You are Home   »   News   »   View Article

How to get value from analytics - Teradata

Friday, January 15, 2016

Other industries have managed to get value from analytics – it is time the oil and gas industry followed, said Teradata’s Oil & Gas Practice Lead,Dr Duncan Irving

In the oil and gas industry it is typical for data to be handed from one expert to another along the workflow, with nothing captured about where the data has come from and whether it can be trusted.

“By the time it has got to the person who needs to make the decisions it’s got to PowerPoint, and there's no understanding of the algorithms (used to create it) or any sense of data lineage – time is wasted re-establishing trust in the data”, said Duncan Irving, principal consultant, Teradata Oil and Gas Team.

“This is the opposite of pretty much any other industry Teradata works in.”

Many industries operate on much thinner margins than the oil and gas sector, but they still thought it was worth spending money on analytics, he said.

Mr Irving’s goal is “trying to explain to the oil industry what this could do for them,” he said.

Analytics systems should “become a more robust and significant part of the IT infrastructure in the upstream domain."

Oil companies are attaching sensors all over the platform which generate data for the life of the field, but “we are not getting maximum value from it,” he said.

Many companies keep data siloed in different departments and domains, with very little data communication between them, and this makes it harder to implement new processes and workflows at scale, he said. “You have to understand the challenges in your own architecture,” he said.

You could say that oil and gas companies have 3 types of data – transactional, technical and scientific.

Transactional data is stuff in the Enterprise Resource Planning (ERP) systems, SAP systems, operations and finance data. Technical data could include designs of casing shoes and well trajectories. Scientific data can include flow simulations and seismic data.

All of this data “comes in different flavours,” he said.

Unconventionals and data

The North American unconventionals business has done much more with data.

For example, ConocoPhillips used analytics to work out the best way to reduce the number of shut-ins (when drilling needs to be stopped, for example due to high pressure gas entering the well).

They managed to reduce the number of shut-ins by 90 per cent, Mr Irving said.

ConocoPhillips is also using analytics to work out how to optimise maintenance on wells and tank emptying, he said.

It has improved its logistics, building on work done by a US parcel carrier, which worked out that (for example) it may be faster to do three right turns around the block than one left turn (the US drives on the right). They found certain routings led to less wear on the vehicles.

The company has been tackling questions like, what is the optimum well spacing (how many wells per mile) - and what is the best way to build the substations, roads and other infrastructure.

The company uses analytics as a continual guide to operations, he said. It brings together an integrated view of “what is the most effective way to develop my fields,” keeping OPEX as low as possible.

The unconventional oil and gas business was basically a re-invention of the oil and gas industry, which meant an opportunity to bring in many new processes. The unconventionals business “looks a lot more like a factory than what we have offshore,” he said.

Retail, manufacturing and refineries

Consider that Walmart has a supply chain so tightly integrated, that by the time you’ve loaded your shopping into your car, replacement items have been loaded onto a truck at the depot. “That's how integrated their supply chain is,” he said.

Walmart has an integrated understanding of how to access the biggest ‘wallet share’ of a neighbourhood, with different prices, promotions and marketing. “They understand the environment in which they are operating.”

It also sees its business as selling shelf space to suppliers, as much as it is about selling everything to consumers, and uses analytics to maximise what it can do with its shelf space.

Looking at car manufacturing, Daimler is doing 190,000 types of component failure analysis akin to ‘decline curve analysis’ work every night based on data from the production line, to work out if there is something which should be tweaked. “They understand the total cost and profitability of every piece of plant on the production line,” he said.

Oil refineries are also getting good with analytics. At one refinery, the refinery managers have sat together with the traders to design a mutual dashboard, so the products being produced can be adjusted according to near real-time changes in the markets.


Mr Irving uses the word ‘sentient’ to describe where he thinks oil and gas companies need to go.

“This is a bit of a hard concept”, it basically means to have a better feeling about what is happening,” he said.

This is a “new style of industry, you have organic sharing of data, you have organic sharing of experience,” he said. “You have infrastructure which allows everyone to add the value they need to using tools they are familiar with.”

If this was happening with offshore oil and gas, all the various technical domains should have much better insight on how their equipment is running today, what it is contributing to the business bottom line, how it is operating together with other facilities, what should be developed next, what rewards are available, and does this link together with understanding of the subsurface.

People could quickly answer questions like, ‘how to optimise the scheduling of maintenance tasks,’ ‘which bit of the reservoir to drill into next,’ ‘are we depressurising the reservoir too early.’

Operational areas

It gets much tougher implementing collaboration tools and analytics for operational areas of the oil and gas business, compared to strategic departments, he said. In operational areas, people are under constraints of cost and time, which make implementing software harder.

There is also more need to build bespoke software tools, because everybody works differently, and every team has a different mix of skill sets. For example, one company might want a specialist software tool for monitoring a reservoir using live data from permanent reservoir monitoring equipment, he said.

Discovery analytics

The “gateway” class of analytics is ‘discovery’, which basically means looking through the data for new insights that potentially could be put onto an operational footing.

As an example, consider that auction website eBay has a large staff of data scientists continually probing its live database to try to find patterns. About 60 per cent of the workload on its master database is from internal data scientists, he said. As a result, they might discover that one user prefers to click on red buttons rather than green ones, and only show them red buttons from now on.

In a similar spirit, Teradata has given a data science project to a Petroleum geoscience MSc student at the University of Manchester, to try to find overlooked stratigraphic relationships by doing analytics on a large amount of wells data, he said. This discovered badly interpreted facies and unseen hot shales, both leading to a re-appraisal of the entire basin.

Enabling data management

Teradata sees its role as ‘enabling data management’ – or helping the people who get very close to the data to help the company do more.

Many of the people employed by data managers are highly qualified, with PhDs in geology or data physics. “They got into data management because they were the ones who cared the most,” he said. “Once you get tarnished with that brush it is very hard to get away from it.”

“But it is a waste of talent, having someone that clever who is responsible for knowing it’s on this or that file system, instead of enabling value to be extracted with the data.”

“That person has to make sure all the metadata and master data. Someone has to care enough to do that and be knowledgeable about it.”

The ‘big data’ label is not very helpful, he said. “To me, big data is anything where you take in data too quickly for your systems and processes to deal with. If your systems are Kingdom suite and a load of spreadsheets to run a brown field - you've probably got a big data problem. If you need to extract value from SCADA in your historians, and SAS analyses on your SAP logistics reports across a regions assets, then that's defintely a big data problem.”

Looking at subsurface data, managing 2D alone can be a ‘very difficult computational problem,” when you get to 3D or 4D (with time) it gets difficult.

You have to find a way to make this data available for the life of the field, not just lock it away in your favourite software – and you have to manage security and safety.


Usually the best language for analytics is SQL, because that is the language which business people’s systems use.

“If you want to show the CFO something - that has to be something he understands – ideally a few numbers (small data!) and a graph. That’s probably the thing he's already got - and it’s probably powered by SQL.”

IT staff in different domains speak different languages. IT staff serving the business tend to use Structured Query Language (SQL) for interrogating databases. Engineers like languages like Matlab, and some scientific people do other languages. “I was talking to seismic people yesterday, they kept saying 'we do this in FORTRAN,” he said.

The challenge is to pull all of these IT domains together under one umbrella for the CFO’s integrated “widget”. It is data-driven analytics which enable these new kinds of transformation insight and as other industries have seen, complement (rather than replace) the experience of the decision makers who have access to such large amounts of data and computational power.

Associated Companies
» Teradata
comments powered by Disqus


To attend our free events, receive our newsletter, and receive the free colour Digital Energy Journal.


Latest Edition Oct-Dec 2023
Nov 2023

Download latest and back issues


Learn more about supporting Digital Energy Journal