey.

If you’re looking to get your company started with Big Data and data analytics in 2016, you’re making a smart move. Data analytics has already helped many companies become more efficient, more agile, and more profitable.

However, implementing a Big Data strategy requires careful planning. One of the things we are commonly asked is ‘How do we start using Big Data?’ In this post, we’ll take a look at two typical ways of using Big Data to solve a business problem. And we’ll examine a third solution that many companies have found even more useful.

The Top-Down Data Analytics Model: Focusing on Business Issues

The top-down model comes naturally to people on the business side of the equation. In this method, an area is targeted for improvement – for example, upping customer engagement through digital channels. When the top-down approach is employed, the focus is on using data to fix the problem. How the data fixes the problem is less important than the result.

One of the pros of this approach is that it is readily adopted by business stakeholders. It’s very popular because it’s a quick and relatively easy, especially when there are already some Big Data deployments in place. But this same emphasis on quick and understandable results also leads to a pair of significant downsides: a lack of unforeseen, out-of-the-box results and a possible uptick in the number of data and technology silos.

The Bottom-Up Analytics Model: Building for the Future

The bottom-up approach actually starts with the processing of all available data. Insights emerge as information is analyzed over longer periods of time. This leads to a more integrated overall response; instead of merely targeting digital customer engagement as a focus, a bottom-down analytics method might call for integrating all available systems and data to get a complete picture of what’s really going on.

As you can imagine, this approach takes time. Quantifiable results do not immediately leap off the screen, ready for business use; instead, results come more slowly and must often be analyzed to align with stated business goals. Given its reliance on technology over business practice, it can be harder to get stakeholder support on this method.

The plus side to the bottom-up is that it’s a deeper, more sustainable system. If your organization is just starting out in its Big Data journey, a bottom-up framework will give you room to grow and learn tech skills. And because you’re not locked into a single business object, a range of explorative insights can emerge that top-down systems miss.

Clearly, there are flaws to both of these approaches. That’s why Absolutdata often employs a third approach that combines the best of both into a hybrid analytics model.

The Hybrid Analytics Model: Switching Things Up

With a hybrid analytics model, you can get the best of both worlds. If you’re starting from scratch, we recommend employing a bottom-up model at first. This will get you familiar with your data and data patterns. Once you understand the type of data pouring into your framework, you can select business goals and give them laser-like focus with the top-down approach.

And then you’ll iterate your model, your goals, and your processes. And you’ll iterate again, as insights continue to make your journey clearer.

Before we move on, let’s examine a real-world example of how this approach works.

A Case History of the Hybrid Model

A large mining company approached Absolutdata with the question ‘What can we do with all the truck ops and sensor data we have?’ Better efficiency was the overall goal here, but we didn’t swoop in on that right away. We let the data tell its own story first.

To start, we went straight to the data. (This is the bottom-down approach.) We harmonized the data coming in from multiple systems, and we set up dashboards and an integrated, Hadoop-based back end. Using this, we were able to establish a single source of truth – in other words, a repository of information that the data was telling us to explore.

This information helped our team and our client to define four areas for potential improvement: the predictive maintenance of trucks; machine failure predictions; the prioritization of the alert system; and optimization of the haul cycle.

Now that these four areas were clear, we used a top-down approach to solve them. This involved deploying machine learning models, integrating oil-test data, and integrating our models with their existing truck Op Management system. With these objectives in place – and with technology working toward solving them – we helped our client shave $20 million off their annual truck operations budget.

This history shows how beautifully these two systems work together. Without the bottom-up initial approach, would we have identified all four areas? It’s unlikely. But without the precision of the top-down approach, would these areas have been improved so quickly? Again, unlikely.

The key here is not to get stuck on one process over the other – let the emerging data guide your path. Next week, we’ll discuss another step in the Big Data journey: technology and Big Data architecture. Join us then, or subscribe to our blog so you don’t miss it!

Authored by Sukhmeet Sethi, Senior Technical Architect at Absolutdata and Titir Pal, Director – Analytics at Absolutdata