Upstream, the 10^1,009 problem and the ranked work plan

A typical mid-tier upstream operator runs around 1,800 active wells and 1,500 inactive wells, dispatched by a team of roughly 26 field operators who each visit somewhere between 15 and 22 wells a shift. Every morning, every one of those operators picks an order, picks a sub-set, picks a route. Multiply across the team and the daily plan is one combination drawn from a number with 1,010 digits in it. We are not exaggerating. We worked the math.

The math, plain

With 1,800 active wells and 26 operators, each operator is responsible for about 69 wells in their territory. A typical shift is 22 site visits, roughly 15 minutes per site. The number of ways one operator can choose 22 wells out of 69 and visit them in some order is the permutation P(69, 22), which works out to about 6.6 × 10³⁸. That is 660 trillion trillion trillion possible day-plans for one operator.

Twenty-six operators making independent choices means the team picks one combination out of (6.6 × 10³⁸)²⁶, which works out to a number in the neighborhood of 10^1,009. To put 10^1,009 in context, the observable universe contains roughly 10⁸⁰ atoms. Your team’s daily plan is drawn from a space 10⁹²⁹ times larger.

Why no human shift solves this

A foreman could be the most experienced upstream operator on the planet and the math still wouldn’t fit a human shift. The problem isn’t intelligence. The problem is the size of the search space and the speed at which conditions change. A tank fills overnight. A pump goes down at 4 AM. Commodity prices move two cents at the open. By the time a foreman has built tomorrow’s plan from yesterday’s context, today’s context is different.

This is the same shape of problem the airlines solved with crew scheduling in the 1990s and the same shape Amazon solved with last-mile routing in the 2010s. The right answer is not smarter humans. The right answer is a constraint-based optimizer that ranks tasks by economic impact, applies safety and qualification constraints, and assembles a ranked plan in the few seconds humans can’t. The pumper still drives the truck. The foreman still makes the calls a human has to make. The agent does the math no human shift can do.

What the ranked plan looks like in practice

The work loop runs on three lenses, scored every night and re-scored mid-shift when conditions change.

Cash-flow delta forecasting. For every well, the system extrapolates the last 48 hours of SCADA data into a short-term production forecast and compares it against the well’s Aries decline projection. Any delta in projected oil production multiplied by current commodity pricing gives a dollar-impact score. Every well is ranked by cash-flow delta. The top of the list is where the operator goes first, not because someone said so, but because that is where the dollars are.

Risk-based scoring. Independent of dollar impact, every well also carries a risk score that combines time since last visit, regulatory constraints, alarm history, SCADA signal degradation, and failure probabilities for major components. Risk scoring is not a weight that gets traded off against cash-flow impact; safety is a hard constraint in the optimizer. Crews don’t get dispatched to a site they aren’t qualified for, regardless of how much money is on the table.

Liquid Management Index. Tank levels, hauling schedules, and fluid production rates feed a forward-looking index that flags wells likely to hit critical inventory thresholds before the next visit. The LMI prevents overflows and the costly downtime that comes from missed pumps, without over-servicing low-risk wells the way a fixed weekly route does.

The three lenses combine into a single ranked task list per operator, route-optimized against the operator’s yard location, vehicle capacity, and shift hours, and pushed to their truck-cab tablet by 6 AM. The pumper picks up the tablet at the yard, knows exactly what to work on, in what order, and why, and drives. Mid-shift, when a tank fills faster than expected or a downhole pump fails, the plan re-optimizes without a phone call.

The proof, anonymized

We deployed this work loop with a top 25 private producer running 5,000+ wells across the Western Anadarko, the Permian, and Wyoming. The operator had spent over $10 million on prior SCADA systems, field automation, and pump-by-exception frameworks without a measurable lift in cash flow. They had the data. What they didn’t have was a system that turned data into the day’s plan.

What changed once the work loop was running:

• ▸+15% gross cash flow uplift over a six-month window, without drilling a new well or adding headcount.
• ▸35% fewer miles driven on the same production coverage (deployment figure). Fewer wasted truck-rolls, more time on high-value wells.
• ▸40% lower liquid-hauling inventory across the tank battery, with 60% less redundant tank gauging (deployment figures).
• ▸Safety runs as a hard constraint, not a weight, in the same operation: an unqualified or unsafe dispatch is blocked at the point of assignment rather than flagged afterwards. Safety as a constraint compounds the cash-flow result, not the other way around.

What the operator actually has to do

The deployment we ran with that operator followed a 90-day phased rollout. Three phases, each with a clear deliverable the field could feel.

That is the outline. The first phase is what most operators already have, in pieces, scattered across SCADA, accounting, and CMMS. The integration and reconciliation work (Chapter 4) is what makes phases two and three possible. The security architecture (Chapter 5) is what makes the CISO sign off on phase one in the first place.

One ranked work loop, not three programs

The most common mistake we see when an operator first considers this stack is treating the cash-flow win, the safety win, and the back-office win as three separate programs. They aren’t. They are the same work loop, ranked differently. The same ranked plan that surfaces a deferred-production well at 6 AM also makes sure the crew dispatched to it is qualified, also generates the JSA from the live hazard profile, also closes the work order back to financial tracking when the job is done. One loop, one data model, three outcomes.

Chapter 7 picks up the same loop in midstream, where the primary data sources are GIS topology and pipeline SCADA instead of well-level production, and the work output is hydraulic models and pipeline-integrity dispatch instead of pumper routes. Same architecture. Different surface.