The Quiet Lie We Tell Ourselves About Terminal Decline

Author: Mathias Lia Carlsen
Written with Raul Rodriguez from Oxy

Let’s talk about something everyone does, few people love, and almost no one feels great about.

Terminal decline.

If you’ve ever built a forecast, you know the moment. The well is old. The curve is flattening. The hyperbolic tail refuses to die. And someone eventually says:

“Let’s just use 6 percent.”

Six percent is the energy industry’s version of “trust me, bro.”

Sometimes it’s fine.
Sometimes it’s wildly wrong.
And most of the time, it’s not really grounded in the data you actually have.

Why Terminal Decline Is So Awkward

Early time behavior is noisy.
Mid-life behavior is chaotic.
Late time behavior is… sparse.

So we shortcut.

We borrow assumptions from other basins.
We reuse numbers from slides made years ago.
We convince ourselves that late-time wells all kind of behave the same anyway.

They don’t.

But here’s the uncomfortable truth: we now have enough data that guessing is no longer necessary.

The Simple Idea That Changes the Game

Instead of arguing about what terminal decline should be, ask a better question:

“What do wells actually decline at once they’ve been producing long enough?”

Not hypothetically.
Not in theory.
But in real data, from real wells, with long production histories.

So you take a big population of mature wells.
You focus on the last part of their decline.
You track how that decline rate evolves with producing time.
And you wait until it stabilizes.

That stabilization point is the key.
That’s when the well stops pretending to be young.

Filtering Is the Unsexy Superpower

Here’s where most workflows quietly fail.
Not all wells deserve a vote.

Wells with gaps.
Wells with operational chaos.
Wells that went offline and came back from the dead.

They all distort the signal.

So you filter aggressively.
Only continuous producers.
Only long histories.
Only wells that have earned the right to define “terminal.”

What you lose in sample size, you gain in truth.

When the Data Finally Settles Down

Once wells pass a certain age, something interesting happens.

The decline rates stop drifting.
The scatter tightens.
The trend flattens.

This is the moment you’ve been waiting for.

At that point, you don’t need a single magic number.
You have a distribution.

Now you can ask grown-up questions:

• What does P50 actually look like?
• Does it differ by formation?
• Is it stable across multiple late-time windows?

And suddenly, terminal decline stops being an assumption.
It becomes a measurement.

The Real Win Isn’t the Number

Yes, you end up with a better Dmin.
Yes, your EURs calm down.
Yes, late-time tails stop inflating quietly.

But the real win is credibility.

When someone asks, “Why did you choose that terminal decline?” you don’t say:
“It’s standard.”

You say:
“This is what hundreds of long-lived wells actually do.”

That changes the conversation.

The Big Takeaway

Terminal decline was always treated like a necessary evil.
Something you slap on at the end of a forecast and hope no one zooms in too far.

It doesn’t have to be that way anymore.

We have the data.
We have the compute.
And we finally have workflows that scale beyond a handful of hand-picked wells.

Guessing was reasonable when we had no choice.
Now it’s just lazy.

Authors: Mathias Carlsen (whitson) and Raul Rodriguez (Oxy)