A well that produces 8,000 bbl in its first 90 days sounds fine in isolation. But if the four nearest offsets averaged 14,000 bbl over the same window, that well is telling you something — a completion problem, a landing-zone miss, a mechanical issue, or a reservoir that thins right where you drilled. The number only means something next to its neighbors.
That's the question operators actually ask when they open the record: not "what did this well make," but "what should it have made, and why didn't it?" Answering it well requires three things — a clean production history, a defensible peer group, and a way to normalize the comparison. Here's how that plays out conversationally against the Wellsite data lake.
Start with the peer group, not the well
The fastest way to get a misleading answer is to compare a well against the county average and stop there. A county spans multiple benches, vintages, and completion designs. The average lumps a 2015 gas well and a 2023 oil well into the same number.
A better peer group is the set of true offsets — wells in the same interval, within a defined radius, completed in a comparable window. You can ask for exactly that:
"Show me the 90-day cumulative oil for my well and every well within two miles completed in the last three years."
Now you have a distribution instead of a single benchmark. The platform pulls the wellbores, the completion timing, and the production history, and lines them up side by side. The subject well either sits inside the pack or falls out of it — and falling out is the signal worth chasing.
Normalize before you judge
Raw cumulative production isn't a fair comparison when lateral lengths differ. A 7,500-ft lateral and a 10,000-ft lateral should not be held to the same total. The honest read is production per foot, or a rate normalized to time on production.
You can frame the ask directly:
"Compare my well's first-6-month oil per lateral foot against its offsets, and flag anything more than one standard deviation below the group."
That one-standard-deviation cut is where outlier detection earns its keep. Instead of eyeballing a table, you get a short list of wells that are genuinely off-trend for their neighborhood — the candidates for a workover review, an artificial-lift change, or a hard look at the original completion.
Read the decline, not just the peak
A well can post a strong IP and still disappoint. Two wells with identical 30-day peaks can diverge hard by month twelve if one declines at 55% a year and the other at 75%. The peak flatters; the decline curve tells the truth about reserves.
So the benchmarking question extends past cumulatives:
"How does my well's decline rate compare to the offset average over the first 18 months?"
When a well's early-time rate is normal but its decline is steeper than its offsets, that points somewhere different than a low IP does — often toward completion or frac-conductivity issues rather than a rock-quality miss. Separating those two failure modes is the whole point of the exercise, and the shape of the curve is what separates them.
Compare against your own book, too
Offsets answer "is this normal for the area." Your own book answers "is this normal for how I operate." Both matter. If a well underperforms its offsets but matches your operated average across the same interval, the gap may be a design choice you make consistently — worth revisiting, but not a surprise. If it underperforms both the offsets and your own book, it's an anomaly inside your own operation.
"Benchmark this well against my operated wells in the same formation, then against all offsets regardless of operator."
Two reference frames, one well. The place where they disagree is usually where the insight is.
Turn the one-off into a standing check
The manual version of this — pulling offset lists, normalizing by lateral, charting declines — is an afternoon per well. Done across a whole book, it never gets done. The better pattern is to let it run continuously: an alert that flags any operated well drifting a set threshold below its offset group, or a decline that steepens beyond its recent trend.
"Alert me when any of my wells falls more than 20% below its offset average on trailing-90-day production."
That converts benchmarking from a one-time diligence task into a monitoring layer. New wells get scored against their neighbors as production comes in; existing wells get watched for the moment they slip out of the pack.
The point
A production number in isolation is trivia. The same number against a defensible peer group, normalized for lateral length, and read alongside its decline curve is a diagnosis. The record already holds the offsets, the wellbores, the completion timing, and the full production history — benchmarking is just asking it the right comparative question, and then asking it to keep asking on your behalf.