Sweep
Product · Uplift

Barrels earned. Against a counterfactual.

Barrels per intervention against what the well would have done anyway, with a placebo and an interval.

The demoThe numbers
1

Barrels per intervention, with an interval

2

A placebo on shuffled dates

3

Staggered campaigns scored the right way

Hands on

Three steps.

2

The counterfactual, month by month

Now showingOne workover A weak producer worked over.
Simulated field
I-01I-02I-03I-04I-05I-06P-01P-02P-03P-04P-05P-06P-07P-08P-09P-10P-11P-12

Line width: estimated gain. Green: identifiable. Grey: the data cannot tell, so no number.

bbl/day · orange: reported · dashed: the well without the change (physical counterfactual)intervention

So far: 45,605 bbl against the counterfactual.

Sweep says

The workover on P-10 earned 45,605 bbl

Over 12 months against the physical counterfactual (90 %: 17,432 bbl to 76,885 bbl).

The other methods

Synthetic control: 46,028 bbl

Difference-in-differences: 52,333 bbl

Before and after: 71,609 bbl

Causal engine as shipped: 53,198 bbl

3

One clear finding

Finding

The workover on P-10 earned 45,605 bbl. Synthetic control says 46,028 bbl. 57 injector–producer pairs cannot be told apart from these data. Proposal: no injection change on this evidence.

Proposal onlyNever sets a rateA person decides

What it needs

  • Monthly well and injection rates
  • Intervention log with dates
  • Separator and fiscal totals, to reconcile rates first

What it is not

  • Not a reservoir simulator
  • Not a reserves booking

Proven, in simulation

  • Calls 11 % of no-intervention windows significant; before–after 59 %, 2 × 2 difference-in-differences 62 % (simulation, t003)
  • Its 90 % interval covers 86 % of planted uplifts, a little under nominal
  • Injection changes, where no panel method applies: 70 % RMS relative error, against 232 % for the open CRM library

Not proven

  • Against the bar: out of distribution its error is 143 % against 119 % for the open CRM library and 84 % for the open staggered DiD library, worse than both beyond the interval
  • On producer workovers and pump changes the panel methods win: 130 % RMS relative error and +50 % median bias, against 47–48 % and 1–4 % for synthetic control and synthetic difference-in-differences
  • Injector–producer gains: every pair flagged unidentifiable from 36 months; rank agreement with the true gains 0.19
  • No field data; public panels for placebo runs only
  • The full-physics deck not run
  • No reserves booking, ever
Works with

Pair

Who supports whom.

Open →

Tank

The field as tanks and flows.

Open →

Placebo

Effects that are not there.

Open →

Allocate

What moving injection would do.

Open →
Uplift — Barrels earned. Against a counterfactual. · Sweep