30-Session Programme · 25 Minutes Per Day · Free · For leaders in Nigeria’s oil and gas sector
WEEK 3 OF 6 · INTELLIGENCE IN THE FIELD · SESSIONS 11–15 · WORK AT YOUR OWN PACE
Part Three · Intelligence in the FieldSESSION 11
Subsurface Maturity, Reserves and Credible Numbers
Why This Matters
Last session placed the subsurface on the Value-versus-Risk Map: accelerate the expert, never replace the judgement. Today is about what earns that acceleration the right to touch a capital decision. The book is plain that reserves are the basis on which capital is raised and partnerships struck, and that uncertainty reported deterministically has repeatedly led to overconfident bookings and eroded credibility. Every company in this sector reports technical numbers someone else relies on. I want you to leave able to say which maturity stage your technical analysis has honestly earned, and whether the numbers your company reports are a range you can defend or a single figure you hope holds.
25-Minute Module
0:00 – 6:00
Earn each stage: the subsurface maturity ladder
- The book names a familiar temptation: jumping to an AI-optimised field development plan before the foundations exist. The result is an impressive model nobody trusts. The disciplined path climbs a maturity ladder, earning each stage before the next.
Stage What it does What it stands on 1. Descriptive Automated interpretation that lets the organisation see the whole data volume reliably Validated data 2. Predictive Property and behaviour modelling that infers the reservoir from sparse data with quantified uncertainty A trustworthy interpretation 3. Simulation Faster, more thorough history matching and scenario testing that honestly characterises the range of outcomes A trustworthy reservoir model 4. Prescriptive Development and recovery decisions ranked by risk-adjusted value The integrity of every stage below - THE DOCTRINE
A development optimisation built on an untrustworthy reservoir model, built on a poor interpretation, built on unvalidated data, is a tower on sand. Phase the ambition so each capability stands on a foundation strong enough to bear it.
- The chapter's own words: subsurface AI, like every capability in the book, must be built in sequence. That is the Sequencing Discipline below the wellhead. The point, the book says, is not to avoid ambition but to phase it.
- No subsurface work? Apply the ladder to your highest-consequence technical judgement, or to the client subsurface decisions that drive demand for your work.
Ask your technical lead which stage they would defend in front of the board, not which stage the vendor's brochure describes.
6:00 – 12:00
The data estate, the chain of trust and the six failure modes
- The subsurface data estate is notoriously fragmented: seismic of different vintages, well logs of varying quality, cores, pressure data, production history and decades of interpretation in different systems. A model built on it without discipline does not fail loudly. It produces plausible, wrong answers.
- The answer is scope, not perfection: build the foundation around the highest-value decisions. A fully integrated data lake before any value is created is, the book says, a multi-year trap.
- On mature fields the first work is data archaeology: reconciling decades-old interpretations and drifted models with the field, work AI can accelerate. The deeper shift is from a static model, used until obviously wrong, to a living subsurface model that updates as production and surveillance data arrive.
- The architecture is a chain of trust from seismic to development decision: data validated, models versioned, uncertainty propagated, output connected to the decision.
Failure mode the book names The discipline that prevents it The black box, ignored by experts, correctly Keep the expert in the loop; explainable methods Ignoring uncertainty Quantify it honestly Building on unvalidated data Do the data work first Disconnecting the model from the decision Connect every model to a decision and an owner Replacing rather than amplifying expertise Make experts faster and fuller Skipping the maturity sequence Respect the ladder
A model nobody owns is a model nobody maintains as it drifts from production reality.
12:00 – 19:00
Reserves, capital and the credibility of the numbers
- Reserves rest, ultimately, on a subsurface understanding and its uncertainty. AI changes what is possible: by quantifying that uncertainty rigorously and probabilistically, it allows reserves to be stated with honest ranges and explicit confidence rather than single optimistic figures.
- The book calls that honesty a financing advantage: the operator that defends its reserves with quantified probability earns trust, and better terms, that deterministic optimism does not. Where, as the book notes citing Reuters reporting on NNPC, mobilising investment at scale is central to Nigeria's production ambition, credible subsurface numbers are part of the financing case.
- THE DOCTRINE
A reservoir model that informs a development decision or a reserves booking must be as governed as any financial model, because its consequences are as large.
- Governance means a model inventory, an owner and quality control for each, versioning so a decision traces to its model and data, and uncertainty conventions. Subsurface data is commercially sensitive; guard it against leakage through uncontrolled tools.
- Know the four limits the book names: data too scarce for any method; non-uniqueness, which AI characterises but never eliminates; explainability, since a model experts cannot interrogate should not carry a hundred-million-dollar decision; and the temptation to automate judgement that should remain human.
- IN THE FIELD · THE WORKED CASE
- Field. A fictional mature-field operator's living subsurface model flags one reservoir block producing differently from expectation. Re-history-matching widens the range of plausible futures.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its capital committee holds a single-figure forecast in the approved plan. The disciplined move is to revisit the decision against the range rather than defend a forecast the field has already contradicted. Same loop, both ends.
Ask for the range before you ask for the answer.
19:00 – 25:00
Before you close the laptop
- Measure against a baseline: interpretation speed and coverage, model fidelity, rigour of uncertainty quantification, barrels added from bypassed zones, forecast accuracy against subsequent production, and cycle time from question to capital decision. Models built are activity, not value.
- The book's starting moves: one mature, high-potential field and its recovery-factor question; the small trusted data foundation that decision needs; experts at the centre; a baseline and a measured outcome.
- THE DOCTRINE
No confident interpretation acts on a well without an expert's judgement behind it. Let the machine accelerate the expert rather than replace the thinking.
- Write down the stage your company has honestly earned, the evidence for it, and the one number you report externally that rests on that stage.
- That is Section 11 of your plan: our maturity stage and the credibility of the numbers we report.
The cheapest credibility you will ever buy is saying the range out loud before the field says it for you.
Key Terms
Subsurface AI maturity ladder
The book's four stages, descriptive, predictive, simulation and prescriptive, each earned before the next because each depends on the integrity of the one below.
Data archaeology
Recovering and reconciling old seismic, well and production data into a trustworthy current picture of an asset before modelling it.
Living subsurface model
A reservoir model that updates continuously as production and surveillance data arrive, flagging divergence before value is lost.
Deterministic optimism
Reporting a single optimistic figure where the underlying uncertainty is large. The book contrasts it with reserves defended by quantified probability.
Self-Check — Before You Submit
1. Your team proposes an AI-optimised field development plan this year. The interpretation is decades old and the data unvalidated. What do you decide?
- A. Approve it; the prescriptive stage carries the most value
- B. Phase it: validate the data one high-value decision needs and earn the descriptive stage first
- C. Reject subsurface AI until the whole data estate is integrated
- D. Approve it if the vendor guarantees accuracy
B. Phase it: validate the data one high-value decision needs and earn the descriptive stage first — The book's ladder requires each stage to stand on the one below, so phasing is right. Waiting for a fully integrated data estate is the tempting opposite error; the book calls that a multi-year trap.
2. Which of these is an outcome measure for subsurface AI in the book's terms?
- A. Number of models built
- B. Volume of seismic interpreted
- C. Forecast accuracy against subsequent production
- D. Number of geoscientists trained on the tool
C. Forecast accuracy against subsequent production — Models built and volumes interpreted are activity. Whether the forecast held against later production is an outcome the board can rely on.
3. A director asks why reserves should be reported as a range with explicit confidence. What is the book's answer?
- A. Regulators prohibit single figures
- B. AI can now eliminate subsurface uncertainty
- C. Ranges are easier to defend in a dispute
- D. Quantified probability earns trust, and better terms, from capital markets that scrutinise technical claims
D. Quantified probability earns trust, and better terms, from capital markets that scrutinise technical claims — The book frames honest ranges as a competitive and financing advantage. The tempting answer that AI eliminates uncertainty is wrong: the book says non-uniqueness can be characterised better but never removed.
Today’s Assignment
Write Section 11 of your plan: the maturity stage your company's technical-AI capability has honestly earned, and how that stage affects the credibility of the numbers your company reports.
- Name the technical judgement in your company with the most capital consequence: for an operator, likely a reservoir model, reserves figure or recovery decision. With no subsurface work, use the client subsurface decision that drives demand for your work, or your highest-consequence technical forecast.
- Place it on the four stages with evidence and the gap to the next. If you use no AI yet, place the current manual workflow and state what the descriptive stage would require.
- List the numbers your company reports to lenders, partners, regulators or the board that depend on it, and say whether each is a single figure or a range with stated confidence.
- Name the failure mode your company is closest to and the owner, by role, who would stop it.
- Give baselines for three of the book's subsurface measures, or state that none exist.
- The honest part: where does your company report deterministic optimism today, and what would saying the range cost? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 11 — Our maturity stage and the credibility of the numbers we report.
How this is marked — the rubric
| Places the capability on the four stages with evidence and names the gap to the next | 30 |
| Identifies the externally reported numbers that depend on it and states single figure or range for each | 25 |
| Names the nearest failure mode and an accountable owner by role, not a department | 25 |
| Sets three outcome measures with baselines, or states honestly that none exist | 20 |
| Total · pass mark 70 | 100 |
Submitting session 11
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 11, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.Capabilities Covered
Subsurface maturity sequencingReserves credibilityModel governanceHonest uncertaintyExpert-in-the-loop judgement
Part Three · Intelligence in the FieldSESSION 12
Digital Twins for Mature Fields
Why This Matters
Few phrases in our industry are sold harder than digital twin, and few are misunderstood more. The book draws the line precisely: a twin is not a 3D picture, a static engineering model or a dashboard. It is a virtual representation kept connected to the asset by live or regularly refreshed data, and trusted enough to support real decisions. Its purpose is to move your company from asking what happened last month to asking what is changing now. Today you decide whether one of your assets or operations warrants a twin at all, what it must answer, and how a 90-day pilot would prove it. No is an acceptable answer if you can defend it.
25-Minute Module
0:00 – 5:00
What a twin is, and what it is not
- Plain idea first. A good station supervisor carries a mental picture of what normal looks like. A twin is that picture made explicit, kept current by data and shared.
- The book's definition: a virtual representation of a physical asset, kept connected to it by live or regularly refreshed data, and trusted enough to support real operating decisions. Its defining requirement is sustained, validated correspondence with the asset.
- THE DOCTRINE
A model that drifted away from field reality two years ago is not a twin. It is a historical document. The loop closes only when the twin's intelligence returns to the field as an owned, tracked action.
- This is the Operating-Intelligence Loop with an asset attached: sense through synchronised data, decide through the model, act through an owner, learn by checking the model against the field.
When anyone shows you a twin, ask when it was last checked against the field, and by whom.
5:00 – 11:00
Six methods built for imperfect brownfield data
- The book says these six methods are specifically suited to environments where instrumentation is uneven and historical data is messy, and that each solves a specific problem mature Nigerian assets actually have.
Method The problem it solves on a mature asset Sensor fusion Combines SCADA tags, well tests, manual readings, inspection reports and field notes into one estimate of current state, flagging drifted instruments Physics-informed machine learning Embeds engineering behaviour so models need less data and extrapolate more safely where failure labels are scarce Model-based surveillance Compares what the field does with what the model says it should; management by exception Simulation and scenario testing Tests a choke change, compressor outage or workover sequence on the twin before the field Anomaly detection Flags departures from the normal envelope before a fixed alarm trips Asset-behaviour forecasting Projects decline, water cut, degradation and the constraint that binds next quarter - Fusion lets a credible twin be built from imperfect data that exists, rather than waiting for a perfect sensor network that may never be funded. Simulation is the most direct payoff: testing interventions on the twin before the field is, the book says, one of the cheapest forms of risk management available.
If nobody can tell you which instruments the twin currently distrusts, it is trusting all of them.
11:00 – 17:00
Where a twin earns its keep, and the ladder it climbs
- The book walks eight asset classes. Mature fields: twin economics are most direct. Flow stations: often decades old and partly instrumented, so data must be hybrid. Pipelines: detection time is loss; NUPRC has reported crude oil losses falling to a 16-year low, which the book reads as evidence that focused surveillance pays. FPSOs: every offshore mobilisation costs far more.
- Gas facilities, where flaring must be modelled and documented as the industry moves toward commitments aligned with the World Bank's Zero Routine Flaring by 2030 initiative. Terminals, where evidence beats assertion in reconciliation. Well pads, surveilled by exception.
- Brownfield infrastructure is the hardest and often the most valuable start, because documentation rarely matches the asset. Begin descriptively: laser scanning, document recovery, P&ID reconstruction and field verification, before any prediction.
- The twin's ladder: descriptive first, a trusted record of what actually exists, then diagnostic, predictive and prescriptive, each depending on the stage below.
- THE DOCTRINE
Refuse the vendor's pretty model. A twin that does not connect six departments into one causal chain is a screensaver. Buy living intelligence, or buy nothing at all.
- IN THE FIELD · THE WORKED CASE
- Field. A fictional indigenous operator's partly instrumented flow station fuses instrument data with structured operator readings; its twin flags a separator departing from its modelled norm.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its executive committee puts the book's test: why did we lose that production, and which handoffs allowed it? A twin that cannot trace that across functions is a visualisation.
Start on the asset where the pain is visible, not where the data is prettiest.
17:00 – 21:00
Measuring a twin the board can trust
- Seven outcome indicators, each against a baseline: model accuracy, asset visibility, production constraints identified, shutdown planning accuracy, maintenance prioritisation quality, capital allocation confidence and operator response time, plus adoption and learning velocity.
- Model accuracy is the foundation. An untracked twin quietly drifts, and the book warns a drifted twin is more dangerous than no twin at all.
- Constraints identified matters only with what was removed and recovered. Shutdown planning accuracy compares planned with actual scope, duration and cost; fewer discovered-on-opening surprises is one of the most bankable forms of twin value.
- Capital allocation confidence asks how many major decisions were scenario-tested on the twin and how predictions compared with outcomes. Operator response time, signal to validated field action, tests the chain of trust you traced in Session 7.
Publish the twin's accuracy openly; a number nobody sees is a number nobody defends.
21:00 – 25:00
Before you close the laptop
- The pilot: one mature field, flow station, pipeline segment or facility, with visible pain, adequate data, an engaged field team and a personally invested sponsor. Discovery and baseline, build and test, adoption; then continue, expand, adjust or stop.
- Design the workflow before any screen: who receives each output, who validates it, who owns the action, the escalation path, the response time and the meeting that reviews it.
- THE DOCTRINE
When the dashboard is designed first, the organisation usually gets a beautiful model that nobody is accountable for acting on.
- Write down the asset, the question its twin must answer, and whether one is warranted.
- That is Section 12 of your plan: whether a living twin is warranted, what it must answer, and its pilot.
A twin earns its name every week it stays true to the field, and loses it the week nobody checks.
Key Terms
Digital twin
A virtual representation of a physical asset, kept connected to it by live or regularly refreshed data and trusted enough to support operating decisions. Not a 3D visualisation, static model or dashboard.
Sensor fusion
Combining sources of different quality into one estimate of an asset's current state, weighting each by reliability and flagging drifted instruments.
Model-based surveillance
Continuously comparing what the field does with what the model says it should, so attention goes only to assets behaving unexpectedly.
Shutdown planning accuracy
Planned versus actual shutdown scope, duration and cost; a test of whether twin-informed condition assessment improves planning.
Self-Check — Before You Submit
1. A vendor shows a detailed 3D model of your flow station built from drawings several years old, not connected to any refreshed data. What do you decide?
- A. Buy it as the descriptive stage of the ladder
- B. Buy it and connect live data later
- C. Decline to treat it as a twin until it has validated, refreshed correspondence with the asset and a stated question to answer
- D. Buy it for the board pack
C. Decline to treat it as a twin until it has validated, refreshed correspondence with the asset and a stated question to answer — A twin requires sustained, validated correspondence with the asset; this is a historical document. Calling it descriptive is the trap: that stage is a field-verified record of what exists.
2. On a brownfield facility with outdated drawings and unrecorded modifications, where does the twin journey begin?
- A. Reconstructing a record of what actually exists, before any prediction
- B. Predictive failure models, because they pay fastest
- C. A full real-time sensor network
- D. Prescriptive optimisation of the operating envelope
A. Reconstructing a record of what actually exists, before any prediction — Brownfield twins begin descriptively, reconstructing what exists before predicting. Waiting for a full sensor network is the tempting error; the book's approach is hybrid.
3. Which of the book's twin measures most directly shows whether the chain of trust from signal to action is working?
- A. Asset visibility
- B. Number of twin users trained
- C. Model accuracy
- D. Operator response time
D. Operator response time — Signal to validated field action is, in the book's words, the most direct measure of the chain of trust. Model accuracy is foundational but says nothing about whether anyone acted.
Today’s Assignment
Write Section 12 of your plan: whether a living twin is warranted for one of your company's assets or operations, the question it must answer, and its 90-day pilot.
- Choose one asset or operation from the book's classes, or your nearest equivalent: your own critical equipment, a vessel, a depot or a client asset you maintain. Say why it carries your company's one loss.
- State the question the twin must answer, in barrels, downtime hours, cost or risk, and test it against the book's: why did we lose that production, and which handoffs allowed it.
- Decide: warranted or not. If not, say what you would do instead. If warranted, name the starting rung on the ladder and the data archaeology the descriptive stage needs.
- Name the first two of the six methods you would use and the data each draws on.
- Write the 90-day pilot: scope, baseline, three of the seven board measures, the workflow by role and your decision rule at day 90.
- The honest part: which model, drawing set or asset record in your company is a historical document that people still treat as current? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 12 — Whether a living twin is warranted, what it must answer, and its pilot.
How this is marked — the rubric
| Makes and defends a warranted-or-not decision for one named asset or operation | 25 |
| States the question the twin must answer in measurable terms and the starting maturity rung | 25 |
| Designs a 90-day pilot with baseline, three board measures and a workflow owned by role | 30 |
| Names one honest example of a drifted model or record still treated as current | 20 |
| Total · pass mark 70 | 100 |
Submitting session 12
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 12, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.Capabilities Covered
Twin versus visualisationBrownfield data fusionScenario testing before capitalTwin maturity ladderTwin outcome metrics
Part Three · Intelligence in the FieldSESSION 13
Marginal Field Economics
Why This Matters
The book calls marginal fields the clearest proof that advantage is shifting from who holds the asset to who can understand it. NUPRC's bid round placed 57 marginal oilfields with indigenous operators, under the Commission's dedicated guidelines. Those assets run on thin margins, where every avoidable loss cuts into survival economics. If you are not a marginal operator, stay with me: every business has an operation where margin is thin and capital scarce, and the discipline transfers. Today you rank the avoidable losses in that operation by risk-adjusted impact, so capital follows evidence rather than the loudest voice in the room.
25-Minute Module
0:00 – 5:00
Capital, not intelligence, is the constraint
- The book's diagnosis: the smaller operator's structural disadvantage is capital, not intelligence. Marginal fields are typically smaller discoveries larger operators left undeveloped because the economics did not clear their thresholds, so they reward operators who run them smarter rather than bigger.
- The book's anatomy of a thin margin, labelled illustrative and not to scale, shows the costs standing between revenue and surviving margin leave little room for avoidable loss. That is why operating intelligence is worth more per barrel on a marginal field than anywhere else.
- THE DOCTRINE
For a marginal operator, time to first oil and the depth of the pre-revenue cash trough are financing decisions, not just engineering outcomes.
- A plain picture: a household on a tight budget is rarely ruined by one large bill, but by many small leaks it has stopped noticing. The book asks you to pick the five small losses your people have quietly normalised, and take them back.
The losses that cost a thin-margin business most are the ones nobody bothers to report any more.
5:00 – 12:00
Six tools: capital discipline first, technology second
- The book is explicit that, for a marginal operator, each of these methods is a capital-discipline tool first and a technology second.
Method What it does for a thin-margin business Production forecasting A range with explicit confidence, built on the field's actual behaviour; it underpins lender covenants, partner cash calls, evacuation commitments and liquidity Economic scenario modelling A living model of field economics re-run in minutes against price, foreign exchange, tariffs, downtime and cost inflation Portfolio prioritisation Ranks every intervention by risk-adjusted value per dollar of scarce capital Workover optimisation Estimates each candidate's success probability and uplift from production signatures and past outcomes Capital sequencing Orders spend under rig, long-lead, logistics, funding and evacuation limits to keep the cash trough shallow Risk-adjusted decision support States each decision as an expected value with assumptions and ranges, so it is auditable - THE DOCTRINE
An evidence-based, repeatable, auditable ranking turns capital allocation from an argument into an analysis.
- Workovers are the marginal operator's most common growth lever and the costliest place to be wrong. The book's arithmetic is simple: avoiding two poor candidates can fund a third good one.
- Sequence matters almost as much as amount: the same capital, sequenced differently, can produce a materially different survival profile. The book's conceptual two-path chart shows earlier first oil and fewer overruns keeping the cash trough survivable and payback earlier.
The book's argument is that a smaller operator with a probabilistic forecast and a credible track record borrows cheaper.
12:00 – 19:00
Where intelligence substitutes for scale
- The book's illustrative portfolio chart plots interventions by risk-adjusted impact against capital intensity, separating quick wins from capital traps. The book's alternative to that ranking is allocation by the loudest voice in the room or the most recent crisis.
- Early production facilities are often the company's entire cash engine: when the EPF is down, revenue is zero. Even simple, low-cost telemetry with anomaly detection, designed in from day one, protects uptime on the single asset the business cannot afford to lose.
- Evacuation is often the most fragile link: trucking, barges, third-party pipelines and shared infrastructure, each with tariffs, losses, security exposure and counterparty risk. Reconciling volumes shipped against volumes received, by route, converts chronic leakage and dispute into a managed, costed decision.
- Small flow stations: a modest intelligence layer gives a two-person shift the situational awareness of a fully staffed control room. Mature wells: per-well gains are individually small, but on a marginal asset they aggregate into the difference between covering costs and not.
- Instrument for decisions, not aspirations. Ask which measurements buy the most decision quality per dollar. The book holds that low-cost automation chosen this way gives a marginal operator eighty percent of the visibility at a fraction of the big-operator price.
- IN THE FIELD · THE WORKED CASE
- Field. A fictional marginal field operator evacuating by barge finds a persistent gap between volumes shipped and received on one route, and its EPF has tripped repeatedly this quarter.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its board ranks both losses against a new drilling idea by risk-adjusted impact and capital intensity. The reconciliation fix is a quick win; EPF protection guards the only cash engine. The ranking sets the order, not the proposer's seniority.
Declining a weak candidate before it takes a rig slot is as valuable as choosing a strong one.
19:00 – 25:00
Before you close the laptop
- Measure in cash. Seven operating KPIs, each against a baseline, feed the metric that funds the future: cash operating cost per barrel, time to first oil, workover success rate, capex efficiency, production uptime, deferment reduction and payback period, all rolling into field-level free cash flow.
- Pilot where the cash impact is most direct: often the workover programme, the EPF or the evacuation chain. A small organisation's decision paths are short, so a well-designed workflow can be running in weeks, not quarters.
- THE DOCTRINE
You will never out-spend a major, so stop trying. Your only path is to out-think them: make a lean team see more, decide faster and waste less.
- That is Section 13 of your plan: our thin-margin discipline, the avoidable losses ranked by risk-adjusted impact.
- Pop Quiz 3 opens now: five questions, closed book, on subsurface maturity and credible numbers, digital twins and thin margins, covering Sessions 11 to 13. Sit it before Session 14.
Intelligence is how the underdog wins, and it only wins when it is aimed at cash.
Key Terms
Marginal field
In the book's terms, typically a smaller discovery that larger operators left undeveloped because its economics did not clear their thresholds. NUPRC's bid round placed 57 with indigenous operators.
Cash trough
The depth of cumulative negative cash before a development reaches revenue. Capital sequencing aims to keep it shallow enough to survive.
Risk-adjusted value per dollar
The basis of portfolio prioritisation: each candidate intervention ranked by its expected value, adjusted for risk, against the scarce capital it consumes.
Field-level free cash flow
The metric that integrates cost discipline, speed to revenue, intervention quality, capital efficiency, uptime and deferment control, and funds the next well.
Self-Check — Before You Submit
1. A fictional marginal operator has capital for only part of its list: a drilling idea backed by the chief executive, an evacuation reconciliation fix and low-cost EPF telemetry. What does the book's discipline require?
- A. Fund the chief executive's proposal
- B. Rank every candidate by risk-adjusted value per dollar of scarce capital, stress-tested against downside scenarios, and fund in that order
- C. Split the capital equally across functions
- D. Defer all three until the field is fully instrumented
B. Rank every candidate by risk-adjusted value per dollar of scarce capital, stress-tested against downside scenarios, and fund in that order — Portfolio prioritisation turns allocation from an argument into an analysis. Deferring until full instrumentation is the tempting error; the book prescribes fit-for-purpose, low-cost measurement on the points that drive decisions.
2. Why does the book say a forecast error can be existential for a marginal operator but only a nuisance for a large one?
- A. Marginal operators face stricter reporting rules
- B. Large operators do not use forecasts for planning
- C. Marginal fields have less reliable reservoirs
- D. Its forecasts underpin lender covenants, partner cash calls, evacuation commitments and its own liquidity planning
D. Its forecasts underpin lender covenants, partner cash calls, evacuation commitments and its own liquidity planning — Thin margins and limited capital mean the forecast carries financing weight. The book's remedy is a range with explicit confidence, not a single optimistic line.
3. Which metric does the book place at the top of the marginal operator's scoreboard, served by all the others?
- A. Field-level free cash flow
- B. Time to first oil
- C. Cash operating cost per barrel
- D. Number of workovers completed
A. Field-level free cash flow — Field-level free cash flow is the metric the seven operating KPIs feed. Cash operating cost per barrel is tempting because the book calls it the survival metric, but it is one input, not the summit.
Today’s Assignment
Write Section 13 of your plan: the thin-margin discipline for your company, with its avoidable losses ranked by risk-adjusted impact.
- If your company operates marginal fields, work on one field. If not, choose your business's most margin-sensitive operation, such as a contract, a vessel, a yard line or a depot, and apply the same discipline.
- List five avoidable losses your people have normalised. For each, give the evidence and your best estimate in barrels, naira, dollars, days or hours. Where you have no number, say so and name the role that would produce it.
- Rank them by risk-adjusted impact against capital intensity, mark which are quick wins and which are capital traps, and give the reason for each.
- Name the single asset your revenue cannot survive losing, your equivalent of the EPF, and how you would protect its uptime cheaply.
- Choose three of the book's scoreboard measures, give baselines, and show how each feeds field-level free cash flow or your operation's free cash flow.
- The honest part: where does capital in your company currently follow the loudest voice or the latest crisis? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 13 — Our thin-margin discipline: the avoidable losses ranked by risk-adjusted impact.
How this is marked — the rubric
| Lists five normalised avoidable losses with evidence and an estimate or a named owner by role for each | 30 |
| Ranks them by risk-adjusted impact against capital intensity, separating quick wins from capital traps with reasons | 30 |
| Names the single revenue-critical asset and a low-cost protection for it | 20 |
| Sets three cash measures with baselines linked to free cash flow | 20 |
| Total · pass mark 70 | 100 |
Submitting session 13
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 13, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.ASSESSMENT NOW OPEN
Pop Quiz 3 — Subsurface, Twins and Thin Margins opens now. Sit it before you start the next session — it is how you find out whether the doctrine landed, and the practice papers may be resat as often as you like.
Sit Pop Quiz 3 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Thin-margin disciplinePortfolio prioritisationWorkover candidate selectionCapital sequencingCash-based scoreboard
Part Three · Intelligence in the FieldSESSION 14
Production Optimisation
Why This Matters
The book puts it in one line: predictive maintenance protects barrels; optimisation recovers them. This is the loop closing on revenue. The prize is the gap between natural decline and what the field could deliver with earlier, better-sequenced intervention, and on a mature asset base that gap is being lost quietly every day. Today I want you to hold the single idea that separates optimisation from activity: every production system has a binding constraint, and capacity added anywhere else adds cost, not barrels. Whatever your business runs, wells, a fabrication line, a vessel schedule or a depot, one stage sets the pace of the whole. You will name it, and name the one most likely to bind next.
25-Minute Module
0:00 – 6:00
From a reporting line to a decision loop
- Reactive field management is linear and slow: an event happens, it is logged, compiled, reported, discussed and eventually acted on, often after the operational moment has passed.
- Real-time decision intelligence is circular: sense, analyse, recommend, act, learn. It is the Operating-Intelligence Loop from Session 1 with the decide stage drawn as two steps, analysis and then a ranked recommendation. Its critical property is that it closes: the result of acting feeds back into sensing.
- THE DOCTRINE
An optimisation programme that recommends but never learns from what happened is not a loop. It is a slow straight line wearing a circle's name.
- The book's conceptual picture of the prize is the shaded area between natural decline and optimised production: barrels protected from deferment and recovered through earlier intervention. It is a continuous flow of value, not a one-time gain.
Ask when a recommendation last fed back into the model that made it.
6:00 – 13:00
Optimisation is constraint management
- Plain idea first. Water through a pipe flows at the rate its narrowest section allows; widen any other section and nothing changes. A production system behaves the same way. Well capacity, flowline hydraulics, separation, compression or evacuation: one stage limits everything upstream and downstream of it. That stage is the binding constraint.
- THE DOCTRINE
Adding capacity anywhere except the binding constraint adds cost, not barrels. The central skill is finding the true bottleneck as it moves through the system.
- The constraint moves. Relieve it and another emerges. The book's illustrative picture is explicit that AI's value is identifying the true binding constraint, and the next one once it is relieved. In Nigerian operations, the book adds, the true constraint is often not where attention is focused.
Where it can bind What the book says to look for Producing wells Departures from expected decline; lift and choke settings across every well, not the priority handful Flowlines Hydraulic limits, slugging, deposition and back-pressure; a cleanout, pressure change or routing decision may release trapped production Separators Carryover, level instability or undersized capacity capping everything upstream Compressors A compression limit can cap oil production and force flaring at the same time Surface facilities Facility-level limits that well-by-well thinking misses; balance the whole rather than maximise one part Allocation systems Weak logic or inputs that create dispute; allocation anchored to well tests and traceable from sensor to statement
If two departments each claim to be the bottleneck, at most one of them is right.
13:00 – 19:00
Six methods, one loop
- Decline-curve analytics industrialises the oldest discipline in reservoir engineering: fitting and re-fitting decline models to every well and quantifying departures in barrels, so an underperformer is flagged within days rather than a quarter late. Well-performance anomaly detection learns each well's normal signature and catches a developing lift, tubing or sand problem early enough to plan rather than absorb an outage.
- Constraint-based optimisation models the system as a whole and computes the settings that maximise total throughput within its limits. Real-time forecasting turns the production plan into a living instrument, so nominations and evacuation commitments rest on confidence rather than hope.
- Recommendation engines convert findings into ranked actions, such as adjusting a choke or prioritising a workover, each with an expected impact and a confidence level. Closed-loop decision support tracks each recommendation through validation, ownership, action and outcome, deliberately human-in-the-loop: the system proposes and learns; people validate and execute.
- Workover portfolios are where optimisation meets capital: rank candidates by uplift, success probability and return, sequence them under rig, logistics and budget limits, and decline the weak ones.
- IN THE FIELD · THE WORKED CASE
- Field. In a fictional onshore operator, engineers upgrade artificial lift on three wells. Flow-station throughput does not move, because separation was already the limiting stage.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its board approved the upgrade on a well-level business case. The question it should have asked is the book's: where is the binding constraint, and what binds next once it is relieved? The field saw wells; the board needed to see the system.
A recommendation without a named owner is a suggestion the field is free to ignore.
19:00 – 25:00
Before you close the laptop
- Measure in two halves, each against a baseline. Barrels not lost: barrels deferred, production variance, well uptime, daily production reliability. Barrels and value added: intervention response time, constraint removal rate, forecast accuracy, optimisation gains. Together they roll up to protected and optimised production.
- Constraint removal rate keeps the method honest: not how many constraints were found, but how many were removed and what they returned. Optimisation gains are credited conservatively, with disciplined before-and-after logging.
- THE DOCTRINE
If your production meeting argues about whose number is right, you have a data problem, not a production problem. The meeting should debate choices, not spreadsheets.
- Pilot one field's decline management, one system's constraint or one asset's workover sequence: ninety days, a written charter, a measured baseline and a senior sponsor, ending with production visibly protected or recovered because a decision was made better and faster.
- That is Section 14 of your plan: our binding constraint, the next one, and how AI would find them.
Every week your team argues about data is a week of barrels you chose not to recover.
Key Terms
Binding constraint
The one stage of a production system that limits everything upstream and downstream of it. Capacity added elsewhere adds cost, not throughput.
Closed-loop decision support
Tracking each recommendation through validation, ownership, action and outcome, and feeding the result back so the models learn which recommendations created value.
Constraint removal rate
The rate at which identified production constraints are actually relieved, and the production released as a result.
Production variance
The gap between forecast or potential production and what the field actually delivered, tracked by well, asset and cause.
Self-Check — Before You Submit
1. Your engineers propose an artificial-lift upgrade across a cluster of wells. The flow station they feed has run near separator capacity for months. What do you decide?
- A. Approve it, because more well capacity always means more barrels
- B. Approve half the wells as a trial
- C. Hold it until the binding constraint is confirmed; if separation is binding, relieve that first and identify what binds next
- D. Reject all spending on the flow station until the next shutdown
C. Hold it until the binding constraint is confirmed; if separation is binding, relieve that first and identify what binds next — Capacity added anywhere except the binding constraint adds cost, not barrels. A partial trial is tempting but tests the wrong stage; the throughput still cannot rise past the separator.
2. What is the critical property that makes the book's optimisation cycle a loop rather than a report?
- A. The result of acting feeds back into sensing
- B. It runs on real-time data
- C. It produces a ranked list of recommendations
- D. It is reviewed in the monthly executive meeting
A. The result of acting feeds back into sensing — The loop only creates value when act feeds back into sense. Ranked recommendations are tempting, but recommendation without learning is, in the book's words, reactive management in disguise.
3. Which of these belongs to the book's barrels-and-value-added half of the production scorecard?
- A. Barrels deferred
- B. Well uptime
- C. Daily production reliability
- D. Intervention response time
D. Intervention response time — Intervention response time, constraint removal rate, forecast accuracy and optimisation gains measure value added. The other three options sit in the barrels-not-lost half.
Today’s Assignment
Write Section 14 of your plan: the binding constraint in your company's production system, the next one likely to emerge, and how AI would find them.
- Draw your system as a chain of stages from source to revenue. For an operator: wells, flowlines, separation, compression, evacuation and allocation. For other businesses, use your own chain and state your unit of throughput.
- Name the binding constraint today and the evidence for it. If you cannot evidence it, say so; that is itself a finding.
- Name the constraint most likely to bind once the current one is relieved, and why.
- Say which of the six methods would find each constraint, and the data it needs. For operators, the book's data spine is well tests, production rates, pressures, facility readings and allocation inputs.
- Choose one barrels-not-lost measure and one barrels-and-value-added measure with baselines, and name the owner of the recommendation loop by role.
- The honest part: where has your company added capacity, or spent capital, somewhere other than the binding constraint? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 14 — Our binding constraint, the next one, and how AI would find them.
How this is marked — the rubric
| Maps the system as stages and names the binding constraint with evidence | 30 |
| Names the next likely constraint with a reason | 20 |
| Matches methods and data to finding each constraint | 20 |
| Sets one measure from each half of the scorecard with baselines and names the loop owner by role | 15 |
| Gives one honest example of capital spent away from the binding constraint | 15 |
| Total · pass mark 70 | 100 |
Submitting session 14
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 14, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.Capabilities Covered
Constraint managementClosed-loop optimisationDecline surveillanceRecommendation ownershipTwo-half production scorecard
Part Three · Intelligence in the FieldSESSION 15
Predictive Operations
Why This Matters
The book's charge on maintenance is the sharpest of this week: reactive maintenance is a decision, not a fate, and every emergency repair is a planned one you declined to see coming. Today we place your reliability practice on the book's maturity journey, reactive to preventive to predictive, with the condition-based stage its figure shows in between, and design the next step. Predictive is not a replacement for the earlier stages; it becomes possible only once their disciplines are in place. That matters here, because procurement cycles, import logistics and foreign-exchange pressure can stretch lead times dramatically, so months of warning are worth more in Nigeria than in markets where a part arrives in days. Practice Exam 3 closes the week.
25-Minute Module
0:00 – 5:00
The maintenance maturity journey
- The book's Figure 3, the maintenance maturity journey, shows four stages, each retaining the disciplines of the stage before it. Reactive (run to failure): failures drive the schedule, with the highest downtime and emergency cost. Preventive (calendar-based): fixed intervals, which over-maintain healthy assets and still miss surprises. Condition-based (evidence-based): monitoring data triggers work, and intervention follows evidence. Predictive (AI-enabled): models anticipate failures, so work is planned, spares staged and losses prevented.
- Keep the last two apart. Maintaining on evidence of current condition is the condition-based stage. Predictive goes further: the model anticipates the failure, so the work and the spares are in place before it arrives. The chapter's subtitle shortens the journey to reactive, to preventive, to predictive.
- THE DOCTRINE
Predictive operations is not a replacement for the earlier stages. It is what becomes possible once their disciplines are in place.
- Why the order matters: each stage keeps the disciplines of the one before, and the book is plain that an alert without a responsible owner is operational noise. Maintenance compliance, planned work executed on schedule, is the book's test of whether planned work survives contact with daily emergencies.
- The book names the binding constraint here as not data volume but the distance between field signal, management decision and corrective action.
Ask what share of last month's maintenance hours went to emergencies; that number tells you where you sit.
5:00 – 12:00
Six methods, from the cheapest start to the fullest picture
- Start where the data already is. Work-order analytics needs no new sensors; the data already exists inside the CMMS. It exposes backlog ageing, repeat failures, chronic bad actors, planning quality and schedule compliance. The book's example of what records reveal: the same pump rebuilt four times in two years.
Method What it changes Time-series anomaly detection Flags deviation from each signal's learned envelope long before a fixed alarm trips Remaining useful life Turns maintenance from a calendar exercise into a risk-based discipline; long-lead spares ordered before they are urgent Failure-mode classification Identifies how equipment is failing, so the right crew, parts, procedure and permit go first time Work-order analytics Turns maintenance history from an administrative archive into a management instrument Condition-based monitoring Fuses vibration, thermography, oil analysis, ultrasonic testing and performance curves; maintain on evidence, not fixed intervals Asset-criticality scoring Ranks equipment by failure consequence, updated as condition and production context change - Criticality connects the rest to resources. The book's example: a pump feeding a high-rate well matters more this quarter than the same pump on a shut-in string.
- THE DOCTRINE
The difference between a planned intervention and an emergency response is usually measured in the weeks of warning the organisation had.
Your cheapest predictive model may be a fortnight spent reading your own work orders honestly.
12:00 – 18:00
Where to point it first
- Pumps are usually where predictive maintenance proves itself first: failure modes are well understood, signals are rich, and avoided losses are easy to count.
- Compressors: a trip frequently means immediate production impact and, in many configurations, flaring exposure. For gas-lift-dependent fields the book says compressor reliability is not a maintenance topic but a production topic.
- Generators are a hidden single point of failure where sites make their own power: production, safety systems, communications and metering can be affected at once. Generator reliability rarely reaches an executive dashboard, until the day it becomes the reason a flow station went dark.
- Valves carry regulatory weight: emergency shutdown and surface safety valve integrity is a compliance obligation, not a preference. Cranes are a quiet use case with outsized HSE value. Turbines, wellheads, flow-station equipment and pipeline systems complete the book's list.
- IN THE FIELD · THE WORKED CASE
- Field. A fictional drilling and well services company reads two years of its own work orders and finds one class of pump failing repeatedly between scheduled overhauls, with every failure waiting on an imported spare.
IN THE BOARDROOM · THE PARALLEL- Boardroom. Its board had read MTTR as a workshop problem. The book's point is that in Nigerian operations the logistics and procurement components of MTTR are often larger than the technical repair. The step up is spares planning driven by remaining-life estimates, owned jointly by maintenance and procurement.
Choose the equipment class where the failure data is richest and the loss easiest to count; credibility comes before coverage.
18:00 – 25:00
Before you close the laptop
- Seven board measures, each against a baseline: reduction in unplanned downtime, MTBF, MTTR, maintenance compliance, spares availability, barrels deferred due to equipment failure and production protected by early intervention, plus adoption and learning velocity.
- Read MTBF and MTTR together. Rising MTBF means failures are being prevented; without it, the organisation may simply be getting faster at repairs. Low compliance today is, in the book's words, the statistical promise of failures tomorrow, and a programme with falling compliance is borrowing from its own future.
- Production protected by early intervention is the most demanding measure: disciplined logging of what was detected, what was done and what would likely have happened otherwise, estimated conservatively. Done well, it is a defensible record of losses that did not happen.
- THE DOCTRINE
Reactive maintenance is a decision, not a fate. The cheapest barrel you will produce this year is the failure you prevent.
- That is Section 15 of your plan: where our reliability practice sits and the step up to predictive.
- Practice Exam 3 opens now: fifteen questions, closed book, covering Sessions 11 to 15. It closes Week 3.
Move up the ladder asset by asset, and do not accept that it has always been done this way as an answer.
Key Terms
Remaining useful life (RUL)
An estimate of how much operating time or how many duty cycles remain before equipment can no longer perform its intended function.
Asset-criticality scoring
Ranking equipment by the consequence of its failure across production, safety, environment, repair complexity and cost, updated as condition and production context change.
Mean time between failures (MTBF)
The average operating time equipment runs between failure events. Rising MTBF shows failures are being prevented, not merely repaired faster.
Maintenance compliance
The percentage of planned maintenance executed on schedule. A leading indicator: low compliance today predicts failures tomorrow.
Self-Check — Before You Submit
1. Emergency work dominates your maintenance hours, compliance is low, and work orders rarely record failure modes. Your chief executive wants to buy a predictive platform for all rotating equipment. What do you recommend?
- A. Buy it now, because predictive maintenance replaces the earlier stages
- B. Start with work-order analytics and compliance, then pilot one critical equipment class with a baseline, owner and workflow
- C. Wait until every asset is fully instrumented
- D. Buy it for compressors only, because they are well-sensored
B. Start with work-order analytics and compliance, then pilot one critical equipment class with a baseline, owner and workflow — The book says predictive becomes possible once the earlier disciplines are in place, and calls work-order analytics, which needs no new sensors, often the fastest place to start. The book does call well-sensored compressors strong early candidates, but buying a platform before the value case, baseline and workflow inverts its roadmap.
2. Over a year, MTTR falls sharply while MTBF stays flat. What does that most likely tell the board?
- A. Reliability is structurally improving
- B. Maintenance compliance has risen
- C. The organisation is getting faster at repairs, not preventing failures
- D. Spares availability has fallen
C. The organisation is getting faster at repairs, not preventing failures — The book calls MTBF the cleanest signal of whether reliability is structurally improving or the organisation is simply repairing faster. Reading a falling MTTR alone as structural improvement is the trap.
3. Why does the book say remaining-useful-life estimation is worth more to a Nigerian operator than in many other markets?
- A. Procurement cycles, import logistics and foreign-exchange pressure can stretch lead times dramatically
- B. Nigerian equipment fails more often
- C. Regulators require it for every asset
- D. It removes the need for condition monitoring
A. Procurement cycles, import logistics and foreign-exchange pressure can stretch lead times dramatically — Seeing a failure coming months ahead lets long-lead spares be ordered in time. The option about equipment failing more often makes a claim the book does not make.
Today’s Assignment
Write Section 15 of your plan: where your company's reliability practice sits on the book's maturity journey from reactive through preventive and condition-based to predictive, and the step up to the next stage.
- Choose the one equipment class that matters most to your company's one loss. For operators this is often pumps, compressors or generators; for services, yards and vessels, your own critical rotating or lifting equipment.
- Place your practice honestly as reactive, preventive, condition-based or predictive, with evidence: the share of emergency against planned work, maintenance compliance, and whether work orders record failure modes. If you do not know these numbers, that is the finding.
- Do a desk version of work-order analytics: what would your last two years of records show about repeat failures and backlog on that class?
- Design the step up: the first of the six methods, the data it needs, who receives the alert, who validates it, who owns the action by role, and the spares and procurement change it requires.
- Set three of the book's seven measures with baselines, splitting MTTR into technical and logistics time if you can.
- The honest part: which recurring emergency in your company is, in the book's words, a planned repair you declined to see coming? Aim for 400 to 900 words.
YOUR 90-DAY PLAN
Section 15 — Where our reliability practice sits and the step up to predictive.
How this is marked — the rubric
| Places the practice on the progression with evidence, not assertion | 25 |
| Designs the step up with a first method, its data and a workflow owned by role | 30 |
| Treats spares and procurement lead time as part of reliability, with a named change | 20 |
| Sets three board measures with baselines and names one honest recurring emergency | 25 |
| Total · pass mark 70 | 100 |
Submitting session 15
Write your answer in Word or Google Docs. The first line must read
Write your answer in Word or Google Docs. The first line must read
Session: 15, or the submission is refused. Upload the Word file (.docx), or switch to the Google Doc tab and paste your document’s share link.Marked against the rubric above and returned as a written letter, saved to your progress page. Do not submit until you have passed the self-check above.ASSESSMENT NOW OPEN
Practice Exam 3 — Intelligence in the Field opens now. Sit it before you start the next session — it is how you find out whether the doctrine landed, and the practice papers may be resat as often as you like.
Sit Practice Exam 3 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Reliability maturityWork-order analyticsRemaining-life planningCriticality-led maintenanceReliability board metrics