Sapping Accident Insight from Big Data

The second season of the television show “Breaking Bad” led to a dramatic mid-air plane crash above Albuquerque. That crash happened because of a single mistake by an air traffic controller that guided two planes into one another.

That fictional airline would likely have benefited from incident management analytics. Instead of investigating the accident and concluding, “perhaps we should not have let a distraught man control the lives of hundreds of people,” they potentially could have identified those warning signs in advance or instituted failsafes that call on another human or computer for backup in case of impending emergency.

To help curb those and other far less dramatic accidents, SAP developed their Incident Management rapid-deployment system. SAP VPs of Sustainable Products James Farrar and Jeremiah Stone spoke to Datanami this week about their new platform.

SAP calls their platform “rapid-deployment” owing to the fact that, according to them, companies who implement it and start computing within 90 days. The lynchpin to that operation is collecting data on not only actual incidents but also near-misses. SAP works with oil refineries and airlines among others, companies whose mistakes, such as crashes and spills, could spark disastrous environmental and human life consequences.

However, for every incident, there exists several hundred events that could have went awry had one or two seemingly consequential factors gone differently. The data from these near-misses are just as valuable to SAP, as they help build a stronger database and would allow them to more effectively identify problem variables.

According to Stone, the most difficult part of developing the platform was making the user interface similar to that of their consumer-based systems. “Our customers are asking for us to provide a tool that everybody can gain actionable information from in order to run safer,” Stone said. “That design challenge was tremendous because it needs to be as user-friendly and as approachable as our consumer-based applications, so we want to provide a high level of performance with professional power but with a consumer-based interface.”

Beneath the top layer and the user interface is the actual ability to identify problem spots in a company’s system. What makes that difficult is that some incident reporting relies on observations from humans that are often input as unstructured text data. In order to garner information from that, SAP structured a significant amount of that. “Part of the key is structuring as much information as possible, from both incidents where there’s harm to a person or the environment to near-misses and observations,” Stone said.

In order to better structure observational data, SAP created classifications such that someone can quantify their qualitative experience. The success of this organizational effort will be key to the platform’s effectiveness.

The next step is implementing a system that identifies incident causation. According to Stone, SAP serves as a foundation for the user to ultimately make the final relationship determination. “Identifying the causes that led to an incident is a combination of guiding users through an investigative approach and providing users a data interrogation and exploration environment so that they can explore their own data to identify potential causal relationships.”

The big data perspective comes in when working through those massive datasets compiled through extensive incident or near-incident reporting.  “In terms of the analytical processing side, this gets into the big data opportunity,” Stone said. “We provide a great deal of flexibility for users to explore and mine the data.”

So far, the Incident Management platform has already shifted the way Tesoro, an oil refinery company, understands their risk. According to Farrar, while the total number of reported incidents has increased, the severity of those incidents has decreased. “At first glance they’d say ‘oh we’re not as safe as we used to be.’ In fact they’re finding they didn’t have all the information.”

In a sense, when applying analytics to incident management, the focus shifts from a simple statistical count of how many problems happened and onto the underlying factors that led to those incidents. With the Tesoro example, simple reporting measures have led to a basic statistical jump in minor incidents reported but have also led to a higher awareness regarding major incidents and, as a result, have curbed them.

The overall goal of the Incident Management platform, according to Farrar, is to turn corporate investigation into operations management. “Once you start looking at the big data around safety, you start getting into the predictive world. You’re moving away from the discipline of managing safety in a very corporate way to instead managing risk.”

When companies increase risk management efficiency, operation costs decrease. As such, Farrar argues that an effective incident management and prevention system essentially turns a hard-nosed corporate investigator into an operations officer. “Managing risk is a different game in a way because if you stop being a corporate policeman and stop investigating the incidents that have happened, and start being a risk manager and seeing how incidents could happen and how to prevent that chain—then you become an operations manager.”

Maintaining safe workplace environments has always been a priority. A big data-fueled, analytical approach to accident awareness could be a significant step in regulating safe practices.

Related Articles

SAP Signals Spawn of In-Memory Era

IEEE Tags Top Tech Trends of 2013

Quartz Crystallizes Long-Term Data Storage