"Petroleum engineer" is one job title covering sub-disciplines that share a hallway and almost nothing else about their day. This is the one that owns the question every investor actually asks: how much is down there, and how fast does it come out?

The reservoir engineer's day starts with a number that changed overnight.

Production came in. Somewhere in a few hundred wells, a handful did something they weren't supposed to — a gas well that flattened instead of declining, an oil well that watered out two quarters early, a new completion that came on at half the rate the type curve promised. Before the first coffee is cold, the job is triage: which of last night's surprises are noise, and which are telling you the forecast is wrong.

The morning watchlist

You open your economics-and-decline package — for most U.S. onshore shops that's ARIES (Quorum) or PHDWin (now Emerson), increasingly ComboCurve or whitson for the teams that have moved to the browser — and you pull the wells flagged by the overnight production load. Each one is a small argument with yourself.

Take a gas well that's declining slower than your last forecast. Good news? Maybe. Or maybe the operator choked it back, and the "improvement" is an artifact that will reverse the moment they open it up. You don't get to be lazy about it, because that single decline curve — the Arps hyperbolic you fit through the dots — is the spine of the well's reserves, its PV-10, and its contribution to the number the company reports to the bank and the SEC.

Here's the part outsiders never quite believe: a lot of this is still judgment, by hand. Ask a working reserves engineer how they fit a curve and a surprising number will tell you, flatly, that they don't trust autofit. They set the b-factor from experience in the play, anchor a terminal decline so the well goes exponential at the tail instead of producing to infinity, and drag the curve until it looks right against analog wells they've seen a hundred times. The software's automatic fit is a starting point you argue with, not an answer you accept.

Type wells, and the tyranny of the analog

By mid-morning you're probably building a type well — a single representative decline for a group of similar wells, used to forecast the undrilled locations nobody has data for yet. You normalize a few dozen Eagle Ford or Permian wells to a common start, average them, and fit one curve through the cloud. That type curve then gets stamped onto every proved-undeveloped location in the area, which means a small error in one shape quietly multiplies across the whole development plan.

This is the quiet weight of the job. A drilling engineer who makes a mistake finds out in days. A reservoir engineer who picks a b-factor a little too optimistic finds out in years — and so does everyone who lent money against it.

Material balance, and the rest of the toolbox

Decline curves are the daily bread, but the reservoir engineer's actual mandate is bigger: estimate the volume in place and the recovery factor, not just extrapolate the trend. So the afternoon might turn to material balance — using the reservoir's own pressure response to back out how much gas or oil was there to begin with (the classic p/Z plot for a gas reservoir), or flowing material balance to separate real incremental reserves from the temporary bump you get just by drawing a well down harder. Knowing the difference between "I added reserves" and "I borrowed tomorrow's production and called it a gain" is most of the craft.

Then there's surveillance on the waterfloods and EOR patterns — watching injection and voidage, chasing why one pattern responds and the neighbor doesn't — and, when an acquisition lands on the desk, evaluating someone else's wells under deadline with worse data than you'd ever accept on your own.

Reserves season

Once a year the whole rhythm changes. Reserves season arrives, and for several weeks the reservoir engineer's life is the year-end report: every producing well re-declined, every PUD re-justified, operating expenses and price decks applied, the whole portfolio rolled up under PRMS-2018 and SEC rules, then handed to a third-party auditor who is paid to disagree with you. The auditors, notably, tend to prefer to receive native ARIES or PHDWin project files — which is exactly why those two tools have outlived every prediction of their demise.

It is the most consequential paperwork in the company, and a shocking amount of it moves through spreadsheets, manual exports, and one engineer's institutional memory of why the 2019 wells in that one section always need a higher b-factor.

The part nobody budgets for: moving data around

If you shadowed a reservoir engineer for a week, the single biggest surprise wouldn't be the math. It would be how much of the day is spent getting data into the tool at all — pulling monthly volumes from a state agency or a vendor feed, reconciling three different IDs for the same well (the regulator's number, the operator's internal number, the API number), patching gaps, and re-importing it every month so the forecasts stay current. The intellectual work is fitting the curve. The time goes to feeding the machine.

Where this connects to what we build

This is the discipline Workollab's Auto-DCA engine grew out of, and it's not a coincidence. The two things that define the reservoir engineer's craft — judgment over autofit, and a defensible, auditable trail of every assumption — are exactly the things a black-box tool throws away. We built a browser-native decline engine that fits the same Arps curves, lets you watch the fit happen and override it, validates against a known-good reference (Equinor's open-source library) to a fraction of a part per million, runs on your machine, and keeps every assumption visible. Not to replace the engineer's judgment. To stop making them fight the software to apply it.

The reservoir engineer holds the number everyone else trusts. The least the tools can do is show their work.

Tools & terms named in this piece

ARIES (Quorum Software) and PHDWin (Emerson) — onshore reserves & economics / decline-curve packages; ComboCurve and whitson — web-based forecasting and reserves tools; OFM (SLB) — production-data analysis; Arps decline (exponential / hyperbolic, b-factor, terminal decline); material balance and p/Z; type wells; PRMS-2018 (SPE) and SEC reserves rules; PV-10. Workflow details drawn from current vendor documentation and practicing-engineer accounts, 2024–2026.

One of Workollab's oil & gas engineering pieces. Related reading covers our browser-native decline-curve (Auto-DCA) and PVT engines, and the reserves-engineering workflow gap.

Need to turn this engineering logic into a production-ready tool? Explore built custom engineering workflows for your team.