Variance analysis is the work of comparing what you expected with what actually happened, and explaining the difference. In production operations, that means planned volumes against actual volumes, forecast against delivered, budget against spend, and the daily question underneath all of them: where did the barrels go?
The explaining part is valuable. It’s where deferments get understood, where creeping losses get caught, and where next month’s plan gets better than last month’s. The problem is that in most operations, the explaining is the last hour of the job. The first three days are spent finding the numbers.
Where the three days go
The pattern is familiar enough to hurt. The plan lives in a planning tool or a spreadsheet. The actuals live in the allocation system, or the historian, or both, disagreeing slightly. The budget lives in the ERP. None of them talks, so the variance work starts with exports.
Then the real fun begins. The plan is monthly by field; the actuals are daily by well. The calendars don’t line up. Someone’s export was Tuesday’s version, and the allocation reran on Wednesday. So the analyst builds the bridge by hand: reshape, match, reconcile, and hope, before a single difference has actually been explained.
By the time the variance is found, dated, and understood, the operation has moved on. The meeting it was needed for has happened. As the saying doesn’t quite go: that’s not analysis, it’s archaeology.
What the delay actually costs
It’s tempting to file this under mild inefficiency, but the cost is higher than the three days.
- Deferments get explained too late to act on. A variance spotted the same day is a valve, a choke setting, a pump to fix. The same variance spotted at month-end is a paragraph in a report.
- Small losses get normalised. When finding a difference costs days, nobody chases the small ones. They quietly become the baseline, and the baseline quietly erodes.
- Your engineers do plumbing instead of engineering. The people qualified to answer “why did this well underperform?” spend their time answering “which of these files is current?”
The stakes are not small. McKinsey’s benchmarking of global oil and gas operations has found production efficiency spreads of more than 40 percentage points between the weakest assets and the top quartile in the same basin. Deferments are a core part of that gap, and you can’t recover a deferment you haven’t found yet.
How to shorten it
The fix isn’t a faster analyst. It’s removing the archaeology, so the analysis can start on day one, hour one.
- Plan and actual live in one environment. When the production plan and the allocated actuals sit on the same platform, the variance isn’t something you assemble. It’s a column.
- Variances are computed as data lands. Daily actuals arrive, get validated, and are compared against plan automatically. The differences are waiting for your team, not the other way round.
- Exceptions come to you. Validation by exception means the system flags what’s outside tolerance, and people look only where they need to. When a Queensland CSG operator configured exception-based field data validation, the well count each engineer could confidently manage went up, because their attention stopped being spread across everything.
- Every variance drills to source. From the difference, straight through the allocation, to the meter reading behind it. No exports, no version doubt.
- Explanations are captured once. The reason for a deferment gets recorded against the variance, feeding the daily report, the monthly pack, and next year’s plan, without being retyped into any of them.
None of this requires the analysis itself to be automated away. It requires the finding to be.
The question only your people can answer
Software can compute a difference. It cannot tell you that the well underperformed because of the workover slipping a week, or that the plan was optimistic because last year’s decline curve flattered the field. “Why” belongs to your production engineers and your accountants, and it always will.
That’s the real argument for fixing variance analysis: not to replace the expertise, but to spend it properly. Three days of finding and one hour of explaining is the wrong way round. Our production operations solutions exist to invert it: with plan, actuals, and allocation on one platform your own team configures and owns.
Bring your longest hunt
If there’s a variance your team is still chasing from last month, that’s the perfect test case. Book a call, and we’ll match you with a partner who knows your domain and can show you what same-day answers look like on your own data.
