Why All the Buzz About Analytics and Big Data in Upstream Oil and Gas?05/10/2016
Analytics can help us all make better decisions, but only if we understand what we’re getting, what it means, and the assumptions and data going into the system
Big data and analytics are probably the most popular buzz words in technology right now. The reality is, big data and analytics have been used in the petroleum industry a long time. P2 Tobin has been in the data-collecting and data-processing business since the 1920s, so we know a thing or two about big data and analytics.
If you had asked a Tobin employee in the 1930s, “What is big data and analytics?” they probably would’ve pointed to stacks of aerial photos and the analysts who were verifying the land grid and putting wells in their exact location.
If you asked someone in the 1980s, “What is big data?” they probably would’ve pointed to the terabytes of 3D seismic data and the dozens of attribute volumes.
With the advance of new geophones, logging tools, and sensors, we now collect a dizzying amount of data. The upstream oil and gas industry probably collects more information than any other industry, and most workers will tell you they never have enough.
Today, big data isn’t about those terabytes of 3D seismic data. In some respects it isn’t even about the amount or mass of data. It’s about the diversity of the data. We’ll get back to that thought in just a few minutes, but first I’d like to describe a few terms that people throw out as part of analytics. Those terms are descriptive analytics, predictive analytics, and prescriptive analytics.
I’ll use something that every one of us uses every day – weather analytics – to put these terms in context. In the U.S., the National Oceanic and Atmospheric Administration gathers data from a host of instruments from a host of locations to give you information about the weather.
Descriptive Analytics – It's currently sunny. The sky is clear. The temperature is 76 degrees Fahrenheit and the wind chill is 72 degrees. Technically, descriptive analytics is the analysis of data from the past (even if it is from a few seconds ago). The description you receive can be anywhere from a few minutes to an hour old. It is the gathering and synthesis of multiple weather instruments to tell you in simple English the current weather conditions.
Predictive Analytics – Scattered showers this afternoon. Winds will be from the north at 10 miles an hour. Temperatures in the 40s. This uses more complex analytics, including a time-predictive model, which includes weather readings from other stations as well as wind models.
Prescriptive Analytics – Bring a light coat and umbrella with you to work this morning. The wet roads will require more time to travel. You will need to leave for work by 7:52 to make your 8:30 meeting. This includes everything from the predictive model, but it also needs to know a few things about me: where I work and what I do. If I were a sea captain, it would’ve told me to bring my foul weather gear (an umbrella would be silly) and some Dramamine for my chronic sea-sickness.
I need to know the weather forecast to know how to dress and be comfortable while I’m outside. Prescriptive analytics tells me how to adjust to events that will happen. In some respects it skips the details and advises how to dress for the day, which is why we listen to the weather forecast in the first place. If you’ve ever used Waze on your phone, you know that it doesn’t tell you about the traffic; it just shows you how to get around it.
In my world, a complete prescriptive system also would’ve adjusted the wakeup time on my alarm clock to account for the longer commute. As you can tell, prescriptive analytics isn’t completely there yet in many industries, but we’re very close to it.
Many industries use predictive analytics to predict machine failures and schedule maintenance – this saves companies millions of dollars each year in downtime costs and revenue losses. Valves and systems shut down automatically when processes become unstable or dangerous.
We have used simple systems like this in the oil field to monitor production or drilling. However, we are only now using big data in conjunction with analytics to help us determine where to drill and how to drill and complete a well.
Someone once said, “Every unconventional well we drill is an experiment.” The sentiment is correct, but the reality is wrong. If we were running a real experiment, we would drill a well, then change one thing, then drill another and compare its success with the previous. We would do it over and over again, changing just one parameter each time until we knew exactly what factors made for a good well and those that made for a bad well. In actuality, many things change from well to well, some in your control and some out of your control. The rocks, fractures, and total organic carbon (TOC) can all be different.. Mechanical things can also happen that might yield different results.
So how do we determine which things are required to make a well a good well and those things that contribute to poor well performance that we shouldn’t do? There are hundreds of variables in play at the same time – how do we make sense of everything? Answer: Analytics (there’s that buzz word again).
OK, that’s a bit like saying the best way to win at baseball is to use a bat and a ball. But there are a lot of things that need to happen and you need to do correctly to win a ball game.
Some analytics can help you uncover the answer. There are all kinds of analytics – one size doesn’t fit all. Asking for the current price of oil is one thing. Asking for the price of oil tomorrow is a completely different question.
There are many books written on analytics, so I could never hope to do any more than just scratch the surface in this column. However, I’ll talk about some generic things. (In future columns I’ll dig deeper into both analytics and oil and gas data).
In the upstream oil and gas industry, I like to classify analytics into two groups: generic and fit-for-purpose.
Generic analytics tools could include simple tools like Microsoft Excel or LibreOffice Calc. These are both spreadsheet applications that can yield an enormous amount of information if you have the time and the technical skills to manipulate them to do what you need. LibreOffice is free and my personal favorite.
The popular ones that handle a lot more data and provide a higher level of analysis are Spotfire, Tableau, and to a lesser degree the open-source options. They include R, Google Fusion Tables, Impure, Tableau Public, ManyEyes, and VIDI. I confess I haven’t tried all of these, but if you have some basic programming skills, R is definitely something to take a look at (many people prefer R over the commercial packages). Google Fusion Tables is easy to use, especially if you use Google Drive. They are very different packages, but for what they do, they can yield very impressive results.
All the tools mentioned above require a certain amount of expertise in both the tool and the method to use them effectively and achieve the desired results. It’s this “toolbox” that makes each tool powerful, but it’s the same toolbox that can make each one difficult to use and learn as well.
The fit-for-purpose tools that have emerged in the past few years have been absolutely amazing. Without mentioning names, there are a host of new companies that have sprung up in the past few years that are focused on solving specific problems. Some of them have produced general interfaces that take the user through several workflows, whether it’s buying production, selling production, or determining why some area has better production than another.
The technology used ranges from simple R to unsupervised neural networks with self-organizing maps to physio-statistical predictive analytics that run a million scenarios on the data to simulate any and every possibility and then return probabilistic answers with risk.
There are also several boutique firms that can customize a solution based on a company’s specific needs and the data available. Like traditional consulting firms, the customer is given an answer and a custom tool that can be used in the future.
So, let’s summarize…
Big data and analytics are the current buzz words in the upstream oil and gas industry. The data you collect today might hold the key to unlocking some important insights in the future. The tools available in analytics are what turn the key, if you will. Although analytics is a broad and general term, these tools can provide you with precise information about what has happened, predict what will happen, and give insight into what course of action should be followed.
Know your data. Any analytics should be done in a context with solid geological and engineering principles. In other words, if your analytics vendor tells you that the analytics reveal that you should spud your wells on even days of the month because it will maximize your production, it’s probably the right day – odd or even – to look at a different analytics provider.
Next time I will talk more about data and the importance of understanding the data you have – what it is, how it was collected, and where it came from – before you run any analysis.
John Fierstien is Director of Product Management for P2 Tobin Data. He started his career in oil and gas in 1978 after finishing his MS in Geology from the University of Pittsburgh and his BS in both Geology and Biology from Central Michigan University. He has worked as both a development and exploration geologist. John has been a product manager in oil and gas for the better part of the last 20 years. He’s also spoken at various meetings and conferences and written about sub-surface modeling, oil and gas software, and oil and gas data. John enjoys photography and growing his home automation system. John currently lives west of Austin, in the Texas Hill Country.