Wellbore Genius
Help

Fleet-cohort drivers

Across similar stages in your fleet, which inputs most correlate with the outcome? Spearman rank correlation — correlation, not causation. Use this to generate hypotheses, then test with Sensitivity sweep or Design Copilot.

Cohort configuration
24 / 40 stages
Bookmarks

Editing a preset JSON file safely

  • Keep the outer envelope intact — version must stay downhole-fleet-cohort-preset-v1.
  • Only change values inside the preset object. Do not add extra keys; the schema rejects unknown fields.
  • Numeric fields are strings — write "8000", not 8000.
  • wellIds must be an array of strings, e.g. ["BAK-01", "BAK-02"].
  • Dates must be ISO — "2024-06-01". Booleans are true or false (no quotes).

Inline JSON validator

Paste a preset JSON here and validate against the schema before applying it to the form.

Units

Drop wells with fewer matching stages.

Drop stages with screen-out rate ≥ this fraction.

Driver tornado · cohort N=24
clusterSpacing
ρ=-0.71
strong
n=24
fluidIntensity
ρ=-0.41
strong
n=24
stageLength
ρ=-0.23
watch
n=24
proppantIntensity
ρ=+0.22
watch
n=24
Narrative

Cohort N=24. Top 3 drivers of eur (Spearman ρ — correlation, not causation):

• clusterSpacing: ρ=-0.71 (n=24) — lower clusterSpacing correlates with worse eur.

• fluidIntensity: ρ=-0.41 (n=24) — lower fluidIntensity correlates with worse eur.

• stageLength: ρ=-0.23 (n=24) — lower stageLength correlates with worse eur.

Method: Spearman rank correlation with two-sided t-approx p-value. Not a causal model — pair with a single-factor sweep before changing a job plan.

Phase 2 · multivariate + drift + meta-cohort

Holds other factors fixed (ridge β), tracks the strongest driver across vintage years, and pools per-basin ρ via DerSimonian-Laird. Hypothesis generators, not causal claims.

Ridge partial coefficients · top 4 factors

n=24 · R²=0.81 · λ=0.1
clusterSpacing
β=-37.44
strong
p≈0.000
proppantIntensity
β=+21.76
strong
p≈0.000
stageLength
β=-6.87
strong
p≈0.094
fluidIntensity
β=+0.94
strong
p≈0.818

Vintage drift · clusterSpacing vs eur

±2 yr rolling window, ρ vs year
2022: ρ=-0.64 · n=182023: ρ=-0.71 · n=242024: ρ=-0.71 · n=242025: ρ=-0.67 · n=18+10−120222025

Cross-basin meta-cohort · clusterSpacing

k=3 · I²=18%
Pooled ρ̂ = -0.80 · τ²=0.030 · Q=2.45
Basinρnvar(z)
Bakken-0.71240.048
Eagle Ford-0.9380.200
Permian-0.8180.200
Debug
mem://features/fleet-cohort-drivers.md
Cohort N = 24
Outcome: EUR [Mboe] (eur)
Drivers ranked: 4
Skipped: 0
CSV columns (drag to reorder)
Top drivers (|ρ|)
  1. clusterSpacing — ρ=-0.705 · p=0.000 · n=24
  2. fluidIntensity — ρ=-0.410 · p=0.035 · n=24
  3. stageLength — ρ=-0.234 · p=0.258 · n=24

The memory entry is a project-memory record (index + body file). The check mark confirms this route is the one described by that entry — sidebar item, Spearman driver finder, and route path all match.