Konstantin “Stan”
PonomarevProgram & portfolio leadership
Menu
All case studies

Data platform delivery · Platform delivery

Data platform delivery 3 min read

Comparable production data with Azure Databricks

At a glance

The challenge

Make historical and current production data comparable for leadership reporting and field analysis.

My responsibility

I owned scope, budget, hiring and internal coordination, and aligned requirements and comparison rules.

The decision

Begin with two sites and three feeds; reconcile the data and confirm receiving-team readiness before expanding.

Delivered uses

Production dashboards and contextual field analytics delivered through a platform using Azure Databricks.

Context
Energy company · Internal data platform
Period
Main project work · 2021–2022
Explore the delivered uses

Context

How the work connected.

Background and context

The internal platform brought historical and current production data into a shared environment for reporting, analytical applications and machine learning-related work. I owned scope, budget, hiring and stakeholder coordination. I organized a core team of about 12, with vendor and research participation bringing peak staffing to about 20, within a roughly $6M two-year planning envelope. Architects and engineers owned technical design and implementation; I aligned business priorities, capacity, dependencies and receiving-team readiness.

Historical records and current production figures could not be compared safely just by putting them on one dashboard. Names, units and well identifiers differed, as did reporting periods, refresh timing and corrections. Leaders needed to compare fields while accounting for field size, operating wells, development stage, geological conditions and plan versus actual performance.

The decision, illustrated

Comparable data needed common meaning.

Historical records and current source data enter a translation bench. Different local identities are aligned into common rules before reaching a shared comparison view.
Historical and current data
Production records and current figures used different names, units, well identifiers and reporting periods.
Common rules
The teams reconciled identities, units, periods and corrections before publishing comparable data.
Useful comparisons
Leaders received production dashboards and contextual field comparisons.

A metaphor for the delivered data work, not a physical Azure architecture. I led delivery; architects and engineers owned technical design and implementation.

operational systems in the wider environment
~100
production sites in the wider environment
35
main project years
2021–2022

Judgment

Connect business meaning with delivery priorities.

The central trade-off

Launching more sources quickly could increase coverage while leaving inconsistent meanings unresolved. Common rules required agreement across departments, and the first release had to fit available people and time. I worked through these linked constraints with the participating teams.

  1. 01

    Define the comparison with its users.

    I aligned the needs and priorities of internal leaders and defined or agreed the metrics and comparison rules. The business uses were concrete: production dashboards for the Head of Upstream, and contextual field comparisons, trends and recommendations for subsidiary leaders.

  2. 02

    Hold publication until the figures reconcile.

    Azure Data Factory coordinated source loading and jobs; Azure Data Lake Storage Gen2 retained inputs; Azure Databricks mapped, transformed and checked the data. When an unknown well code prevented reconciliation, the team held publication, agreed the mapping with the domain owner and replayed retained input. I coordinated the dependency and release decisions; the teams implemented the data work.

  3. 03

    Start with a release the receiving teams could accept.

    The first release focused on two sites and three feeds, connecting production actuals, well identity and plan data. Later waves extended that priority track to ten sites and nineteen feeds within the wider environment. I balanced scope, shared specialist capacity and release timing. Data owners accepted mappings, users accepted the output, and operations accepted monitoring and recovery before release.

Outcome

Reporting and analytics that people could use.

  • The team delivered production dashboards to the Head of Upstream and contextual field comparisons, trends and recommendations to subsidiary leaders.
  • Common dictionaries, mappings and reporting periods made historical and current production figures comparable; retained input supported correction and replay.
  • The platform also supported well or geology analytics, equipment-related predictive modeling and corporate management reporting.
  • For later company and vendor product connections, my retrospective estimate was about two weeks instead of six months. This is not a measured average across all systems or an effect attributed to Databricks alone.

Responsibility

My role and the scope of this case.

I owned delivery scope, budget, hiring and stakeholder coordination; architects and engineers owned technical design and implementation. This was a separate internal platform project from the earlier foundation-governance work and wider transformation program. The roughly 100 systems and 35 sites describe the wider environment, not the first-release footprint. The roughly $6M figure is the two-year planning envelope, not a claim of actual spend or savings. The later-connection comparison is my retrospective estimate for company and vendor products, not a measured average across every system.

Hiring for similar responsibilities?

Discuss a role

A conversation about the work

Let’s talk about
the work ahead.

If you’re hiring a leader for complex programs, portfolios or products, I’d welcome a conversation.

Discuss a role

Based in Tampa, FL

Program & portfolio leadership
Product & AI leadership

U.S. work authorization
No sponsorship required

One-page profile (PDF) Connect on LinkedIn (opens in a new tab)