How a major North American upstream operator replaced a fragile, expert-dependent data quality process with a continuous one, and went from 125 chained scheduled jobs to 4 continuous processes.
The team: A data governance advisor with 40 years in oil and gas, responsible for well data quality across the enterprise for the past 20.
This operator had already solved the data quality problem once. In 2016, they built a solution that automated data loading, applied business rules, and improved the quality of well headers and directional surveys across their Kingdom and GeoGraphix projects. It worked for eight years.
But keeping it running required what their data governance advisor called “a unicorn”: someone who understood the business day to day, had the technical knowledge to support the solution, and understood the data deeply enough to keep it all alive. That person had become a single point of failure for the company.
“I was turning into a point of failure for our company. And a solution shouldn't really require this type of expertise.”
For a single asset, over a single week, the team saw an average of 5,000 changes to well data. Most users assume it's a handful. That gap is why manual, scheduled processes fall behind, and why so much good work gets lost.
When a geologist finds an issue in their project, they fix it and keep working. The correction lives in one project, never reaches the others, and gets overwritten on the next load. Multiply that across teams and offices, and the same well has different answers in different systems.
“The biggest challenge is users don't trust the data that they're using to make their decisions.”
A custom in-house replacement was off the table. In this advisor's experience, custom solutions take a year to implement, last a couple of years, and then decline as the people who understood the code leave. Other proofs of concept hit the same wall: heavy resources, heavy customization, and weaker exception handling than they already had.
They ran a proof of concept with ODIN on one asset team in January 2024. It took about four weeks. The requirement was clear: keep everything that made the old solution successful, and remove every roadblock to scaling.
Under the old model, nobody actually knew the state of data quality at any given moment. It depended on when the last job ran. Now the picture is always current, across every project.
“No more scheduling jobs. Say it again. No more scheduling jobs. Game changer.”
ODIN arrived with subsurface and data management expertise built in: coordinate system transformations, unit handling, matching the same well across sources, and pre-built adapters for the operator's Kingdom and GeoGraphix projects.
What made the unicorn unnecessary was guided configuration: a short set of questions that set the solution up correctly. Who should participate? What data should be managed? Where are the sources? Which attributes are critical and which are important? And most importantly, how should the data flow?
That last question is where trust is built. Classifications let the team control exactly how corrections move, so routine changes flow automatically while changes that need judgment go to a person. Setting those classifications reduced exception handling by 75%.
“You don't just blast the data and automate everything. You need to control it. This is where you build trust with the users.”
Deployed June–August 2024 across 4 asset areas, 17 data sources, and 10.8 million data objects, with few resources.
Initial resource load, total team, not per person.
Today one person manages all four areas, running continuously, with no scheduled jobs and a holistic view of data quality across every project.
This team has continued expansion to support additional sources and a more extensive list of data types, scaling ODIN further across their organization.
“The deployment was so smooth I didn't hear a peep from anyone, which is remarkable considering it touched all of our active exploration and development projects. Our users now fully trust the automated process. Most importantly, I can sleep easier knowing the solution won't disappear when key personnel retire.”
A 40-year industry veteran walks through the full journey, the challenges, the numbers, and how the process works, in her PPDM conference presentation.