alphaspirit - Fotolia
As more and more enterprises master the basics of business intelligence reporting and descriptive analytics, the real value from analytics is moving into more advanced territory, like predictive and prescriptive analytics. The problem, particularly for businesses that sell analytics-based products, is how to explain this value to customers.
"In some instances, people get what we do in a flash," Boris Savkovic, lead data scientist at BuildingIQ, wrote in an email interview. "In some cases, we have a lot of educating to do."
BuildingIQ, based in San Mateo, Calif., is a software-as-a-service company that helps building managers monitor and adjust facilities' heating and air conditioning to improve efficiency and reduce costs. The product is built around advanced machine learning algorithms that factor in historical energy use data, weather forecasts, data streaming off buildings' HVAC systems and energy cost data. The machine learning algorithms, developed using MathWorks' MATLAB software, continually determine optimal HVAC settings for a building throughout the day, and these settings are passed on to the building's HVAC controls via Java code.
The company claims it can save building operators between 10% and 15% of their energy consumption costs, but this doesn't make the sales pitch a slam dunk. Savkovic said a lot of work goes into making sure building managers understand the value proposition.
Part of the difficulty is that customers generally don't interact with the machine learning algorithms. They are developed entirely by BuildingIQ's data scientists and run on the company's own servers. The whole thing works pretty much on autopilot from the perspective of the customer. And while this can be an advantage for customers, in terms of not having to invest in hardware or spend their own staff time developing algorithms, it makes the whole thing abstract.
This is why, Savkovic said, it's important to do some work ahead of time. He and his team will take data from prospective customers and develop models for their buildings, showing how much they could have saved in the past had they used the service. Visualizing this data is also key. This helps translate a complicated predictive analysis into something that staff from the business side of the facility management company can understand, even if they don't have sophisticated data analysis skills themselves.
"Presenting visuals and outputs go a long way to explaining what BuildingIQ does," Savkovic said.
Ultimately, it may be changing attitudes that spur adoption of technologies like advanced machine learning algorithms and other data-driven technologies. Savkovic said the buildings industry hasn't traditionally been the most forward-looking when it comes to new technologies. But he sees this starting to change in ways that could make data-driven products more common.
"The buildings industry has been very slow to adapt to the digital disruption," he said. "But given the investments being made into smart cities and smart buildings, the adoption and maturity of the market will grow."
Parallel data processing pushes machine learning algorithms into the future
Machine learning libraries are a big part of the draw of Apache Spark
Advanced machine learning algorithms will play big role in future of many industries