Ask any completions engineer whether wells are "getting better" and you'll get a confident yes: longer laterals, more proppant, tighter stage spacing. But confidence isn't a number. The question a reservoir engineer, an A&D analyst, or an operator planning next year's program actually needs answered is: in this specific county, does a 2023 well outproduce a 2019 well over the same time on production — and by how much?
That's a vintage question. And it's one the record can answer directly, because every well carries a spud or first-production date and a full production history behind it. Grouping wells into drill-year cohorts turns a vague impression into a curve you can plan against.
Why cohorts beat a single average
A plain county type curve blends everything together — a 2015 well limping along at 15 bbl/d gets averaged with a fresh 2024 completion still flowing 600 bbl/d. The result is a mushy "average well" that describes no one's actual program and understates what modern completions do.
Vintage analysis fixes this by slicing the population by year of first production, then comparing each cohort on equal footing — normalized to months on production rather than calendar date. A 6-month IP from a 2020 well sits next to the 6-month IP from a 2023 well. Now you're comparing like to like, and the year-over-year lift (or plateau) shows up cleanly.
A typical question to the Wellsite data lake looks like this:
"Group all horizontal wells in this county by first-production year since 2017. For each vintage, show me the average cumulative oil at 6, 12, and 24 months."
What comes back is a small table — one row per cohort — that tells the whole story at a glance.
What the pattern usually reveals
Three shapes tend to emerge, and each means something different for a decision.
Steady improvement. Each vintage's 12-month cumulative steps up over the one before — say 85 MBO, then 98, then 110. That's the signature of completion intensity paying off, and it justifies underwriting new drills to the most recent cohort rather than the field-wide average.
A plateau. The steps flatten. 2021, 2022, and 2023 all land within a few percent of each other at 12 months. This is the more common and more important finding: it says the county has hit the point of diminishing returns on bigger completions. If you're modeling next year's wells off a still-rising trend, you're overbooking.
Degradation from spacing. Newer cohorts come in below older ones despite bigger jobs. That's often parent-child interference — infill wells draining into depleted rock or getting hit by frac communication. When a vintage curve rolls over like this, it's a flag to look at well spacing and offset timing, not completion design.
Normalize, or the comparison lies
A vintage table is only honest if the cohorts are comparable. A few adjustments matter:
- Lateral length. If 2018 wells averaged 7,500 ft and 2023 wells 10,000 ft, some of the "improvement" is just more rock, not a better well. Looking at production per lateral foot separates design gains from geometry.
- Time on production. Younger cohorts have fewer producing months. Comparing a 2024 cohort's cumulative against a 2019 cohort's cumulative is meaningless; comparing both at month 6 is fair. That's why the normalized milestone (6/12/24 months) framing matters.
- Survivorship. Wells that were shut in, plugged, or never delivered still belong in the cohort. Dropping them flatters the average. The full record — from first permit to last production — keeps the dead weight in the count where it belongs.
From cohort table to a decision
The value shows up when you attach the vintage read to a real choice.
A&D underwriting. You're evaluating an operator's book in a county. Pull their wells by vintage and compare the recent cohorts to the county-wide vintage curve. If their 2022–2023 wells sit above the county's same-year cohorts, you're paying for genuine execution edge. If they track the county exactly, you're paying for acreage, not skill — and the price should reflect that.
Program planning. An operator setting a 2025 AFE needs a defensible type curve. The most recent stable vintage — not a ten-year blend — is the right anchor, adjusted for the lateral length you actually plan to drill.
Reserves sanity check. If your booked type curve implies every new well behaves like the best vintage ever recorded, the cohort table is the reality check. It shows what the field has actually delivered, year after year, at each month milestone.
Ask it, then drill in
The point of running this conversationally is that the first answer is never the last. See a plateau and you can immediately follow up — split the cohorts by operator, by lease, by lateral bucket — to find who's still improving when the county as a whole has stopped. The record holds every well's permit, wellbore, and production history; the vintage cut is just one way of asking it what's changed over time.
Wells getting better isn't a belief. It's a table you can pull in a minute and defend in a room.