Photo by Danil Sorokin on Unsplash
In the E&P planning domain, it is tempting to spend a lot of time talking about algorithms, metrics, solvers, and all of the powerful technology that underlies modern decision science. I can’t help but gush about how far things have come since I started working in business modeling and optimization over 20 years ago. It’s fun to talk about the flashy stuff, especially in today’s environment where AI and analytics are such hot commodities. Notwithstanding the appeal of AI and analytics, we need to remember where everything starts: data.
The happy path for analytics is where we go without thinking, as a function of the tools and data before us.
It comes as no surprise that the quality of planning data impacts the quality of the planning outputs– garbage in garbage out. What is often overlooked is that the nature of planning data significantly impacts the planning process itself– the medium is the message. Available data establishes an analytical path of least resistance, down which any analyst is likely to be drawn, especially under time pressure and stress. The happy path for analytics is where we go without thinking, as a function of the tools and data before us.
We must account for the gravity created by the data... that we provide to users.
The happy path of analytics implies a need for much more than just ETL (extract, transform, load) and data hygiene. We must account for the gravity created by the data (and structure of the data) that we provide to users. Thankfully we’re not talking about a crazy three-body gravitational problem. Expanding our concept of data, from seeing it just as an input to considering how it influences our perceptions and analytical decisions, suggests some general principles for sourcing and formatting data.
the more data preparation required, the greater the opportunity for prejudices and systematic errors to enter the decision making process
At the highest level, the more data preparation required, the greater the opportunity for prejudices and systematic errors to enter the decision making process, not to mention friction, frustration, and operational risk. Specifically for E&P planning data, we often see complex multistep workflows. E.g. well cases go somewhere for analysis to generate type cases, which go someplace else to be married with rig data, which then get joined with operations and financial constraints, and finally get pulled into a planning module. It is not uncommon for planning data to be five or six steps removed from the source of record, with multiple transformations in between, and no opportunity for planners to view the original data or control any of the transformation steps. The telephone game consequences of extensive data preprocessing are the same, whether the steps are performed in separate software systems or within a monolithic product suite.
planning decisions and reserves data share a mutual feedback loop
At the detailed planning level, the situation gets even worse. The critical point is that planning decisions and reserves data share a mutual feedback loop. The forward loop is clear: reserves numbers are the basis of many economic parameters that are essential for financial planning. The reverse loop is more subtle: development plans impact how PUDs are booked, so planning projects must be tied back to actual PUD locations in the reserves data. Failing to tie planning decisions directly back to reserves data creates the need for a whole other set of business processes to correct reserves numbers once the planning process is complete. It is better to close the loop between planning and reserves, and avoid all of this complexity in the first place.
a better approach is to establish the shortest possible path to the source of record
Instead of introducing additional complexity with extensive preprocessing, a better approach is to establish the shortest possible path to the source of record. For E&P planning, reserves data is the ultimate source of truth. However, data schemas for reserves systems are notoriously opaque, and reserves systems typically don’t play well with others. Integrated product suites purport to unify planning, reserves, and operations data, but under the hood such suites are amalgamations of disparate acquired products. Reserves systems are complex enough that legacy code hangs around for decades, and unless vendors reimplement new reserves solutions from the ground up, all of the same issues of the legacy product will persist.
No amount of heaving, banging, bypassing, or extending data flows between legacy systems will result in the clean and direct flow that is necessary
So how is it possible to flange up planning and reserves data? The key lies in implementing a single shared data model between planning and reserves data. The SPE Petroleum Resources Management System (PRMS) and Professional Petroleum Data Management (PPDM) standards provide a starting point, and should be regarded as table stakes for any compliant E&P reserves and planning system. Implementing those standards with a holistic, efficient, and enduring design is the crux of the issue, which is only possible with simultaneous development of reserves and planning functionality, starting with a clean slate. No amount of heaving, banging, bypassing, or extending data flows between legacy systems will result in the clean and direct flow that is necessary to make planning data synonymous with reserves data.