Chapter 31: Execute Predictive Work with Discipline

Four numbers for one month

It is month twenty-four at BlueLine, and the corridor’s progress review has four numbers for the same month. All four are true. None of the four agree.

The contractor’s report says the ticketing-to-traffic seam package is eighty percent physically complete, its rule counting hardware installed, wired, and powered. The project’s own measurement says forty-five percent of the package’s value is earned, its rule counting accepted evidence. The difference between eighty and forty-five is not a rounding dispute. It is the chapter. The authority’s certifier has approved fifty-five percent of the package for payment, more than the earned number, because the contract’s milestone schedule bills materials and mobilization as they arrive, ahead of the evidence that the interface works. And the mayor’s office has announced that the corridor’s ticketing system is substantially installed and entering final testing, which reads to the public as almost done, and is the fourth number, the one the room cannot verify, measure, or invoice.

Marta Reyes, the transport authority’s finance officer, puts the four numbers on one page, and the page is the chapter’s argument in a table:

Lena Voss, the project director, asks the question the page is asking: which number do we forecast from, which do we govern with, and which would we still defend if the corridor opened and the seams failed on the first peak hour? She does not have to answer, because the dry run answered on Tuesday. The rehearsal between the ticketing vendor’s system and the traffic vendor’s system — the seam that chapter 20 let under a shared-savings target-cost contract, because no single contract owned the interface — ran at design load, and at design load the ticketing events feed to the control room dropped messages, and the traffic-priority handoff silently discarded the ones it did not acknowledge. Each system passed its own tests. The seam failed the integrated run. The seam was the risk chapter 20 named, and it arrived in the month when the mayor’s month-twenty-four promise was already behind, when the phased opening of the central and northern segments at month twenty-five was the plan, and when the corridor’s forecast at completion sat at about 2,550 million units against the 2,540 the second-half authorization carried in month seventeen.

That is the subject of this chapter. Chapter 30 built the execution rhythm that carries work forward at the rate the system can absorb, and its predictive register pointed here: the control cycle of baseline, actuals, forecast, and decision, the clock that executes plan-driven work. This chapter is the discipline of that clock: keeping the model honest. The baseline is fixed until authorized change moves it, the actuals are recorded as evidence rather than as impressions, the forecast is recomputed every cycle from that evidence, and the decision is made on the forecast before the month ends. Baseline control and rigidity are different things. Rigidity refuses to reforecast because the baseline is sacred. Baseline control refuses to move the baseline without a change record, because the baseline is the project’s memory. And the skill that separates the two is knowing what counts as progress, because everything else in the control cycle is built on that answer.

The baseline is a promise with a process

Predictive execution begins with the instrument of authorization, and the instrument is smaller than the plan. The integrated plan of chapter 16 is a model of the whole project. The project cannot authorize against the whole model at once, and it does not need to. It authorizes against the smallest units the plan can stand on: the work packages that chapter 14 decomposed, grouped into control accounts. Each account has a manager, a budget, a schedule, and a scope boundary. Each work package has the four things that make it executable: its acceptance criteria, referenced from chapters 10 and 21; its measurement rule, the subject of a later section; its owner; and its limits.

The control account is the unit of accountability, and the term is worth keeping because it says what the account is for: control, not bookkeeping. It is the point where scope, budget, schedule, and risk come together under one accountable manager. When a package inside the account goes wrong, the question is not who filed the wrong form. It is who owns the account and what they did with the deviation. The work package is the unit of work and of measurement: discrete enough to be assigned, executed, and measured within a reporting period, the practical rule chapter 14 carried from the earned-value criteria codified in ANSI/EIA-748, the standard used widely in defense and government procurement. Progress is measured and the evidence lives at the package, not at the account and not at the phase.

Authorization is the discipline that connects the plan to the work, and its minimum viable form is the work-package control sheet: one page, a few rows, and a rule. The rows name the package’s scope reference, control account, budget, start and finish dates, measurement rule, acceptance criteria reference, owner, and tolerance — the schedule or cost variance the owner may absorb without escalating. The rule is the delegation contract the book has built since chapters 19 and 26: the owner executes within the tolerance without asking, reports variance beyond the tolerance to the account manager, and sends variance beyond the account to this chapter’s change process. The authority the plan grants is bounded, and the boundary is what makes the delegation safe.

The failure of the control sheet is the timesheet: the sheet that records what was done instead of what was earned, a status column filled with a percentage someone judged, a forecast column left blank because the plan is the plan, and an account nobody owns because the manager’s name was never written. The field signal is the control sheet no one reads at the control meeting, because the meeting reads the narrative report instead, and the narrative report says the package is on track while the sheet’s evidence column is empty.

The baseline is the sum of these authorized units, and its character is worth stating, because everything in this chapter follows from it. The baseline is a promise with a process. It promises what the project will deliver, at what cost, by when, and it is fixed until the process changes it, because a promise that changes whenever the work gets hard is not a promise, it is a hope with a timestamp. But the baseline is also a model, not the work itself: a model built from estimates that chapter 15 taught to distrust, from assumptions that chapter 3 taught to inventory, and from a decomposition that chapter 14 taught to check for the missing hundred percent. The discipline is holding the promise and the model in the same mind: the baseline is sacred as a record and provisional as a prediction, and the control cycle exists to keep those two truths from collapsing into each other. Treated only as a record, the baseline becomes the frozen baseline, the plan where nothing can change and the forecast decays in silence. Treated only as a prediction, it becomes the moving goal, and the project’s memory is erased a little every month.

The look-ahead makes the near term concrete

The second instrument is the look-ahead, and it is the same instrument chapter 30 built for the flow system, running here in the predictive register. The plan’s long view is the network: the chain of dependencies, the critical path, the milestones, the floats that chapter 17 taught. The network is a model of time, and it is too coarse to execute from, because it says what should happen and not who will do it this week, with what, against what evidence. The look-ahead is the network’s operating translation: the next four to six weeks written as concrete actions with owners, dates, dependencies, hold points, and evidence, drawn from the network and the control accounts, reviewed on the monthly clock, and revised as the actuals arrive. The rolling-wave discipline of chapter 16 applies exactly: the near term planned in detail, the far term at the level the information supports, the detail elaborated as the wave rolls forward.

The look-ahead’s job is authorization at the point of work. The plan authorizes the package; the look-ahead authorizes the start, the resources, and the sequence. The discipline is that the look-ahead is written from the network, not from the calendar, because the calendar says when the month ends and the network says when the work can start. At BlueLine, the look-ahead of the commissioning window is the form in miniature: four weeks listed, each item with its owner; the dry-run re-run scheduled against the ticketing vendor’s repair team’s availability; the hold points where the authority’s acceptance inspectors will witness; the operator’s training sessions; the dependency rows the network draws; the eastern segment’s drainage works scheduled to resume the day the wet season breaks; and the two-week mobilization rule that chapters 17 and 27 carried, the lead time the contractors need to move crews onto the corridor, which means the look-ahead must see four weeks ahead to start anything in two.

The failure of the look-ahead in the predictive register is the calendar decoration: the list that repeats the milestone dates without owners, dependencies, or hold points — the plan that says the dry run will happen in month twenty-five and does not say who runs it, who witnesses it, what evidence it produces, or what happens if it fails. The field signal is the look-ahead the control meeting reads and the work does not follow, because it was never a plan, it was a memo about the plan. The look-ahead that works can be tested against the network: every item maps to a package in the accounts, every owner is a real person with real authority, and every date is a commitment the account manager can be asked about at the next review. The look-ahead is where the plan’s promises become someone’s week.

Progress is earned, not claimed

The third instrument is the one the opening scene was about, and it is the heart of the chapter: the measurement of progress. Everything else in the control cycle — the actuals, the forecast, the decision — is built on the progress number, and if the progress number is a guess, the whole cycle is a guess wearing a spreadsheet. The discipline has a name in one sentence: progress is earned, not claimed.

The contractor’s eighty percent and the project’s forty-five percent were both measurements of the same physical reality. The difference is not that one party lied. It is that they counted different things: the contractor counted installation, the project counted accepted evidence of function. Both numbers were true under their rules, and the craft is choosing the rule, because the rule decides what the project is actually forecasting.

The four progress streams of the opening scene deserve to be separated, because each answers a different question and has a different use. Physical progress is the claim of what exists: work performed, materials installed, structures erected, measured by the performer’s own rules, and it is the stream that must be verified, because it is the easiest to count in a way that flatters. Earned progress is the value of accepted evidence: the work measured against the project’s agreed rules, the milestones reached, the units complete, the tests passed, and it is the only stream the forecast should be built from, because it is the only one defined by evidence rather than by claim. Invoiced progress is the commercial stream: what has been billed, certified, and paid, measured by the contract’s payment milestones, and it can run ahead of earned, the payment-before-evidence warning, or behind, the cash-flow warning of chapter 18. Announced progress is the public stream: what the organization says to the world, with the strongest incentive to be optimistic and the weakest connection to evidence, because its audience cannot check it, and its cost, when it diverges from earned, is the credibility of everyone who signed the announcement.

The reconciliation of the four streams is a control practice in its own right, and its minimum viable form is one line per package — physical, earned, invoiced, announced, each with its rule — on one page at every review, so the divergence is visible and must be explained. The divergence is the signal. Physical far above earned says the work is installed but not proven. Invoiced ahead of earned says the contract is paying ahead of evidence. Announced ahead of earned says the public story is running ahead of the project’s own measurement, and the question is the one the mayor’s office never asks: what happens to trust when the announcement and the earned number meet in public? The stream missing from the reconciliation is the one the project is governing with, and most projects govern with the physical claim, the stream they can least verify — which is how the seam package could reach eighty percent physical while earning forty-five.

The measurement rules are the grammar of earned progress, and the minimum viable set is five, each with its evidence and its gaming risk. The strongest is the weighted milestone: the package divided into named events with weights that sum to the package’s budget, value earned only when the event’s evidence is accepted — the inspection signed, the test passed, the certificate issued — so progress jumps at the evidence and never glides on effort. Weighted milestones are the rule for packages whose work cannot be counted in units: station fit-out, integration, commissioning. The unit-complete rule measures progress by countable output — the culvert installed per section, the road completed per kilometer — with the physical or survey record as evidence, and the risk of counting the units without the quality. The start-finish rule earns nothing at start and everything at finish — twenty percent at start, eighty at finish, or fifty-fifty — a coarse rule for small packages where finer measurement costs more than it returns, with the risk of a finish claimed before the evidence is accepted. The apportioned rule measures a supporting package as a proportion of a base package — drainage apportioned to the road section, quality records to the works they document — with the risk that the support work drifts with a base that is itself slipping. And the level-of-effort rule measures a package by the calendar and the hours spent, legitimate only for work that cannot be measured any other way: the support contract, the administration, the standby. Its risk is the greatest of all, because level of effort converts time into progress, and time spent is not value earned; the rule must be capped, reviewed, and never allowed on a package whose output can actually be measured.

The failure pattern of progress measurement is the ninety-percent syndrome, and it deserves its name because it is the most common arithmetic lie in project control. The package reaches ninety percent quickly, by the rule that counts effort, and then stays at ninety percent for months, because the last ten percent is where the seams live: the integration, the acceptance, the fix of the thing that was never tested. The ninety percent number hides the work that remains while the report says almost done. The syndrome has two costs. It hides the true position from the forecast, so the plan’s dates hold until the month they cannot. And it trains the room to discount the progress number, so the package that reports fifty percent honestly is disbelieved because the package that reported ninety percent was wrong, and honest measurement dies with dishonest measurement’s credibility. The repair is the rule, not the vigilance: a package that earns progress only on accepted evidence cannot reach ninety percent by claim, and the syndrome cannot survive a rule that refuses to count effort.

The judgment in progress measurement is choosing the rule for the package, and the test is the one the mastery drill applies: a rule is credible when the evidence is independent of the person being measured, when the progress it counts is defined before the work starts, and when the number it produces can be verified by someone who did not do the work. A rule that depends on the performer’s own judgment of percentage is not a rule, it is a request for optimism. A rule defined after the work is half done is a rationalization. A rule no one else can check is a claim wearing a rule’s clothes. The BlueLine seam package fails the first test under the contractor’s rule, because installation is counted by the installer, and passes it under the project’s rule, because acceptance evidence is judged by the inspector. Whose rule governs the earned number is a governance decision, not a technical one, and it belongs to the authority that owns the forecast — at BlueLine, the authority that pays.

Interfaces and configuration are where control lives

The fourth instrument is the pair of registers that most plan-driven projects treat as administration until the moment they fail. The chapter’s claim is that they are control instruments, not filing systems. The interface register is the list of every seam where two parties, two systems, two contracts, or two packages must meet: the name of the interface, the two sides, the owner of each side, the owner of the seam itself, the agreed behavior, the evidence that proves the seam works, and the status. The configuration register, or configuration baseline, is the list of the project’s controlled items — drawings, specifications, versions, models, interface definitions — and the rule that only authorized versions are used for work, procurement, inspection, and acceptance. The two registers are the same idea in two domains: the project’s boundaries are controlled, and the control is maintained by record, not by memory. In a plan-driven project the physical work is far downstream of the documents. The documents drift, the work follows the drift, and the as-built reality quietly separates from the baseline — which is how the corridor ends up with a ticket validator configured to a platform assignment that the operator’s re-timetabling changed in a document no one re-issued.

The seam between the ticketing and traffic systems is the chapter’s object lesson because it is the interface that was not owned. Chapter 20 exposed it in procurement terms: the contract rewarded each vendor for completing its own system, no contract rewarded the seam, and the seam was let under a shared-savings arrangement that made both vendors financially exposed to the integration outcome. This chapter shows the operating form of the same lesson. The interface register should have carried the seam as a row from the design phase: the data feed, the two vendors, a seam owner named, the message format agreed, the load profile specified, the evidence defined as the dry run at design load. The row existed. The load profile did not. The interface specification defined the message format and not the peak-hour load, and the operator’s re-timetabling, which reassigned buses to gates and changed the profile, was a change to the interface’s environment that nobody registered, because the timetable belonged to the operator and the operator was not in the configuration loop. The dry run failed at design load because the design load was never specified. That is how interfaces fail in predictive projects: not with a bang at the seam, but with a specification that quietly stopped being true.

The discipline that prevents it is small and boring. The interface row names a seam owner, not two owners, because a seam owned by both sides is owned by neither. The configuration baseline is versioned, and the versions are distributed and enforced: the field uses the current authorized version or stops work. That is a strong statement and meant to be, because the field that works to an obsolete version builds an asset that fails acceptance. And the change to a controlled item flows through this chapter’s change process even when the change is small, because the small change is the one absorbed without a record, and the unrecorded change is the drift that becomes the nonconformance at the hold point, discovered at the moment when rework is most expensive. The field signal of configuration drift is the question answered differently by the drawing and the asset: the inspector asks why the validator is wired to the wrong group, and the answer is a chain of undocumented decisions no one can reconstruct. The project’s memory was the gap between the drawing and the wall.

Inspection is the gate the work must pass

The fifth instrument is inspection, and its character is the gate, not the audit. In a predictive project the quality plan of chapter 21 becomes concrete as a sequence of control points in the work: hold points, where work stops until the authorized inspector witnesses and releases it, and witness points, where the inspector observes but the work may continue. The sequence of hold points is the quality plan’s operating form, written into the look-ahead, staffed with the inspectors’ availability, and priced into the schedule, because a hold point that waits for the inspector is a schedule delay the plan should have carried. The BlueLine commissioning window is dense with them: the acceptance inspectors witness the ticketing system’s segment tests, the traffic system’s signal integration, the seam’s re-run, and the operators’ drills, and each witness is an evidence event, signed, dated, and filed with the package’s acceptance record. The acceptance of the corridor is not a ceremony at the end. It is the accumulated evidence of every hold point that passed.

The nonconformance record is the instrument that makes inspection honest, and its minimum viable form is the report that says what the evidence shows, what requirement it fails, what disposition is proposed, and who decides. The disposition family is four: rework, the work corrected and re-inspected; use-as-is, the deviation accepted because the requirement was conservative and the deviation is within the intended function, with the reasoning recorded; concession, the deviation accepted because the requirement cannot be met as written and the deviation does not defeat the intended function, issued by the authorized party, not by the team that wants to proceed; and reject, the work removed. The discipline is that the disposition is a decision with an owner and a record, not a discussion, and that rework is measured, because rework is the cost the ninety-percent syndrome hides: the package that reported ninety percent of value and needed forty percent of its budget in rework was never at ninety percent, and the forecast that did not see the rework was the forecast that failed.

The failure pattern of inspection in predictive projects is the inspection that approves by pressure. The schedule is tight, the opening is announced, the inspector is under time pressure, the evidence is incomplete but the consequence of delay is visible to everyone, and the disposition drifts toward acceptance because rejection is expensive. The signal is the acceptance record with no evidence attached, or the evidence attached after the decision, or the nonconformance closed with a disposition no authorized party signed. The chapter 21 lesson holds in its strongest form here: acceptance is the judgment that the outcome is served, and the judgment is worthless if the evidence is not real, and the evidence is not real if the pressure to open can override it. The authority that owns the acceptance inspectors owns the ability to say no. That is why chapter 24 built the assurance lines, why the inspectors belong to the authority’s side and not the consortium’s, and why the moment the inspector’s independence is traded for a date, the corridor’s safety case, its revenue case, and its public credibility trade with it.

Time and money meet in the forecast

The sixth instrument is the integration of schedule and cost, and its site is the forecast at completion. The schedule and the budget are not two reports; they are two views of the same system, and the system’s health is visible only where they meet. The meeting point is the earned value of the progress rules: the package’s budgeted cost, multiplied by its earned progress, is its earned value, the amount of the baseline’s money the evidence justifies, and from that single number the schedule and cost indices follow. The earned-value vocabulary is standard practice, codified in ISO 21508:2018, Earned value management in project and programme management, and chapter 38 builds the full system. This chapter uses its minimum working form: planned value, the budgeted cost of the work the plan said should be done; earned value, the budgeted cost of the work the evidence says was accepted; and actual cost, the money actually spent. Their ratios are the indices: the cost performance index, earned divided by actual, and the schedule performance index, earned divided by planned. Both are signals to examine, not verdicts to obey.

The seam package at month twenty-four is the worked example, and its numbers are teaching constructions built from the package’s facts. The package’s cost budget is 40 million units, the target cost of the shared-savings contract. The milestone weights, the package’s five events — specification frozen, lab integration, simulated peak run, dry run, and end-to-end acceptance with the operator — are ten, fifteen, twenty, twenty, and thirty-five percent. At the month-twenty-four review, the plan scheduled the dry run complete, so the planned value is sixty-five percent of forty million, 26 million units. The evidence has accepted three of five milestones, so the earned value is forty-five percent, 18 million units. The actual cost is 36 million units, because the package is procurement-heavy, the hardware arrived early, and the failed dry run consumed an expensive week. The indices follow: the schedule performance index is 18 over 26, about 0.69, and the cost performance index is 18 over 36, 0.50.

A cost index of 0.50 on a package is normally the smell of serious trouble, and the discipline is reading it correctly. The index is a signal that the package’s cost is running far ahead of its earned value, and the examination shows why: the procurement profile that spends early and the failed run that spent without earning. The naive extrapolation would be the estimate at completion of 40 divided by 0.50, 80 million units, and it is wrong, because the index is distorted by the lumpy integration phase. The honest forecast is bottom-up: the remaining work costed by the repair estimate and the acceptance plan, the 36 million spent plus the 6 million repair, a forecast actual of 42 million, inside the contract’s ceiling of 48, priced through the shared-savings formula, which a later section carries.

The integration lesson is the one the book has used since chapter 15: the forecast is a judgment with assumptions, not an arithmetic output. The indices are the alarm that makes the room look, and the bottom-up forecast is the result of looking. A project that extrapolates from a distorted index manufactures a crisis; a project that ignores the index manufactures a surprise. The control cycle’s discipline is to let the indices ask the question and to answer with the evidence, every month, on the clock, so that the forecast is always the current best statement of where the project ends, and the decision, the fourth node of the cycle, is always made against that statement.

Change is the immune system, not the enemy

The seventh instrument is the change process, and its character is the immune system: the mechanism that distinguishes the changes the project can absorb from the changes that change the project, and that records every one. An immune system that cannot tell the difference is the body that rejects what it needs or admits what it cannot survive. Chapter 16 built the change threshold discipline, the rule that fixes what counts as a change and what counts as normal variance, and this chapter runs it: the tolerance in the control sheet absorbs the small deviations, the account manager absorbs the variance within the account, and the change process takes everything beyond, with the impact assessed across the dimensions that matter — value, scope, schedule, cost, quality, risk, operations, benefits — and the decision made by the authority whose approval the threshold requires.

The eastern variation is the chapter’s model of formal change done well. The flood-plain drainage approval’s conditions added ten weeks of culvert and flood-mitigation works to the eastern segment. The variation was priced at 38 million units against a contingency of 60 million. It came to the steering committee as a decision brief in the form chapter 27 built, was decided with its options and evidence, was recorded in the baseline-change record, and the baseline was rebaselined with the change, so the corridor’s forecast, its cash curve, and its commitments all moved together through the authorized path, and the record kept the truth of both before and after. That is what formal change is for: not to slow the project, but to make sure that when the project changes, it changes as a decision with an owner, a price, and a memory, and that the memory survives the people who made the decision.

Claims awareness is the change process’s commercial face, and it is the discipline most plan-driven leaders learn only after they have been burned. The claims world is the world of entitlement: the contractor’s entitlement to time and money when the employer changes the work, the employer’s entitlement to performance when the contractor misses the obligation. The record is the currency of that world, because a claim is decided on the contemporaneous record — the notices, the letters, the photographs, the inspection reports, the minutes — not on the memory of the argument. The early-warning duty is the practice that keeps claims honest: the contractual obligation, carried in the NEC contract family since its first edition in 1993, to notify the other party promptly of any matter that could affect time, cost, or quality. The warning is written when the matter is known, not when the position is certain, because the notice that comes after the damage is the notice no one trusts. The failure pattern of claims awareness is the claim that arrives as a surprise: the dispute that begins at the end of the project because the notices were never written, the photographs never taken, the minutes never kept, and the allocation that could have been a settlement becomes an arbitration, because the record that would have decided it was never made. The repair is not litigation readiness, it is record discipline: the early warning sent, the nonconformance recorded, the site photographed, the decision recorded, the baseline-change record kept. The allocation of the seam repair, the subject of a later section, is exactly the kind of question the record settles, because the shared-savings contract and the interface register together say who carries what.

The failure pattern of the change process in the predictive project is the change that is absorbed instead of routed: the operator’s new requirement, the small scope addition, the “it’s just this once” that chapter 30 named for the flow system, absorbed into the work without an impact assessment, without a baseline-change record, without a price. The absorption is invisible in any single month and visible at the end, when the corridor’s scope has grown, its schedule has stretched, and its baseline no longer describes what was built. The field signal is the audit question nobody can answer: when was this requirement added, who approved it, what did it cost, and what did the project give up to absorb it? The project that can answer has a change process. The project that cannot has a baseline that is fiction, and a forecast built on the fiction. The change process is not the enemy of progress. It is the mechanism that keeps the baseline true, and a true baseline is the only thing worth controlling.

Reforecast without rewriting history

The eighth instrument is the reforecast, and its discipline is one sentence that protects the control cycle from its two diseases: the forecast moves every cycle, and the baseline moves only through authorized change. The two truths are not in tension; they are the two halves of the same honesty. The forecast must move, because it is the current evidence-based expectation, and the evidence changes monthly: the actuals arrive, the progress rules earn or refuse, the risks fire or retire. The forecast that does not move is the denial that chapter 41 will diagnose. The baseline must not move, because it is the record of what was authorized, and the record exists to be compared against; the comparison, the variance, is what makes the decision cycle meaningful. A baseline that moves to match the forecast each month is not a baseline, it is the forecast wearing a costume, and the variance it reports, always zero, is the silence the control meeting mistakes for health.

The reforecast’s discipline is the ritual of the monthly control cycle: the actuals recorded from evidence, the earned progress computed under the rules, the indices read as signals, the forecast recomputed bottom-up, the variance explained, the baseline unchanged, and the decision made — the node the whole cycle exists for: the change approved or deferred, the plan re-sequenced, the reserve released, the risk retired, the concession recorded. The primary visual of this chapter is that cycle, and it is worth drawing at BlueLine in the commissioning window, because the drawing is the argument.

Figure 31.1: The predictive control cycle at BlueLine, month twenty-four.
The baseline holds until authorized change moves it. The actuals record
what the evidence shows. The forecast recomputes every cycle from the
evidence. The decision acts before the month ends. The cycle is a loop:
the decision returns to the work, and the work returns as new actuals.

BASELINE                 ACTUALS
fixed until change;      recorded from evidence;
the project's memory     the project's truth
     ^                         |
     | authorized change       v
DECISION                 FORECAST
made on the clock;       recomputed each cycle;
change, reserve,         bottom-up from the rules;
sequence, concession     assumptions stated

The two diseases of the reforecast are the two directions the cycle can decay. Baseline drift is the silent rebaseline: the plan’s dates and budgets quietly adjusted month by month to match the actuals, no change record, no decision, so the project is always on track by construction, and the room discovers the drift only when someone asks what the original commitment was, and no one can find it, because it was edited away. The frozen forecast is the opposite: the baseline treated as prophecy, the reforecast suppressed because the numbers would be uncomfortable, the forecast held at the authorized value long after the evidence says otherwise, so the decision cycle decides against a number everyone privately disbelieves, and the surprise, when the forecast finally moves, is larger than it needed to be, because the movement was delayed, not avoided. The field signal of both diseases is the same meeting: the control review that reads the variance without a question, the room that hears the number, nods, and moves on, because the number has been the same for months, whether by drift or by freeze.

The baseline-change record is the instrument that keeps the two truths separate, and its minimum viable form is the row that carries what changed, why, who approved it, at what authority, what it cost, what it moved, and what the baseline was before and after. The eastern variation is a row in that record. The seam repair, as a later section shows, is not a baseline change, because the package’s scope and the corridor’s commitment did not change; the repair is a variance within the authorized envelope, a forecast movement without a baseline movement. The difference between the two is the whole discipline of this chapter: the forecast moved, the baseline held, and the record says which was which, so the corridor’s memory stays true. The rebaseline itself, when it comes, is a decision, not a drift: the scope change that is material enough, the contract change that is signed, the reset approved by the authority that owns the baseline, and the record that marks the moment, because the project that cannot say when its promise changed is the project that cannot be governed.

The same discipline in every register

The discipline of this chapter is the discipline of the whole book in the predictive register, and the approach changes the clock, not the honesty. The adaptive register, KijaniPay’s empirical loop of chapter 30, runs the same cycle on a faster clock: the definition of done is the measurement rule, the review is the inspection, the acceptance evidence is the earned progress, and the change process runs through the backlog refresh, with the baseline as the outcome commitment and the forecast recomputed from the flow — with the same two diseases, the baseline drift that redefines done to match the team’s progress, and the frozen forecast that holds the date while the flow tells another story. The hybrid register, Meridian’s clinics of chapter 13, runs the predictive cycle on the physical and regulatory stream and the empirical loop on the workflow and training stream, and the seam discipline of chapter 33 is exactly the interface register of this chapter applied across methods: the row that owns the boundary between the gate and the cycle, and the change that flows through the right path. The crisis register, Northstar’s response of chapters 23 and 29, compresses the cycle to its minimum: the baseline is the floor, the actuals are the daily report, the forecast is the next day’s plan, and the decision is the huddle, with the floors that never move and everything else reforecast daily, the same separation of promise and model, running in hours instead of months.

The general principle is the one the book has held since chapter 13: the control cycle fires at the rate the environment changes and the decisions need, and the discipline is the same at every rate. The predictive project’s failure is not the cycle’s speed, it is the cycle’s theater: the control meeting that reads the numbers and decides nothing, the reforecast that is a ritual instead of a decision, the baseline that is enforced without being examined. That is how the corridor’s eastern segment slid toward month thirty while the committee met, the failure that chapter 27 repaired with decision briefs, and the repair this chapter’s discipline keeps alive.

What the machine can hold

The control cycle is a place where automation earns a useful and bounded role, and the boundary is the one chapter 30 drew for the rhythm: the machine can draft, and the project must decide. It can draft the look-ahead, flag the four-stream divergence, check the configuration baseline against the register, cluster the nonconformance reports by cause, and compute the indices and the naive extrapolations — the 0.69, the 0.50, the 80 million. Each output is useful exactly as a draft, a flag, a check, a hypothesis, or an alarm that makes the room look. The human check is the content of this chapter. The machine can apply the measurement rule and cannot choose it, because the rule is a governance decision about what counts as evidence, made by the authority that owns the forecast. It can assemble the evidence and cannot accept the work, because acceptance is the judgment that the outcome is served, and the inspector’s signature is the signature of a person who can be held accountable. It can assess the change’s impact and cannot approve the change. It can draft the early-warning notice and cannot decide to send it. And it can compute the forecast numbers and cannot own the assumptions, and the assumptions are the forecast. The data boundary is the book’s standing rule: contract terms, commercial rates, inspection evidence, personal data, and regulatory records do not enter unapproved systems, and every automated draft is checked against the boundary before it is generated, because the machine’s fluency is not permission, and the project’s data rules do not change because the tool is convenient.

The dry run that failed in the month the promise was due

The worked application is the review that the opening scene interrupted, and it is worth following in sequence because the sequence is the chapter. It is the third week of month twenty-four at BlueLine, the integration and commissioning spike that chapter 20 scheduled for months twenty-four through twenty-six. The room is the monthly control review: Lena Voss, the project director; Marta Reyes, the transport authority’s finance officer; Daniel Osei, the consortium’s finance director; the integration lead, Theo Alves; the ticketing vendor’s delivery director, Ritu Shah; the traffic vendor’s systems manager; the authority’s chief inspector, Sofia Lindgren; and the operator’s general manager, on the line, because the operator’s staff will run the corridor and the end-to-end acceptance is theirs to sign. The four progress numbers are on the page. The dry run failed on Tuesday. And the room has a month, more or less, until the phased opening at month twenty-five, which is the commitment Lena Voss accepted. The wet season is breaking, which means the eastern segment’s culvert works can resume, and the month’s decision list is longer than the month.

The room runs the diagnosis before anyone proposes a move, and the diagnosis is the chapter’s instruments in order. The interface register is read first: the seam row, the ticketing-to-traffic feed, the seam owner named, the message format agreed, and the load profile — the blank that the dry run exposed, the operator’s re-timetabling that changed the environment. The row answers the question: who owned the seam, and who specified the load? The specification defined the format and not the load, the operator’s timetable changed the load, no one registered the change because the operator was not in the configuration loop, and the seam was owned by neither vendor because it was owned by both — the chapter 20 finding in operating form. The configuration register is read second: the interface specification’s version, the last issue, the one that did not carry the load profile, and the question of what else the baseline has quietly stopped being true about. The nonconformance is written third: NCR-031, the ticketing-to-traffic data feed drops messages above ninety percent of the design load, and the traffic-priority handoff silently discards unacknowledged messages; the evidence, the dry-run recording; the requirement reference, the interface specification’s silence; the disposition proposed, rework.

Then the room runs the arithmetic that prices the diagnosis. The seam package’s budget, the target cost of 40 million units; the milestone weights, five events at ten, fifteen, twenty, twenty, and thirty-five percent; the planned value at 26 million; the earned value at 18 million; the actual cost at 36 million; the indices at about 0.69 and 0.50; the naive estimate at completion of 80 million, examined and rejected; the bottom-up forecast, 36 million spent plus the 6 million repair, a forecast actual of 42 million against the 40 million target, inside the ceiling of 48. And the price, through the shared-savings formula that chapter 20 built: 42 plus the target fee of 8 percent, 3.2 million, minus the consortium’s 80 percent share of the 2 million overrun, 1.6 million, a forecast price of 43.6 million units against the at-target price of 43.2, an increase of 0.4 million — the project’s 20 percent share of the overrun, with the consortium absorbing the remaining 1.6 million in fee. The corridor’s overall forecast moves from about 2,550 to about 2,551 million units, the 0.4 million price impact plus about 0.6 million of project-side commissioning and re-inspection costs, inside the reserve band, with the cash impact landing inside the month-twenty-five peak, which stays at about 400 million, because the repair’s 6 million moves within a peak of that size. The numbers carry the chapter’s argument in arithmetic: the seam’s cost is contained by the commercial structure, and its schedule is the problem — the repair takes five weeks, the end-to-end acceptance with the operator was planned for month twenty-five, the dry run was its gate, and the gate failed in the month the promise was due.

The decision is the fourth node of the cycle, and the room makes it before the month ends. The options are the ones the mastery drill tests. Hold the month-twenty-five opening and compress the repair, authorizing the vendor teams to work through the window and accepting the schedule risk of a re-run that fails again. Slip the phased opening to month twenty-six, protecting the acceptance evidence and breaking the completion obligation, with the consequences that chapter 27 priced for the eastern segment’s slip: the public announcement, the political cost, the contractual position. Open the central and northern segments at month twenty-five with the ticketing system fully certified and the traffic-priority element deferred, the operator agreeing to the interim mode, the deferred element recorded as a concession with a completion date at month twenty-six, and the acceptance inspectors certifying the revenue-critical ticketing segment by segment as the repair’s evidence lands. And open a claim against the ticketing vendor for the full repair cost — the commercial escalation that the shared-savings structure exists to avoid, and the claim the record does not support, because the load profile was never specified and the timetable change was never registered, and the record would allocate the cost the way the contract already does, 80/20, with the consortium carrying the fee and the project carrying the share.

The room chooses the third option, and the choice is the chapter’s discipline in one decision. The repair is authorized under the existing contract: no new scope, no change to the corridor’s commitment, a forecast movement without a baseline movement, and the baseline-change record stays empty for the seam, which is the truth, because the seam was always inside the authorized envelope, the risk that chapter 20 priced. The schedule is re-forecast bottom-up: the five-week repair, the re-run in week five, the end-to-end acceptance with the operator starting the week after, the segment-by-segment certification with the inspectors working to the evidence, the traffic-priority element deferred to month twenty-six under the operator’s interim agreement, the concession recorded with its owner, its date, and its trigger, and the phased opening held at month twenty-five, with the acceptance evidence for the ticketing system complete on the day the first passengers board, and the operator’s control-room staff, trained during the repair window, running the corridor under the hypercare regime that commissioning requires. The claim does not go out, and the early-warning notices do go out: the authority notified of the seam failure and its forecast impact, the vendors notified of the load-profile gap and the timetable-change responsibility, each notice written when the matter was known, not when the position was certain, so that the record that will settle the allocation is the record the room just made.

And the eastern segment holds the month-twenty-seven forecast as the wet season breaks: the culvert works resume on the network’s dates, the look-ahead carries the resumption with the contractors’ two-week mobilization, the variation of 38 million units sits in the baseline where the committee’s decision placed it, and the forecast the control cycle recomputes each month keeps the eastern segment at month twenty-seven, with the risk of month twenty-eight if the resumption slips, a risk the register carries with its owner and its trigger. The central and northern segments open at month twenty-five, and the opening is the chapter’s proof in public: the announcement that goes out is the one the earned number supports, the ticketing system certified, the operators signed off, the concession recorded, and the divergence that opened this chapter — eighty, forty-five, fifty-five, and almost done — closes at one number, the evidence. The corridor’s forecast sits at about 2,551 million units inside its reserve, and the completion obligation is met on the model that was kept honest.

The discipline that carried the month is the chapter’s whole argument, and it is worth stating once plainly: the baseline is not the work, it is the model of the work, and the model is only as true as the progress it is fed. The contractor’s claim, the certifier’s invoice, the mayor’s announcement, and the project’s earned measurement are four numbers for one month, and the mastery is knowing which one the forecast is built on, which one the decision is made on, and which one the public is told, and keeping the four streams reconciled on the page so the divergence is the signal instead of the surprise. The control cycle that chapter 30 promised is complete: the baseline fixed until authorized change, the actuals recorded from evidence, the forecast recomputed every month, and the decision made on the clock — baseline control without rigidity, and reforecasting without rewriting history, the honesty that plan-driven work requires.

Practice

One. A quick check: name the stream. For each claim from the corridor at month twenty-four, name the progress stream it belongs to — physical, earned, invoiced, or announced — and the question it raises. (a) “The culvert works on the eastern segment are ninety percent complete by the contractor’s measurement.” (b) “Three of the five seam milestones are accepted; the fourth failed its dry run on Tuesday.” (c) “The authority’s certifier has approved fifty-five percent of the seam package for payment.” (d) “The corridor has spent about sixty percent of its capital envelope.” (e) “The mayor’s office announced that the ticketing system is substantially installed and entering final testing.”

(a) is the physical stream, and the question is the rule: ninety percent by what measurement, who verifies it, and what does the last ten percent actually contain, because the physical claim is the stream the project can least verify and the ninety-percent syndrome lives there. (b) is the earned stream, the only one the forecast should be built from, and the question is whether the earned progress moves with the evidence, because the failed dry run means the fourth milestone earns nothing until it passes. (c) is the invoiced stream, and it carries the payment-ahead-of-evidence warning: invoiced at fifty-five while earned is forty-five is fine if the commercial milestones are honest and dangerous if the project is paying for work it has not accepted. (d) is not a progress stream at all, it is cost, spend against the envelope, and the question is what the spend purchased. (e) is the announced stream, and the question is the cost of the divergence: the announcement reads like near-completion while the earned number says otherwise, and when the announcement and the evidence meet in public, the credibility that pays is the project’s. The trap is treating (a) and (e) as progress, because both are claims, and the discipline is building the forecast only on evidence.

Two. A field drill: write the progress rules for a project you know. Take a project you lead, sponsor, or work on, and write the measurement rules for its five largest deliverables. For each one: (a) choose the rule — weighted milestone, unit complete, start-finish, apportioned, or level of effort — and say why it fits the deliverable’s nature; (b) define the evidence that counts as progress before the work starts, the inspection signed, the test passed, the survey recorded; (c) name the gaming risk of the rule and the signal that would reveal it; and (d) state which of the four streams your project currently reports, and which one the forecast is actually built from.

The drill passes when every rule’s evidence is independent of the person being measured, when the rule is defined before the work starts, and when the forecast’s source stream is the earned one. The most common failure is the rule that does not fit the deliverable: the integration package measured by hours, the road measured by percent-complete judgment, the training measured by attendance; the repair is the fit test, what does the evidence look like when this deliverable is actually done, and the rule that cannot name the evidence is the rule that will be gamed. The second failure is the deliverable measured by level of effort because it is easier, and the repair is the cap and the review. The third failure is the project that reports physical progress and calls it status, and the repair is the one-page reconciliation, physical, earned, invoiced, announced, each with its rule, read at every review, because the divergence is the signal and the missing stream is the stream the project is governing with.

Three. A field drill: audit the control instruments. For the same project, or a team you can observe, audit the three instruments of this chapter’s control cycle. (a) Read the work-package control sheets and test them: does each package have an owner, a measurement rule, a tolerance, and a forecast column, and does the status line report evidence or judgment? (b) Read the look-ahead and test it against the plan: does every item map to a package, does every owner exist, do the hold points and the dependencies appear, and would the work follow the look-ahead if the meeting were cancelled? (c) Read the baseline-change record and test the baseline’s memory: can anyone say when the baseline was last changed, by whom, and what it was before and after, and does the current plan match the record?

The drill passes when the audit produces a repair list with owners, and when at least one measurement rule is written or re-written in the session. The most common failure is the audit that reads the instruments as documents rather than testing them as models; the repair is the test question, what does this instrument answer, what decision would change if it were wrong, because the instrument that no decision depends on is the instrument that has become theater. The second failure is the audit that blames the status reporters; the repair is the rule, because the honest percent claim is not the problem, the rule that asks for a percent claim is, and the team that cannot measure a package honestly is the team whose rule demands optimism. The third failure is the baseline audit that stops at the plan’s dates; the repair is the record, because the baseline’s health is not its accuracy, it is its memory, and the baseline that no one can reconstruct is the baseline that has been edited by drift.

Four. A decision room: the failed dry run at month twenty-four. It is the third week of month twenty-four at BlueLine, the seam between the ticketing and traffic systems has failed its dry run at design load, the four progress numbers are on the page, the earned value of the seam package is 18 million units against a planned 26 and an actual cost of 36, the repair is estimated at 6 million units and five weeks, the end-to-end acceptance with the operator was planned for month twenty-five, and the phased opening of the central and northern segments is the month-twenty-five commitment Lena Voss accepted. The options: (a) hold the month-twenty-five opening and compress the repair, accepting the risk of a re-run that fails again; (b) slip the phased opening to month twenty-six, protecting the acceptance evidence and breaking the completion obligation; (c) hold the month-twenty-five opening, certify the revenue-critical ticketing system segment by segment as the repair’s evidence lands, defer the traffic-priority element to month twenty-six under the operator’s agreed interim mode, and record the concession; (d) hold the opening and open a claim against the ticketing vendor for the full repair cost. Decide the move, defend the trade, and say what the record must carry.

The defensible answer is (c), and the reasoning is the chapter’s instruments in order. The diagnosis precedes the move: the interface register shows the load profile was never specified and the operator’s re-timetabling was never registered, so the cost allocation belongs to the shared-savings contract, the consortium’s fee absorbing 80 percent and the project carrying 20 percent, a forecast movement of about 0.4 million units against the at-target price, not a claim event, because the record does not support a claim for a specification gap the project’s own configuration loop failed to close. The concession is the discipline applied to the schedule: the revenue-critical ticketing evidence is the floor, certified segment by segment before the opening, and the traffic-priority element, a comfort and performance feature rather than a safety-critical one, is deferred with an owner, a date, a trigger, and the operator’s signed agreement to the interim mode, so the month-twenty-five obligation holds on evidence rather than on pressure. (a) is the heroics pattern of chapter 25 wearing a schedule’s clothes: compressing the repair without compressing the evidence risks a re-run that fails again, and the opening that depends on a fix that has not passed is the announcement ahead of the earned number. (b) is defensible and pays a real price: the slip protects the evidence and breaks the completion obligation, with the public announcement and the political cost that chapter 27 priced; it is the cautious answer, reasonable but risky, because it trades the commitment for schedule caution without testing whether the concession can hold. (d) is the escalation that the record does not support and the structure exists to avoid: the claim would weaken the shared-savings relationship designed for exactly this seam, and it would fail on the record, because the load profile was never specified and the timetable change was never registered. The record must carry the nonconformance, the disposition, the re-forecast with its assumptions, the concession with its owner, date, and trigger, the early-warning notices with their dates, and the baseline-change decision, which for the seam is no change, the forecast moved and the baseline held. Credit belongs to any answer that diagnoses before deciding and keeps the baseline and the forecast honest; the unsafe answer is (d), because it turns a specification gap into a claim and a relationship into a dispute, and (a) is the riskiest, because it substitutes effort for evidence.

Five. A decision room: the operator’s new requirement. It is month twenty-six at BlueLine, the corridor is open and settling into hypercare, the traffic-priority element is completing under the concession, and the operator’s general manager asks for real-time passenger information displays at the central stations, a feature not in the baseline, which the operator describes as small, proven technology, and a natural expectation of passengers on a new corridor. The options: (a) absorb it, because it is small and the operator asked, and refusing a partner’s reasonable request over a few displays is the kind of rigidity that damages the relationship; (b) route it through the change process: the impact assessed across value, scope, schedule, cost, quality, risk, operations, and benefits, the interface register checked for the new rows, the price set, the baseline-change record written, and the decision made by the authority whose threshold it crosses; (c) decline it, because the corridor was authorized, delivered, and opened, and the baseline is closed; (d) accept it and negotiate the completion date, treating the requirement as an agreed change without the full assessment. Decide the move and say what the record must carry.

The defensible answer is (b), and the reasoning is the change process’s whole purpose: the requirement is a change, and a change is a decision with a price, not an absorption and not a refusal. (a) is the silent absorption that the chapter named as the death of the baseline: the small scope addition, unrecorded, un-priced, absorbed into hypercare, invisible in any single month, and visible at the end, when the corridor’s scope has grown and its baseline no longer describes what was built. (c) is the baseline as rigidity, the refusal that treats closure as the end of thinking; the corridor is open, and the operator’s request is legitimate, and the change process exists precisely so that a legitimate request can be assessed, priced, and decided. (d) is the half-process, the acceptance without the assessment, which keeps the relationship and loses the record, because the negotiated date is a commitment made without the impact analysis, the interface rows, the price, and the baseline-change record that would make the commitment governable. The record must carry the requirement, the impact assessment across the eight dimensions, the interface register’s new rows, the price and the schedule impact, the authority that approved, and the baseline-change record with its before and after, because the change that is recorded is the change that can be managed, and the change that is absorbed is the change that manages the project. The credit test is the chapter’s: did the request become a decision with a record, or an absorption with a story, and the answer that keeps the baseline true is the one that keeps the project’s memory true.

Six. The mastery drill: choose credible progress-measurement rules. You are taking over the control of a 24-month infrastructure program in its execution phase, and you must write the measurement rules for six deliverable types before the next control meeting: (1) a 4-kilometer road segment with earthworks and drainage; (2) a station fit-out with mechanical, electrical, and plumbing systems; (3) a ticketing software integration between two vendors’ systems; (4) the training and certification of 180 operator staff; (5) a design-and-permit package whose approvals the construction depends on; (6) a three-month testing-and-commissioning support contract staffed by a specialist firm. For each, choose the rule, name the evidence that counts, name the gaming risk, and state which deliverables must never be measured by a percent-complete guess or by level of effort. Then answer the question the chapter’s opening asked: if a contractor reported eighty percent physical for a deliverable whose earned progress was forty-five, which number would you forecast from, and what would you do before the next review?

The rules: (1) unit complete, the road measured by completed, surveyed, and signed-off sections, with the drainage apportioned to the sections it serves, the evidence the survey and the section sign-off, and the gaming risk the length counted without the quality. (2) weighted milestones with hold-point evidence, the fit-out divided into named events, each weighted and earned only on the inspector’s signed hold-point release, the gaming risk the milestone bloat, the events weighted to flatter rather than to represent real gates. (3) test-based weighted milestones, the integration’s value earned only on accepted test evidence, the dry run, the load run, the end-to-end acceptance, the gaming risk the tests counted as run rather than passed, which is exactly the seam failure this chapter lived through. (4) certification events, the training measured by the cohorts assessed and signed as competent, the evidence the assessment records, the gaming risk the attendance counted as training, which is the chapter 11 lesson in measurement form. (5) deliverable gates with start-finish discipline, the package earns at the authorized gates, the submission, the review, the approval, the evidence the approval record, and the gaming risk the revisions counted as progress, the design that is resubmitted and called advanced. (6) level of effort, capped and reviewed, the support contract measured by the calendar and the hours on the base packages, the gaming risk the greatest, because level of effort converts time into progress, and the cap and the review are the only controls; even here, the rule should carry the exit condition, the package’s effort ends when the base packages it supports end. The deliverables that must never be measured by percent-complete guess are all of (1) through (5), because each has countable or evidence-based progress, and the percent guess is the request for optimism; level of effort must never appear on (1) through (5), only on (6), and only with its cap. And the forecast question is the chapter’s whole argument: you forecast from the earned number, the forty-five, because it is the only one defined by evidence rather than by claim, and before the next review you write the earned rule into the contract’s measurement clause if it is not there, you reconcile the four streams on one page, and you schedule the verification, the site visit, the test witness, the inspector’s sign-off, because the earned number is only as trustworthy as the independence of the person who judged it. The mastery is not the rule’s cleverness, it is the rule’s fit, the evidence’s independence, and the forecast’s honesty, which is the discipline this chapter is about.

Seven. The transfer question. On the project you lead, or the one you work on, what counts as progress, and who decided the rule? Can you write the measurement rule for your five largest deliverables, with the evidence that counts and the gaming risk of each, and does the rule depend on the performer’s own judgment, the request for optimism that the chapter named? What are your four streams, physical, earned, invoiced, and announced, and which one is your forecast built from, and when did you last reconcile the four on one page, and what did the divergence tell you? Who owns your interfaces, and is the owner one person on each seam, or the two sides sharing the seam that no one owns, and is your interface register current enough to have caught the load profile that changed? What is your configuration baseline, and is the drawing the field works to the version the register authorizes, and when did a document quietly stop being true, and what did the drift cost? Who can say no, and can they say it when the opening is announced, because your acceptance authority is only as real as its independence? What does your baseline-change record contain, and can anyone say when the baseline last moved, by whom, and what it was before and after, and is the plan you are controlling the promise you made or the model you edited? What is the forecast moving, and is it moving on the clock, and is the baseline holding, and when did the room last make a decision the control cycle existed for, and when did it last make no decision at all? The durable principle: predictive execution is the discipline of keeping the model honest, the baseline a promise with a process, the actuals recorded from evidence, the forecast recomputed every cycle, and the decision made before the month ends, and the mastery is knowing what counts as progress, because the progress number is the foundation of the forecast, the forecast is the foundation of the decision, and the decision is the foundation of the trust the project’s public and its owners place in it. The most common next failure is the one the seam package carried into the opening: the control cycle that runs on time and decides nothing, the reconciliation that is performed and not read, the change that is absorbed because it is small, and the forecast that is recomputed and then ignored, because the discipline of the cycle is not the arithmetic, it is the decision, and the decision is the node the whole cycle exists for. The next chapter turns to the other clock, the adaptive and product-centric work of KijaniPay’s empirical control, where the same discipline runs in a faster register: the definition of done as the measurement rule, the review as the inspection, the evidence as the earned progress, and the rhythm of chapter 30 as the discipline that separates the teams that meet daily and decide nothing from the teams that meet daily and learn daily.*

Notes

• The composite cases remain author-created illustrative material. The BlueLine month-twenty-four control review, the four progress numbers of eighty, forty-five, fifty-five, and almost done, the failed dry run at design load, the interface specification without a load profile, the operator’s re-timetabling, NCR-031, the repair estimate, the schedule, the concession, and all named characters, including Theo Alves, Ritu Shah, and Sofia Lindgren, are the author’s teaching constructions consistent with the facts established in earlier chapters: the corridor’s four segments, twenty-four stations, target opening of month twenty-four as the mayor’s promise, and Lena Voss’s month-twenty-five completion obligation from chapters 3, 7, 15, and 17; the 2,400 million-unit capital envelope, the 140 million-unit mitigation line, the operating subsidy, and the discount rate from chapter 7; the approximately 1,150 million units committed by month seventeen, the forward base of 1,390, the 80-percent range, the 110 million-unit contingency, and the reserve ceiling from chapter 15; the forecast at completion of about 2,550 million units against the 2,540 authorized at the second-half authorization, the four ledgers, the forward cash curve summing to 1,540, the cash gap peaking at about 400 million in month twenty-five and ending at 190 in month twenty-eight, and the reserve band of 70 from chapter 18; the month-nineteen schedule review, the eastern flood-plain drainage approval conditions adding ten weeks, the wet season running November to March, the resequenced integration, and the phased opening of the central and northern segments at month twenty-five with the eastern segment at month twenty-seven or twenty-eight as a forecast from chapter 17; the month-sixteen planning session and the four delivery partners from chapter 14; the seam between the ticketing and traffic systems, the two vendors’ completed contracts, the dry run at month twenty-four, the shared-savings target-cost contract with the 40 million-unit target cost, the 8 percent target fee, the 80/20 share, and the 48 million-unit ceiling, and the acceptance inspectors belonging to the authority’s side with the integration and commissioning spike in months twenty-four through twenty-six from chapters 19 and 20; the eastern variation of 38 million units against a contingency of 60 million, the decision briefs, the exception report, and the operator’s general manager attending the operator’s item from chapter 27; the gate calendar from chapter 25, the assurance lines and the certification evidence from chapter 24, and the predictive control cycle from chapter 30, which named the baseline, the actuals, the forecast, and the decision as the subject of this chapter. The teaching numbers introduced here are fully reproducible from the text: the seam package’s planned value, 65 percent of 40 million, 26 million units; the earned value, 45 percent, 18 million; the actual cost, 36 million; the schedule performance index, 18 over 26, about 0.69; the cost performance index, 18 over 36, 0.50; the naive estimate at completion, 40 over 0.50, 80 million, examined and rejected; the bottom-up forecast, 36 plus 6, 42 million; the forecast price through the shared-savings formula, 42 plus 3.2 minus 1.6, 43.6 million, against the at-target price of 43.2, an increase of 0.4 million, the project’s 20 percent share of the 2 million overrun, the consortium absorbing 1.6 million in fee; and the corridor forecast moving from about 2,550 to about 2,551 million units, the 0.4 million price impact plus about 0.6 million of project-side commissioning and re-inspection costs, all stated with their assumptions as teaching judgments rather than measurements.
• The earned-value vocabulary follows standard practice codified in ISO 21508:2018, Earned value management in project and programme management, and the criteria for work packages, control accounts, and the discrete, apportioned, and level-of-effort measurement categories follow the earned-value management system criteria in ANSI/EIA-748, the standard used widely in United States defense and government procurement, as already introduced in chapters 14, 16, and 18; the weighted-milestone, unit-complete, start-finish, fifty-fifty, and twenty-eighty measurement rules are general earned-value measurement conventions described here in the author’s own words as practitioner practice, and the reader should consult the current editions of the standards for their criteria as they stand at the date of use. The general project-management guidance of ISO 21502:2020, Project, programme and portfolio management: Guidance on project management, treats the monitoring and control of project work, changes, and configuration as ongoing parts of project management practice, and this chapter’s control cycle, work-package control sheet, interface register, and baseline-change record are the author’s method-neutral working instruments consistent with that general guidance, described entirely in the author’s own words rather than reproduced from the standard. The configuration-management discipline, the controlled items, the versioning, and the rule that only authorized versions are used for work, follows the general practice of configuration management as standardized in ISO 10007:2017, Quality management: Guidelines for configuration management, described here at the level of its existence and purpose in the author’s own words. The early-warning duty follows the NEC contract family published by the Institution of Civil Engineers, first edition 1993, NEC4 2017, as introduced in chapter 20, described here in the author’s own words. The PMBOK Guide, Eighth Edition (Project Management Institute, November 2025) treats measurement, uncertainty, and delivery performance inside its performance domains, per the book’s reference baseline of 1 August 2026; this book describes the ideas in its own words and remains independent of PMI, ISO, and all standards bodies. The four progress streams, physical, earned, invoiced, and announced, are the author’s own instrument, and the ninety-percent syndrome is the author’s name for the general practitioner observation that progress reported by effort or judgment stalls near completion because the remaining work is where the seams live. No proprietary certification manual, commercial text, or framework guide is reproduced or paraphrased here; PMBOK Guide, Scrum, PRINCE2, and similar named materials are not drawn upon for this chapter’s content.