15-Session Programme · 25 Minutes Per Day · Free · For leaders in Nigeria’s gas and industrial energy businesses
WEEK 1 OF 3 · LEADERSHIP, GAS ECONOMICS AND PROCESSING · SESSIONS 1–5 · WORK AT YOUR OWN PACE
Orientation · The Promise at the Delivery PointSESSION 1
The Promise at the Delivery Point and the Reader's Toolkit
Why This Matters
A gas business has to connect physical supply with the promise made to a customer, and its records then have to show what was delivered, billed and collected. Artificial intelligence can support those decisions when the business has dependable evidence and an accountable route for acting on it. Over fifteen sessions you will build a signed 90-Day AI Leadership Plan for your own company around one real operating decision you choose today: one with a feasible response, that affects a customer promise and currently costs your company time, money or service. The customer promise remains the test.
25-Minute Module
0:00 – 5:00
The argument, and the finish line
- THE DOCTRINE
An operating forecast does not create gas that is unavailable. A recommendation does not restore a compressor or secure an access route. Management remains responsible for funding the work, agreeing priorities and following the action through.
- The three weeks: 1 Leadership, Gas Economics and Processing; 2 Supply, LNG, Industrial Energy, CNG and LPG; 3 Integrity, Governance and Scale. The finish line: fourteen plan sections, one per session, signed in Session 15.
- House rules: write about your own company; bring a real decision; no value claim without a baseline; no decision without a named owner; report a blocker the same day. The pass mark is 70.
- Said once: this is leadership training, not investment, legal, regulatory or engineering advice. The six frameworks are author frameworks carried forward from The Intelligent Barrel, not regulatory requirements or official technical standards. Technical, commercial and legal acceptance stay with those authorised to give them.
Ask which promise is at risk before you ask what to buy.
5:00 – 10:00
A business defined by interfaces
- The book's fictional industrial customer restarts on Monday. A window is agreed, a compressor is expected back on Sunday, the receiving station has a maintenance restriction and credit was approved against an earlier quantity. Every department can report progress while the combined promise remains impossible.
- The executive question is which constraint must change before delivery can be accepted. A forecast might expose the mismatch; it cannot establish permission, readiness or acceptance.
- Forecasting estimates an uncertain future. Anomaly detection identifies departures requiring investigation. Optimisation compares feasible actions under an objective and constraints. Simulation tests how a proposed action may affect a physical or commercial system. A language model can help retrieve and organise text. Each has different failure modes and acceptance evidence; know which component produced each conclusion.
- Processing, pipelines, LNG, CNG, LPG and industrial supply have problems that are not interchangeable. Apply an example from an activity you do not run to your nearest equivalent decision, without assuming its evidence or controls transfer.
- Nigeria's Petroleum Industry Act, 2021 establishes separate upstream and midstream/downstream regulatory institutions (Petroleum Industry Act, 2021, sections 4 to 8 and 29 to 33). Map an asset's activity and permissions before treating another operation as its regulatory equivalent; legal and technical owners establish the actual obligations.
Count accepted outcomes, not predictions, alerts or documents.
10:00 – 18:00
The six frameworks, as questions
Framework What it asks of a gas decision 1. Operating-Intelligence Loop Where does sense, decide, act, learn stop, and who owns that stage? 2. Data-Foundation Precondition Does the evidence support this decision? 3. Autonomy Ladder Which verbs has the system been delegated? 4. Value-versus-Risk Map What does a wrong recommendation cost, in each direction? 5. Sequencing Discipline Baseline and simplest alternative first: which gate is next? 6. Operations-Excellence Lens Which linked measures represent the mechanism? - The loop. Record separate times for observation, output, decision, execution and outcome acceptance. The delay between prediction and authorised action may be the real bottleneck.
- The data floor and the ladder. Keep measured, calculated, estimated and manually entered values distinct, and ask whether a reviewer can reconstruct the evidence available when the decision was made. On the ladder, use a state model: proposed, checked, approved, released, executed, reconciled and closed.
- THE DOCTRINE
Keep non-negotiable conditions outside the economic ranking. A profitable delivery is not feasible merely because its margin outweighs a scored safety concern.
- Risk, sequence and lens. Weigh each failure direction separately. This use of context is consistent with NIST's voluntary AI Risk Management Framework, which organises risk work through Govern, Map, Measure and Manage (NIST, AI RMF 1.0, 2023, sections 3 and 5). A stopped pilot can be a useful outcome.
- IN THE PHYSICAL SUPPLY · THE WORKED CASE
- Physical supply. At a fictional compression station, a sensor departure becomes a warning. The loop breaks at act if it reaches the maintenance owner too late for a useful response.
AT THE CUSTOMER PROMISE · THE PARALLEL- Customer promise. In a fictional industrial supply contract, the same warning should change the delivery plan or the customer communication. The loop breaks at learn if nobody later records what happened to the affected deliveries. The same loop, seen from both ends.
Find the broken stage before you price the clever one.
18:00 – 22:00
A completed decision specification
- The book's completed specification is illustrative, with assumed scope, not a project approval or engineering recommendation.
Field Illustrative entry Decision Whether to revise tomorrow's industrial delivery allocation Decision time Before the agreed customer confirmation cut-off Evidence and capability Accepted supply, receiving windows, equipment status, contract priorities; forecast plus constrained allocation comparison Existing alternative and boundary Planner's accepted rule-based schedule; one supply point and three named internal customer identifiers Authority Planner proposes; operations checks; commercial owner accepts changes Stop condition Stale supply status, unconfirmed receiving restriction or infeasible allocation Acceptance and value evidence Less accepted shortfall without additional unauthorised exposure or disproportionate review; delivery records, agreed consequences, incremental cost - THE DOCTRINE
Before approving development, ask what information the model could reveal that would actually change the decision. If no feasible action follows the output, redefine the use.
- Such a forecast may still support communication; evaluate it as that, not as additional supply. If no evidence could change your recommendation, the review has become advocacy.
A specification you cannot fill in honestly is itself a finding.
22:00 – 25:00
Write it down, and Section 1
- Before you close the laptop, write one sentence naming your decision and the promise it affects, and one naming where its loop stops.
- I would ask you to choose a decision with a feasible response, not one that sounds like an AI project. The right answer may later be a simpler method or no model at all.
- Section 1 of your plan is Our company, the one decision we will carry, its decision specification and where its loop breaks.
The decision you choose today is the one you will defend in Session 15.
Key Terms
Operating-Intelligence Loop
Sense, decide, act and learn, with separate recorded times for each stage through to outcome acceptance.
Data contract
A short record naming a source's owner, refresh method, acceptable age, unit, reference basis, treatment of missing information and restriction on reuse.
Stop condition
The stated circumstance in which the application must not be used, such as stale supply status or an infeasible allocation.
Decision specification
The Chapter 0 fields for one decision, from decision time to stop condition and value evidence, written before development is approved.
Self-Check — Before You Submit
1. A fictional supplier's forecast shows that Monday's industrial delivery is at risk, but the receiving station's maintenance restriction is unresolved. What can the forecast do, and what can it not establish?
- A. Establish that the delivery can be accepted
- B. Expose the mismatch; it does not establish permission, readiness or acceptance
- C. Authorise automatic reallocation of supply
- D. Show that the commercial team has failed
B. Expose the mismatch; it does not establish permission, readiness or acceptance — A forecast might expose the mismatch, but none of the analytical outputs establishes permission, equipment readiness or customer acceptance; those are resolved in the systems and authority that govern them. Automatic reallocation is tempting but skips that authority.
2. A delivery option has a high margin but carries an unresolved safety condition. How does the Value-versus-Risk Map treat it?
- A. As a weakness scored against the margin
- B. As acceptable if the margin is large enough
- C. As a matter for the vendor
- D. As outside the ranking until the condition is satisfied
D. As outside the ranking until the condition is satisfied — Non-negotiable safety, integrity, permission and authority conditions stay outside the economic ranking and must be satisfied before options are compared. Scoring it against the margin is the trap the book names.
3. Your team proposes a demand forecast, but no feasible action would follow its output before the confirmation cut-off. What should you do?
- A. Redefine the use, or evaluate it as a communication capability
- B. Approve it, since better information always helps
- C. Credit it with the shortfall it predicts
- D. Ask for a more accurate model
A. Redefine the use, or evaluate it as a communication capability — If no feasible action follows the output, redefine the use. It may still support communication, but must not be credited with additional physical supply. A more accurate model does not create a response.
Today’s Assignment
Name your company and where it sits in the gas value chain, choose the one real operating decision you will carry, draft its decision specification and show where its Operating-Intelligence Loop breaks.
- Your company. Its name and where it sits: processing, compression, pipeline transport, LNG, CNG, LPG, industrial gas supply, an industrial energy user, or other. Name the customer promise involved.
- The decision. One real operating decision with a feasible response that affects a customer promise and currently costs your company time, money or service. State what must be decided, by when, the physical and contractual constraints, who can authorise the action, the current alternative and what evidence would show the decision improved.
- The specification. Draft the Chapter 0 fields: decision, decision time, evidence, proposed capability, existing alternative, boundary, authority, stop condition, acceptance and value evidence. Flag any you cannot yet fill.
- Where the loop breaks. Name the stage where it stops, why, and who owns it by role.
- The honest part. What your company gets wrong about this decision today. No AI in use yet is normal.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 1 — Our company, the one decision we will carry, its decision specification and where its loop breaks.
How this is marked — the rubric
| Names the company, its place in the gas value chain and the customer promise involved | 15 |
| States one decision with deadline, physical and contractual constraints, authority by role, current alternative and improvement evidence | 30 |
| Completes the decision specification fields honestly, flagging any that cannot yet be filled | 25 |
| Locates the loop break at a named stage with a reason and an owner by role, and names a real weakness | 30 |
| Total · pass mark 70 | 100 |
Submitting session 1
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: 1, 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
Decision-first thinkingThe six toolkit frameworksTelling analytical capabilities apartWriting a decision specificationLocating the broken loop stage
Chapter 1 · Leadership and Gas EconomicsSESSION 2
Leading a Gas Business with Operating Evidence
Why This Matters
In the book's fictional supplier, the problem is not a shortage of optimism or dashboards. It is a commitment that nobody has checked as one decision: the order is booked, the vehicle awaits a part and the gas purchase must be paid before the customer pays. This session teaches you to define the promise backwards from the delivery point, to give every assumption an owner and an evidence state, and to run a daily review that follows the action through to acceptance. It also shows, with a worked example, how a correct warning can reveal a management decision rather than a technology gap. Your decision from Session 1 is the case.
25-Minute Module
0:00 – 5:00
The unfunded commitment
- The book's fictional gas supplier opens its meeting with good news: orders up, compression expecting higher output, more vehicles, higher projected turnover. Then customer service finds overlapping receiving windows, one vehicle awaits a critical part, and the additional gas purchase needs payment before the customer pays.
- THE DOCTRINE
The problem is not a shortage of optimism or dashboards. The proposed commitment has not been checked as one decision.
- The business needs an accepted operating plan, a commercial consequence and a funding decision. AI can help expose inconsistencies and compare alternatives; management has to make the commitment concrete.
- Ask of your own decision: which of those three is missing today?
Every department reporting green is not the same as one feasible promise.
5:00 – 10:00
Define the promise backwards
- Begin at the delivery point. Record the product, quantity, accepted quality basis, receiving window, location and the conditions under which the customer accepts delivery. Then trace backward through receiving capability, transport, compression or processing, feed availability and financial release.
- Do not equate production capacity with accepted delivery capacity. A compressor may support more output while the receiving station, the fleet cycle or the customer's planned shutdown restricts the sale. Capacity added upstream of a downstream bottleneck may create cost without additional accepted value.
What the team reports What it is not yet An engineer's expected return-to-service time A completed functional check A customer's forecast An accepted order A proposed credit increase An approved limit - Give each assumption an owner and an evidence state. These distinctions belong in the decision pack, not in a later dispute.
Walk backwards from the gate the customer opens, not forwards from the plant.
10:00 – 15:00
Specify the intelligence design
- An effective first design may combine three simple layers, kept separate in the executive pack. Reconciliation finds inconsistent quantities and times across records. Forecasting estimates what is genuinely uncertain. Constrained comparison produces feasible allocation alternatives.
- Build deterministic reconciliation checks before adding a model. Does the released quantity exceed the accepted order? Is the receiving window already closed? Does current equipment status permit the movement? These checks do not become more useful merely because they are described as AI.
- For prediction, ask what will be known at the decision time. A model trained with delivery outcomes available only after dispatch cannot reproduce a real planning decision. Keep historical snapshots of requests, revisions and constraints, and test on later periods.
- THE DOCTRINE
The objective might be to minimise accepted contractual shortfall at a stated cost, subject to feasible supply and approved priorities. The chosen objective is a leadership decision.
- It should not silently become maximisation of gross margin if management has promised priority service to another customer. I recommend asking your team to state the objective in plain language before they show you a single result.
Name the layer before you trust the number.
15:00 – 21:00
The daily review, and a warning with no response budget
- Use a short exception pack: the affected commitment, the constraint, the deadline for useful action, feasible alternatives and the approval needed. Include the no-change case. An option that assumes unavailable funding is shown as conditional, not as the recommended plan. The chair resolves competing priorities and records each acceptance in sequence.
- The action register follows execution: approval to purchase a part is not receipt; receipt is not fit or installation; installation is not readiness; readiness is not the customer accepting delivery.
Worked example (fictional, assumed figures) Amount Possible shortfall × assumed contribution: 20,000 standard cubic metres × ₦120 ₦2.4 million at risk Assumed accepted recovery 75%: 20,000 × ₦120 × 75% ₦1.8 million protected Less additional approved vehicle ₦900,000 Simplified net benefit, before omitted tax, overhead, financing or alternative-use effects ₦900,000 - Engineering and the customer have confirmed the vehicle arrangement is feasible in this base case. It is a decision illustration, not a Nigerian transport quotation or a safety justification.
- IN THE PHYSICAL SUPPLY · THE WORKED CASE
- Physical supply. Now the vehicle cannot reach the customer within the accepted window. The warning remains correct, but the recovery is no longer feasible. The ₦1.8 million must not be retained as a benefit simply because the model predicted the threatened volume. A different response might be early customer communication, an agreed rescheduling or a contractual settlement.
AT THE CUSTOMER PROMISE · THE PARALLEL- Customer promise. Now the response is feasible but funding is unavailable. Management must arrange funding, reallocate a resource or accept the commercial consequence within authority. The warning has revealed a management decision; another dashboard will not resolve it.
A correct warning can still carry no value if nobody can act on it.
21:00 – 25:00
Measures that reveal leadership, and Section 2
- Track the interval from material exception to authorised decision, then to accepted execution, and record overdue resource and approval dependencies. Classify rejected recommendations by reason: incorrect evidence, infeasible action, missing constraint, commercial choice or protection requirement.
- Judge outcomes against the information available at the time. Do not grade a planner with hindsight, and do not let a model team exclude the difficult regimes that matter most to the customer promise. Each week, reconstruct one failed or nearly failed commitment and fund the intervention at the point where a better decision could have changed it.
- Before you close the laptop, write the delivery-point conditions of your decision's promise in one line.
- Section 2 of your plan is The promise defined backwards, its assumptions owned and the review that follows execution.
Fund the intervention where the chain actually broke.
Key Terms
Evidence state
Where an assumption actually stands: an expected return-to-service time versus a completed functional check, a forecast versus an accepted order, a proposed versus an approved credit limit.
Exception pack
A short daily review of material exceptions, each with the affected commitment, constraint, deadline, feasible alternatives including no change, and approval needed.
Constrained comparison
An allocation layer that produces only feasible alternatives against a stated objective and stated constraints, with the objective chosen by management.
Action register
A record that follows a decision through purchase, receipt, installation, readiness and the customer's accepted delivery, rather than closing at approval.
Self-Check — Before You Submit
1. In the book's worked example, the vehicle cannot reach the customer within the accepted window. What happens to the ₦1.8 million protected contribution?
- A. It stays in the business case because the warning was correct
- B. It is halved to reflect uncertainty
- C. It is no longer a benefit; a different response must be considered
- D. It becomes the cost of a new dashboard
C. It is no longer a benefit; a different response must be considered — The warning remains correct but the stated recovery is no longer feasible, so the model must not retain the benefit simply because it predicted the threatened volume. Keeping it because the warning was right is the trap.
2. Your allocation model has been quietly configured to maximise gross margin, although management has promised priority service to another customer. What is the right response?
- A. Restate the objective as a leadership decision in plain language
- B. Accept it, since margin is the best measure
- C. Ask the vendor for higher accuracy
- D. Let each planner choose their own objective
A. Restate the objective as a leadership decision in plain language — The chosen objective is a leadership decision and must be shown in plain language. It should not silently become margin maximisation where a priority promise exists. Accuracy is irrelevant if the objective is wrong.
3. Which item belongs in the daily exception pack?
- A. A catalogue of every observation
- B. Only the recommended option
- C. Options assuming unavailable funding, presented as executable
- D. The no-change case
D. The no-change case — The pack identifies the commitment, constraint, deadline, feasible alternatives and approval, and includes the no-change case. An option needing unavailable funding is shown as conditional, not as the executable plan.
Today’s Assignment
Define the promise behind your decision backwards from the delivery point, own its assumptions, specify the intelligence design and design the review that follows execution.
- The promise backwards. For your decision in your company, record the product, quantity, quality basis, receiving window, location and acceptance conditions, then trace back through receiving, transport, compression or processing, feed and financial release.
- Assumptions owned. List the key assumptions, each with an owner by role and an evidence state.
- The intelligence design. Keep reconciliation, forecasting and constrained comparison separate, and say honestly which your decision actually needs. Name two deterministic checks and state the objective in plain language.
- The daily review. Design the exception pack, including its no-change case, and the action register that follows execution to customer acceptance.
- The measures. Name the measures that would reveal leadership performance, including how rejected recommendations will be classified. Say what your company currently gets wrong in its review.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 2 — The promise defined backwards, its assumptions owned and the review that follows execution.
How this is marked — the rubric
| Defines the promise backwards from the delivery point to financial release with the acceptance conditions stated | 25 |
| Gives each material assumption an owner by role and an evidence state | 20 |
| Separates reconciliation, forecasting and constrained comparison and justifies which are needed | 25 |
| Designs an exception pack with a no-change case, an action register to acceptance and leadership measures | 30 |
| Total · pass mark 70 | 100 |
Submitting session 2
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: 2, 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
Defining the promise backwardsOwning assumptions with evidence statesSeparating the intelligence layersRunning the daily exception reviewTurning a warning into a decision
Chapters 2 and 3 · Leadership and Gas EconomicsSESSION 3
The Gas Value Chains and Their Different Constraints
Why This Matters
A gas group may process gas, run a pipeline, distribute CNG and sell LPG, each constrained by something different. Money behaves the same way: a signed contract is not a delivery, a delivery is not an accepted bill and a bill is not cash. Today you locate the binding constraint behind your decision, follow the revenue chain through its accepted events, and run your decision through the use-case screen that applies hard gates before anyone ranks opportunities. These disciplines help your company avoid funding the wrong bottleneck or counting the same naira twice.
25-Minute Module
0:00 – 5:00
Map the business before selecting the model
- A common reporting layer can help, but a common optimisation problem is unlikely to represent every activity.
- Natural gas is predominantly methane. CNG keeps it in a compressed gaseous state; LNG is its liquid form. LPG belongs to a different product family and commonly contains propane and butane (U.S. Department of Energy, AFDC, Natural Gas Fuel Basics and Propane Fuel Basics; U.S. EIA, Hydrocarbon Gas Liquids Explained).
- Adding tonnes of LPG to standard cubic metres of pipeline gas creates a number with no physical meaning. A common energy basis must preserve the original quantities, and it does not make contractual risks interchangeable.
Model the actual business, not the most familiar image of the product.
5:00 – 10:00
The binding constraint in each chain
- A binding constraint is the limit that actually caps the accepted result. The book's Comparative Application Map is an author-developed design guide, not a technical standard.
Business Useful analytical question Necessary acceptance check Processing Which feed and operating regime threatens accepted product? Engineering envelope and product evidence Compression Which developing condition warrants a useful response? Equipment review and response feasibility Pipeline Which feasible allocation meets approved commitments? Network check and commercial authority LNG How do supply, inventory and transfer windows interact? Cryogenic-system and interface acceptance CNG Which schedule completes deliverable cycles? Loading, route, receiving and return constraints LPG What stock can be sold, funded and reconciled? Product, container, measurement and release status Industrial supply Which supply plan protects an accepted customer window? Customer receiving and continuity agreement - THE DOCTRINE
A single platform may support several workflows, but each use still needs its own evidence, acceptance and benefit mechanism. Reusing software is not equivalent to transferring technical validity.
- The book's fictional CNG business proposes an extra compressor, but trailers spend most of their time at customer receiving points. Compare a receiving-window change, compatible distribution equipment, a better return cycle and the compressor, including under an outage and a customer restart.
- NMDPRA's Automated Gas Systems portal distinguishes activities such as gas processing, CNG compression and industrial cascade systems, and LPG storage and filling (NMDPRA, Automated Gas Systems, official portal). A portal category is not itself a permission to operate. Keep an activity register linking each facility to its current permissions, conditions and accountable owner.
The correct answer may still be the compressor, but only for the changed system.
10:00 – 16:00
The revenue chain is a set of accepted events
- Keep distinct states for contracted entitlement, requested quantity, accepted nomination or order, scheduled, dispatched, delivered, accepted billable quantity, invoiced amount and collected cash. A single status called completed cannot explain a dispute or a late payment.
- Treat contracts as operating constraints: extract the delivery point, product and quality basis, unit, window, measurement, billing and payment terms, each with its source location and reviewer. Commercial or legal acceptance establishes its use.
- Take-or-pay, make-up rights, imbalance treatment, capacity reservation and force majeure have contract-specific meaning. The signed clause governs.
- An uncollected invoice is not automatically a default: it may be disputed, missing delivery evidence, awaiting a credit note or genuinely overdue.
- Build the exception engine on reconciliation rules first. An assistant may draft a dispute pack; it should not amend a price, waive an obligation, raise a credit limit or commit supply.
The unresolved gap between two events is where your exposure accumulates.
16:00 – 21:00
Profit, funding and customer prepayment
Worked example (fictional, assumed) Amount Revenue: 100,000 standard cubic metres × ₦500 ₦50 million Supply cost: 100,000 × ₦260 ₦26 million Incremental delivery cost: 100,000 × ₦90 ₦9 million Contribution before overhead, tax and financing ₦15 million Expenditure to fund if paid before collection: ₦26m + ₦9m ₦35 million Funding ₦35m for 60 days at an assumed 24% simple annual rate, 365-day year about ₦1.381 million - This is an assumed simple-interest calculation; it does not represent current Nigerian borrowing terms, tax treatment or a continuously replenished portfolio.
- IN THE PHYSICAL SUPPLY · THE WORKED CASE
- Physical supply. If only 80,000 units are accepted, revenue falls to ₦40 million; if supply and delivery costs both vary fully with accepted quantity, contribution becomes ₦12 million. If the original supply commitment remains payable, the answer changes. Finance applies the actual contractual cost behaviour.
AT THE CUSTOMER PROMISE · THE PARALLEL- Customer promise. A ₦35 million customer prepayment could cover the simplified cash requirement, but it is not ₦35 million of profit. Its benefit may be reduced financing exposure, lower collection risk or a funded delivery, each needing separate evidence. The board should see contribution, peak funding, expected collection and downside separately.
- THE DOCTRINE
A high-margin expansion can still fail for lack of cash. Count the relevant mechanism once.
- For one disputed invoice, a delivery dispute pack links the signed basis, order, delivery evidence, measurement version, price and receivable. A prepared pack is a process output, not a cash benefit. Preserve an unresolved dispute as unresolved.
Contribution, funding and collection are three numbers, not one.
21:00 – 25:00
The use-case screen, Section 3 and Pop Quiz 1
- Appendix D applies hard gates before ranking. Ask: which decision changes; what is the existing alternative; is a useful response available; are the relevant records dependable; what must remain constrained; how will analytical performance be tested; how will value be established; what happens when the application fails. A high estimated benefit does not compensate for an unacceptable authority or operating boundary.
- Record a failed gate explicitly: a contract assistant may retrieve well and still fail the authority gate. The outcome is proceed to a bounded evaluation, repair a prerequisite, redesign or stop, each with an owner and a reason.
- Before you close the laptop, write your decision's binding constraint and the first gate it fails, if any. Section 3 of your plan is The binding constraint, the revenue chain's accepted events and the use-case screen.
- Pop Quiz 1 opens now: five questions, closed book, on Sessions 1 to 3. Sit it before you start Session 4.
An attractive but infeasible proposition does not belong in an indefinite pilot queue.
Key Terms
Binding constraint
The limit that actually caps accepted delivery or value; capacity added upstream of it may create cost without additional accepted value.
Activity and permission register
A record linking each facility and activity to its source document, validity or review condition, accountable owner and evidence of compliance.
Accepted billable quantity
The quantity the customer has accepted on the contract's basis for billing, distinct from what was contracted, scheduled, dispatched or delivered.
Hard gate
A screening condition, such as authority, response or constraint, that must be passed before an opportunity is ranked; no benefit score compensates for failing it.
Self-Check — Before You Submit
1. A fictional distributor's trailers spend most of their cycle waiting at customer receiving points. Sales propose a second compressor. What should management do first?
- A. Approve the compressor on nameplate output
- B. Compare receiving, equipment, return-cycle and compressor options, including under stress
- C. Buy a forecasting platform
- D. Cut the sales campaign
B. Compare receiving, equipment, return-cycle and compressor options, including under stress — The binding constraint is at receiving, so the compressor's value must come from the changed system, not nameplate output. The book does not rule the compressor out; it requires the comparison, including an outage and a customer restart.
2. A customer pays ₦35 million in advance for the book's assumed delivery. How should the board treat it?
- A. As ₦35 million of additional profit
- B. As revenue recognised on receipt
- C. As cash that may reduce financing exposure, evidenced separately
- D. As proof the customer will accept delivery
C. As cash that may reduce financing exposure, evidenced separately — The prepayment covers the simplified cash requirement but is not additional profit. Its economic benefit, such as lower financing exposure or collection risk, needs separate evidence. Treating it as profit counts the wrong mechanism.
3. A contract assistant retrieves clauses accurately and the sponsor wants it to issue binding instructions to customers. On the use-case screen, the proposal...
- A. Fails the authority gate and may need redesign
- B. Passes, because retrieval performance is strong
- C. Should be ranked highest because benefit is high
- D. Should wait indefinitely in the pilot queue
A. Fails the authority gate and may need redesign — Appendix D's example is exactly this: strong retrieval can still fail the authority gate. A high benefit does not compensate, and the book warns against leaving infeasible propositions in an indefinite pilot queue.
Today’s Assignment
Locate your decision in your own value chain, name its binding constraint, follow its revenue chain to cash, record the permission it depends on and run it through the use-case screen.
- The chain and the constraint. Where your decision sits in your company's value chain and its binding constraint. If your business does not run the activity, apply it to the nearest equivalent decision or the counterpart you depend on.
- The revenue chain. List the accepted events from contracted entitlement to collected cash that this decision touches, and say where the unresolved gap lies.
- The permission record. The activity and permission record the decision depends on: owner by role, source document and condition. If you do not know, say so.
- The use-case screen. Answer the eight Appendix D questions; record any hard gate the decision fails, with the outcome: proceed, repair, redesign or stop.
- The honest part. Where your company currently counts a mechanism twice or funds the wrong bottleneck.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 3 — The binding constraint, the revenue chain's accepted events and the use-case screen.
How this is marked — the rubric
| Names the binding constraint with evidence and places the decision correctly in the student's own value chain | 25 |
| Lists the accepted revenue-chain events and locates the unresolved gap | 25 |
| Records the activity and permission dependency with owner, source document and condition, or states it is unknown | 15 |
| Runs all eight screening questions, records any failed hard gate and states a reasoned outcome with an owner | 35 |
| Total · pass mark 70 | 100 |
Submitting session 3
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: 3, 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 1 — The Promise, the Toolkit and the Value Chains 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 1 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Locating the binding constraintKeeping products and quantities distinctFollowing the revenue chain to cashSeparating profit from fundingApplying hard gates before ranking
Chapter 4 · Processing and SupplySESSION 4
Feed Gas Quality and Processing
Why This Matters
Throughput is not the same as saleable product. A processing plan can show inlet volume rising while accepted product falls, because the feed has changed, an analyser has drifted or a treatment unit is restricted. A similar trap can arise in any business whose output must meet a quality condition before a customer accepts it. This session shows you how to keep the product record honest, how to match a model to the evidence actually available, and how to separate a model's technical accuracy from the value of the decision it informs. If your company has no processing plant, apply it to the acceptance condition that governs your own decision.
25-Minute Module
0:00 – 5:00
Throughput is conditional on product acceptance
- EIA describes natural-gas processing as separating products and removing constituents to meet transport and use requirements, with the treatment depending on the gas and intended product (U.S. EIA, Natural Gas Explained, overview).
- The management implication: connect feed identity, operating regime and outlet acceptance, rather than apply a generic performance model to every feed. An application that forecasts only inlet quantity can miss the condition that determines saleable output.
- THE DOCTRINE
A processing plan can show increasing inlet volume while accepted product falls. Forecast the condition that determines saleable output, not only the quantity entering the plant.
- For an industrial energy user, the parallel is the accepted quality basis at your receiving point. For a CNG or LPG business it is the product and release condition on which a sale depends.
Ask what makes the product saleable before you ask how much there will be.
5:00 – 10:00
The product and transformation record
- Record each inlet stream's identity, time basis, quantity and accepted composition or analysis; link it to the process configuration and outlet products. Preserve sampling method, analyser status and laboratory reconciliation. Where composition is estimated, carry that status into every downstream calculation.
- Build a product-quality register from the accepted specification: each variable, measurement point, unit, source, method, review owner and the response to an uncertain or nonconforming result.
- THE DOCTRINE
The engineering team establishes technical requirements and limits. The analytical team represents them; it does not invent a universal acceptable range.
- Keep production, fuel use, internal transfer, venting, flaring and recovered liquids separately identifiable where they matter to the plant balance. A change in product mix can make a volume-only indicator misleading; the right balance may need mass, molar or energy accounting with engineering acceptance.
An estimate that loses its label becomes a false fact downstream.
10:00 – 15:00
Choose the model to match the evidence
- A soft sensor estimates a variable that is not continuously measured from other available observations, much as a planner might infer a tank level from inflows and outflows. It can support advisory review, but its output remains an estimate. It does not replace a required measurement or acceptance procedure without the appropriate technical and commercial authority.
- Start with a process-informed baseline. Compare estimated outlet quality with independent accepted results, grouped by feed and operating regime, then ask whether a more complex model improves the useful prediction horizon or reduces review effort. A lower average error is insufficient if the model performs poorly near the decision boundary.
- Interpolation is estimating inside the range the model was tested on; extrapolation is estimating outside it. A model tested across a known feed range may be unreliable after a materially different feed appears. Build a check for operation outside the validated range: the result should be unavailable or flagged for engineering review, not silently extended.
- For optimisation, compare alternatives only inside the accepted engineering envelope, preserving constraints on quality, equipment, utilities and downstream demand. Otherwise a model can recommend extra output that consumes unavailable utility capacity or creates an unacceptable product.
Ask your team what the model does when the feed changes.
15:00 – 21:00
Technical accuracy is not decision value
- When an estimated quality variable nears the accepted limit, the decision is whether to seek confirmation, adjust the plan or hold product under the approved procedure. Evaluate missed decision-relevant deviations, unnecessary review burden and response time. A broad correlation plot does not answer these questions.
- Evaluate chronologically across genuine changes, keep one event's overlapping records out of both sides of the test, and include start-up, changed feed and analyser degradation where the use covers them. If evidence is thin, restrict the use explicitly. Record the missing condition when an operator rejects an output; repeated overrides may reveal an omitted valve restriction or a weak operating record.
Worked example (fictional, two technically accepted cases) Case A Case B Accepted units × contribution per unit 80,000 × ₦150 = ₦12m 95,000 × ₦120 = ₦11.4m Less incremental operating expenditure ₦2m ₦2.5m Net contribution ₦10m ₦8.9m If Case B's margin rises to ₦140 ₦10m ₦13.3m − ₦2.5m = ₦10.8m - IN THE PHYSICAL SUPPLY · THE WORKED CASE
- Physical supply. Higher accepted volume yields less net contribution under these assumptions, but that does not make Case A operationally preferable; commitments, product mix, fixed-cost consequences and engineering conditions still apply. If Case B's feed has not been technically accepted, its result is a conditional scenario, not an executable alternative.
AT THE CUSTOMER PROMISE · THE PARALLEL- Customer promise. At ₦140 the ranking reverses. The review should expose the price or cost assumption that changes the answer, rather than present one recommendation to the commercial owner as permanently correct.
The number that reverses the ranking is the number to defend.
21:00 – 25:00
A controlled quality advisory, and Section 4
- Select one outlet-quality decision with dependable accepted measurements and a useful response horizon. Define the regimes in scope and run the advisory beside the existing procedure, recording prediction, evidence, operator review, response and accepted product result.
- Periodically reconstruct selected recommendations with the measurement status and conditions at the time. Scale only when evidence supports the new regime; the strongest decision may be to improve analyser availability or feed records first.
- Before you close the laptop, write the acceptance condition that decides whether your decision's output is saleable or usable, and who owns its limits. Section 4 of your plan is The acceptance condition, the product record and technical accuracy separated from decision value.
Fixing the analyser can be more progress than a new interface.
Key Terms
Soft sensor
A model that estimates a variable not continuously measured from other observations; its output remains an estimate and does not replace a required measurement or acceptance procedure.
Validated range
The feed and operating conditions across which a model has been tested; outside it the output should be unavailable or flagged for engineering review.
Engineering envelope
The accepted technical limits within which operating alternatives may be compared; the engineering team sets them and the analytical team represents them.
Conditional scenario
A result that depends on something not yet accepted, such as a feed without technical acceptance, and therefore cannot be compared as an executable alternative.
Self-Check — Before You Submit
1. In the book's worked example, what does Case B's net contribution become if its margin rises to ₦140?
- A. ₦8.9 million
- B. ₦13.3 million
- C. ₦10 million
- D. ₦10.8 million
D. ₦10.8 million — 95,000 × ₦140 = ₦13.3 million, less ₦2.5 million expenditure, gives ₦10.8 million, above Case A's ₦10 million, so the ranking reverses. ₦13.3 million is the tempting answer but ignores the incremental expenditure.
2. A vendor's soft sensor tracks outlet quality well and the sponsor proposes dropping the laboratory acceptance test. What is the right position?
- A. The estimate cannot replace the required acceptance procedure without the appropriate technical and commercial authority
- B. Agree, because the correlation is strong
- C. Agree for low-volume periods only
- D. Let the analytical team set a new acceptable range
A. The estimate cannot replace the required acceptance procedure without the appropriate technical and commercial authority — A soft sensor's output remains an estimate and does not replace a required measurement or acceptance procedure without authority. Letting the analytical team set a range is also wrong: engineering establishes limits.
3. A new feed arrives that differs materially from anything in the model's test range. What should the advisory do?
- A. Extend its prediction as normal
- B. Show the output unavailable or flagged for engineering review
- C. Average it with the previous day's result
- D. Raise the confidence threshold
B. Show the output unavailable or flagged for engineering review — This is extrapolation outside the validated range, so the result should be unavailable or flagged for engineering review rather than silently extended. Adjusting a threshold does not make an untested regime valid.
Today’s Assignment
Name the acceptance condition behind your decision, describe the record it depends on, match a model to the evidence, separate technical accuracy from decision value and outline a controlled advisory.
- The acceptance condition. The condition that determines whether your decision's output is saleable or usable in your company. Without a processing plant, use the product-acceptance or quality condition that governs your decision.
- The record. The product and transformation (or quality) record it depends on, which values are measured and which estimated, and who owns the limits by role.
- Matching the model. The baseline you would start from, the validated range, and what happens outside it.
- Accuracy versus value. Which errors matter near the decision boundary, the review burden your team can absorb and the response time available. Name one assumption that would reverse your preferred option.
- The controlled advisory. Outline a run beside the existing procedure: scope, records kept and who reviews. Say what your company currently treats as measured that is in fact estimated.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 4 — The acceptance condition, the product record and technical accuracy separated from decision value.
How this is marked — the rubric
| Names the acceptance condition and the owner of its limits by role | 20 |
| Describes the record and distinguishes measured from estimated values | 20 |
| States a baseline, validated range and behaviour outside it | 25 |
| Separates technical accuracy from decision value, names a reversing assumption and outlines a side-by-side advisory | 35 |
| Total · pass mark 70 | 100 |
Submitting session 4
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: 4, 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
Connecting throughput to acceptanceKeeping the product record honestMatching models to evidenceSeparating accuracy from valueRunning a controlled advisory
Chapter 5 · Processing and SupplySESSION 5
Compression and Equipment Reliability
Why This Matters
Compression may determine whether your business can move gas at the required conditions. The question this session teaches you to ask is not whether a model can predict a failure, but whether its warning arrives early enough for a feasible, funded response that changes an operating consequence. You will learn why a headline accuracy figure can hide a model that finds nothing, how to count warnings honestly, and why avoidability, not downtime, drives the business case. If your company runs no compressors, apply it to the warning on which your own decision depends.
25-Minute Module
0:00 – 5:00
The warning must precede a useful response
- A reliability application needs an equipment identity, a defined condition, a useful warning horizon and an accepted response. The business case depends on whether that response can change an operating consequence.
- Begin with machine type, service, duty, configuration, redundancy and relevant failure mechanisms. A model tested on centrifugal equipment should not be transferred to a reciprocating compressor merely because both compress gas.
- Baker Hughes' technical discussion describes component-specific monitoring considerations for reciprocating machinery (Baker Hughes/Bently Nevada, Condition Monitoring of Reciprocating Compressors Oil & Gas). It is manufacturer technical material, not independent validation of an AI investment return.
A warning nobody can act on in time is a report, not a reliability programme.
5:00 – 10:00
The fault and evidence map
- For each advisory, name the mechanism to be detected and the observations that could support or contradict it. A temperature or vibration departure is an observation; a diagnosis needs competent interpretation alongside load, operating state, sensor status and other evidence.
- Separate confirmed faults, suspected conditions, planned maintenance, external supply interruptions and protective trips. A shutdown caused by a downstream restriction is not automatically a compressor failure. A work order closed by an administrative update does not establish repair or return to accepted service.
- THE DOCTRINE
Preserve unknown causes as unknown. A model built on an apparently complete set of invented labels is less defensible than one built on a smaller set of dependable events.
- The reliability owner should help define event labels, linked to equipment, component, observation time, review time, intervention and follow-up evidence.
Ask who labelled the failures before you read the model's score.
10:00 – 16:00
Episodes, decision time and the rare-fault trap
- Many sensor samples can belong to one developing event. Counting each as a separate success inflates the evidence. Group related warnings into episodes and examine each one's useful lead time, review burden, accepted diagnosis and response. Keep training and evaluation separated chronologically, and test transfer to another machine rather than mixing its samples.
- Two terms in plain words. Sensitivity is the share of real faults the model catches. Precision is the share of its warnings that turn out to be real faults. Accuracy alone can mislead when faults are rare.
Assumed evaluation Result 990 normal and 10 fault episodes; every episode called normal 990 of 1,000 correct: 99% accuracy, no faults found Model flags 8 of the 10 faults Sensitivity 80% Same model raises 32 false warnings: 8 real out of 40 warnings Precision 20% - Neither figure is a universal acceptance threshold. The test concerns consequences and resources: missed relevant episodes, warning time remaining, unnecessary specialist reviews, and whether a false warning displaces attention from another important asset.
When faults are rare, ask how many it found, not how often it was right.
16:00 – 21:00
Protection, authority and avoidability
- THE DOCTRINE
An analytical advisory is not automatically a protective function. Machine limits and operating procedures come from the accepted technical arrangements, not from a model's confidence score.
- Engineering establishes the advisory's role, permitted influence, failure response and relationship to existing protection. Another company's reported experience is not a basis for extending operation on this machine, and the economic comparison never cancels a necessary protection requirement.
Worked example (assumed) At 50% avoidable At 20% avoidable 40 outage hours × 8,000 standard cubic metres per hour × ₦120 ₦38.4m exposure ₦38.4m exposure Protected contribution ₦19.2m ₦7.68m Less incremental annual cost (monitoring, review, response, support) ₦12m ₦12m Simplified net value, before tax, financing and other consequences ₦7.2m negative ₦4.32m - IN THE PHYSICAL SUPPLY · THE WORKED CASE
- Physical supply. The response may be specialist review, inspection, planned repair, a delivery revision or more evidence. A nominal 24-hour warning may have limited value where a compatible part or specialist cannot be mobilised in that period. The quantities and avoidability are teaching assumptions, not a compressor record.
AT THE CUSTOMER PROMISE · THE PARALLEL- Customer promise. If a delivery can be recovered later, its contribution may be delayed rather than lost. If another machine covers the commitment, count the incremental system consequence, not the unavailable machine's theoretical output. Require evidence for the response mechanism, not a generic avoided-downtime percentage.
Avoidability, not downtime, is the number that carries the case.
21:00 – 25:00
One machine, one response chain, Section 5 and Practice Exam 1
- Pilot with a named reliability owner, defined regimes and available response resources. Record every eligible episode, including rejected recommendations and unavailable outputs, and fix the comparison and benefit calculation before anyone reports success. Review model performance separately from action performance: a useful warning followed by an unfunded repair reveals a management dependency to resolve before scaling.
- Before you close the laptop, write what counts as an episode for your decision, its useful warning horizon and the response that must fit inside it. Section 5 of your plan is The warning, its episodes and horizon, and the response it must precede.
- Practice Exam 1 opens now: fifteen questions, closed book, covering Sessions 1 to 5. It closes Week 1; sit it before you start Session 6.
Dependable monitoring includes the ability to execute.
Key Terms
Episode
A group of related warnings belonging to one developing event, evaluated as one unit for lead time, review burden, diagnosis and response.
Sensitivity
The share of real fault episodes a model identifies; in the book's assumed case, eight of ten, or 80%.
Precision
The share of a model's warnings that are real faults; in the book's assumed case, eight of 40, or 20%.
Avoidability
The share of exposure that the advisory and a feasible response can actually prevent; the book's 50% and 20% cases show how this assumption can change the investment case.
Self-Check — Before You Submit
1. A vendor reports 99% accuracy on a dataset of 990 normal and 10 fault episodes. What should you ask first?
- A. Whether the accuracy can reach 99.5%
- B. How many of the 10 faults it identified, and with how many false warnings
- C. Whether the model is centrifugal-ready
- D. How many sensor samples were used
B. How many of the 10 faults it identified, and with how many false warnings — Calling every episode normal already gives 99% accuracy while finding no faults. Ask about sensitivity and precision by episode. Counting samples is the inflation trap, not the answer.
2. Using the book's assumptions at 20% avoidability, what is the simplified net value?
- A. Negative ₦4.32 million
- B. ₦7.68 million
- C. ₦7.2 million
- D. ₦19.2 million
A. Negative ₦4.32 million — 40 × 8,000 × ₦120 × 20% = ₦7.68 million protected, less ₦12 million cost, gives negative ₦4.32 million. ₦7.68 million is protected contribution before cost; ₦7.2 million is the 50% case.
3. A model gives a 24-hour warning of a developing compressor condition, but the compatible part takes longer to mobilise. The sponsor wants to count the full avoided outage. What should you do?
- A. Count it, because the warning was correct
- B. Turn the advisory into a protective trip
- C. Treat the missing response as a management dependency and count only what a feasible response can avoid
- D. Extend operation, citing another company's experience
C. Treat the missing response as a management dependency and count only what a feasible response can avoid — Value depends on a response feasible within the warning horizon; an unfunded or unmobilisable response reveals a management dependency. An advisory is not automatically a protective function, and another company's experience is no basis for extending operation.
Today’s Assignment
Draw the fault and evidence map for the warning your decision depends on, define its episodes and horizon, name the response and its funding, state the avoidability evidence and outline a one-machine, one-response-chain pilot.
- The map. The condition or fault mechanism in your company that your decision depends on, the observations that would support or contradict it, and how confirmed faults, suspected conditions, planned work, external interruptions and trips are separated. Without compression, use the equivalent warning in your business.
- Episodes and horizon. What counts as one episode and the useful warning horizon.
- The response. The response that must be feasible within that horizon, who funds it and who authorises it by role.
- Avoidability. The evidence you would require before counting value, and whether lost delivery is truly lost or only delayed or covered.
- The pilot. One machine (or equivalent) and one response chain: owner, regimes, records kept including rejected recommendations. Say honestly where your company's response chain is unfunded today.
- Four hundred to nine hundred words.
YOUR 90-DAY PLAN
Section 5 — The warning, its episodes and horizon, and the response it must precede.
How this is marked — the rubric
| Draws a fault and evidence map that separates observation from diagnosis and keeps unknown causes unknown | 25 |
| Defines an episode and a useful warning horizon | 20 |
| Names a response feasible within the horizon with funding and authority by role | 25 |
| States avoidability evidence and outlines a one-machine, one-response-chain pilot with a named owner | 30 |
| Total · pass mark 70 | 100 |
Submitting session 5
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: 5, 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 1 — Leadership, Gas Economics and Processing 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 1 →Closed book, no notes, and Coach Ritchie switches himself off while you sit it.Capabilities Covered
Mapping faults to evidenceCounting episodes, not samplesReading sensitivity and precisionTesting avoidabilityKeeping protection with engineering