When Trading Teams Stop Trusting Their Own Data
Key Takeaways
- Data quality failure in trading operations is a behavioral problem as much as a technical one
- Reconciliation failures get absorbed into workflow because fixing them has no clear owner and no clear cost
- AI can make these problems visible faster, whether organizations are ready or not
The number gets chosen more than it gets produced.
Risk positions assembled end of day from multiple sources. A reconciliation step that started as a quality check and became the process by which the number gets built. A spreadsheet that bridges two systems nobody fully aligned, maintained by whoever was there when it was built.
From the outside, it looks like a team managing complexity. But when fixing it has no clear owner and no clear cost, the problem persists.
Table of Contents
- How Data Workarounds Become Standard Operating Procedure
- Why Data Quality Problems in Trading Operations Don't Get Resolved
- Signs That Data Trust Has Broken Down in a Trading Operation
- How Data Trust Failure Affects Risk Management
- Why AI Makes Unresolved Data Problems More Urgent
- Frequently Asked Questions
How Data Workarounds Become Standard Operating Procedure
Power trading operations handle real data complexity: positions across books, marks from multiple curve sources, settlements that touch scheduling, finance, and risk on different timelines. When those systems produce numbers that don't align, the fix requires knowing which system is wrong, who owns it, and how long resolution will take.
In many operations, these answers aren't obvious. The problem sits somewhere between the ETRM, the scheduling system, and the settlements workflow.
The makeshift solution, for many teams, is to find a way to work around it. A spreadsheet gets built to bridge two systems that don't agree. A manual step gets added before the morning report goes out. A reconciliation that was designed as a quality check becomes the process by which the number gets built in the first place. Soon enough, the patch outlasts the problem it was built to solve.
Why Data Quality Problems in Trading Operations Don't Get Resolved
Fixing a data trust problem in a live trading operation is slow, risky, and nearly impossible to scope from the outside. Common reasons:
- Formulas don't have a clear owner. If they were built under time pressure, with logic that made sense at the time, those processes are rarely documented. Auditing them means reconstructing decisions that nobody may remember.
- Logic predates the current team. If implementation was years ago, configurations set during that time may not have been revisited since.
- The same term means different things in different systems. Although each system may be internally consistent, the numbers won't align.
- Nobody owns the gap. A mismatch between trading, risk, operations, and finance lives in the handoffs. No single team is accountable for where the number goes wrong.
- When reconciliation produces a mismatch, it gets resolved locally. The person who noticed it fixes it, but the root cause goes unaddressed. Instead, the workaround gets added to the stack.
Data trust failures can be difficult to spot because workarounds often succeed in the short term.
Signs That Data Trust Has Broken Down in a Trading Operation
Trading operations show specific signs when data trust has eroded, and most of them appear in daily workflow before they surface in risk reporting.
The reconciliation step has become a build step
The morning run starts before anything else can move. By the time the number is ready to share, an hour has passed. A position that exists inside the system still can't be used until someone has verified it against something outside of the system.
Shadow systems run alongside the primary system
The systems that actually run the operation aren't in any system. They live in files and scripts maintained by specific people, built to bridge gaps that were never formally closed. Nobody owns them, and nobody fully understands them except the people who built them.
The same question gets two defensible answers
Two people asked the same question will produce two defensible answers, each drawn from a different source. The conversation that follows is familiar to everyone in the room. It happens regularly enough that nobody questions why it keeps happening.
The risk function spends more time explaining than preparing
The number is right, but explaining it takes as long as producing it. More of the function's time goes toward reconstructing the last position than informing the next one.
How Data Trust Failure Affects Risk Management
The hours spent on reconciliation each day are real. So is the concentration of knowledge in the workarounds themselves: what they do, when they apply, and why they were built.
But the larger problem is what happens to risk confidence over time.
When the number in a risk report is one the team chose rather than one the system produced, the report carries assumptions that aren't documented anywhere. The people who made them understand them. Leadership doesn't. Auditors won't. When market conditions shift and someone asks why reported exposure diverged from realized P&L, the answer lives in a layer of reconciliation logic that nobody fully owns or is able to explain.
Risk management depends on a view that holds up when a decision has to be made — when that view is built on assumptions that live outside the system, the organization is carrying risk the system isn't showing.
Why AI Makes Unresolved Data Problems More Urgent
In a 2024 study of 565 data and analytics professionals conducted by Drexel University's LeBow College of Business and Precisely, 67% said they don't completely trust the data their organization uses for decision-making. Only 12% said their data was of sufficient quality and accessibility for effective AI implementation. In energy trading, that distrust has a specific shape and a specific cost.
AI won't resolve the underlying problem. But it will accelerate whatever is already there.
When a risk team asks an AI tool what the net open position is, the tool answers from the data source it's connected to.. If the reconciled version lives in a shadow spreadsheet rather than the system of record, the answer comes back confident and complete. The tool has no way to know its source is compromised.
Ultimately, an organization that can no longer tell the difference between what it knows and what it believes is already carrying risk — whether or not anything has gone wrong yet.
Frequently Asked Questions
Why do trading teams stop trusting their own data?
Data trust erodes when reconciliation failures get absorbed into workflow rather than resolved at the root. Teams build workarounds because fixing the underlying problem has no clear cost and no clear owner. Over time, the workaround becomes standard operating procedure.
What is the real cost of data quality problems in energy trading?
The visible cost is time: hours spent on reconciliation, manual checks, and maintaining shadow systems. The compounding cost is risk confidence. When the number in a risk report is chosen rather than produced, decisions rest on assumptions that aren't documented and can't be defended under pressure.
Why don't trading firms fix data quality problems when they identify them?
Because the fix is genuinely hard to scope and own. Data trust problems fall between multiple teams with no single clear owner. Unclear ownership and unclear fix costs often make local resolution the path of least resistance.
What is the difference between a data problem and a data trust problem?
A data problem has a defined source: a broken feed, a bad integration, a calculation error. A data trust problem is operational: teams have built verification processes around system output because they've stopped relying on it directly. The system may be functioning correctly. The reliance is gone.
How does data trust failure affect risk management outcomes?
When risk teams can't fully trust what they're publishing, more of the function's time goes toward explaining past positions than informing current decisions. That shift is hardest to see from the outside and most consequential when markets move and the operation needs clarity quickly.
