Chapter 44: Realize Benefits and Evaluate What the Project Changed

The meeting the trigger called

The trigger fired in month thirty-six, and the committee did not argue about whether to meet. Chapter 7 had written the ridership trigger before the corridor opened: weekday trips more than ten percent below the 180,000-trip forecast band for two consecutive months. The counters at all twenty-four stations, calibrated at last during the opening gate, recorded month thirty-six and month thirty-seven at about 161,000 weekday trips, about 11 percent below the band. The trigger said the steering committee reconvenes, so it reconvened in month thirty-eight, in the transport authority’s boardroom — thirteen months after the phased opening of the central and northern segments at month twenty-five, ten months after the eastern segment’s opening at month twenty-eight, delivered on the rebased plan of chapter 42 with the re-run of the seams that chapter 41 had scheduled behind it. The risk that chapter 38 had carried publicly had arrived.

Abena was back. She had led the independent readiness review of the eastern segment in month twenty-six, the former transit program director who could not be removed by the people she reviewed, and the authority had asked her to lead the post-opening evaluation the way she had led the pre-opening review, with the same independence and the same question behind it: what is true, and who is telling it. She stood at the board with the numbers in three rows, because the first lesson of the evaluation is that the corridor-level number is not the story.

The first row was the corridor’s row, and it read nearly like the case. Weekday ridership at about 161,000 against the 180,000 forecast, the trigger crossed for the second month. Station-to-station journey time held at about six minutes of saving, the 12-minute crossing of the baseline down to about 6, the arithmetic of chapter 7 still true at the level where the corridor itself was measured. The automatic passenger counters, all twenty-four calibrated, the data-quality log of chapter 37 clean for the operating period. The operating subsidy, 90 million units a year, the number chapter 18 put in the same conversation as the 2,400 million-unit capital envelope, spent as budgeted. At the corridor level, the case was close.

The second row was the neighborhoods’ row, and it was not close at all. Abena had split the corridor’s catchment into six evaluation areas, three pairs, and the pairs did not agree. In the two feeder-served neighborhoods, buses to the stations every eight to ten minutes with an integrated fare, ridership ran about three percent above the area’s forecast share, and the door-to-door travel-time surveys showed a realized saving of nine to eleven minutes, because the old journey had been a slow feeder bus to a distant stop, and the corridor saved the feeder time too. In the two partial neighborhoods, feeder service every half hour, ridership about six percent below the area’s share and the realized saving three to four minutes. And in the two neighborhoods with no feeder service at all — no bus, poor walking access, scarce parking — ridership about 26 percent below the area’s forecast, the realized saving about one minute, and the cars never left. The corridor saved six minutes between the stations. The question was how people reached them, and the answer decided how much of the six minutes they ever saw.

The third row was the evaluation’s row, the rows the case had counted and the field was now paying for, and Abena read them with the source beside every number. Modal shift: the model’s 70/30 split that chapter 9 had flagged against the field’s 55/45 — the corridor had pulled about 55 percent of its trips from cars, not 70, and the shortfall lived in the neighborhoods where the car was still the only reliable way to the corridor. Emissions: the case’s reduction target met on the corridor itself, and eaten back by the park-and-ride traffic, about 4,100 cars a day at the three outlying stations, the vehicle-kilometres moved from the corridor to the feeder roads. Crashes: down on the corridor, and up in the conflicts at three station forecourts where the mixed traffic met the footways. Shop viability: two of the three trading estates recovered to pre-construction revenue; the third, beside the no-feeder neighborhood, still below, with closures continuing. Access for carless households, Grace Njoroge’s row: 52 percent of carless households within a safe ten-minute walk of a station, against the case’s 65 percent target, and 31 percent in the no-feeder neighborhoods.

The room sat with the three rows. The corridor was on budget and on its operating subsidy. The value the project had been built to create was arriving, but it was arriving in the neighborhoods that were already served, and it was not arriving in the neighborhoods the case had promised it to — and the promise was the reason the corridor existed. Councillor Diana Kamau, who had spent the construction years carrying her constituents’ trust from one phase to the next, asked the question the third row made inevitable. “The trigger has called us here. What do we do with it?”

That question is this chapter. The corridor had done the hard part of chapter 43: the transfer, the acceptance, the closure, the records, the lessons. The value question had only just begun, because the certificate is the record of a transfer and the benefit is the proof of the project, and the two are different instruments with different clocks. Delivered is not realized. The output the project hands over is a promise wearing a deliverable’s clothes until the behavior changes and the outcome appears. This chapter is the discipline of the space between the certificate and the value: the benefit profile that makes the promise accountable, the clock that keeps it honest, the curve that shows it rising or eroding, the evaluation that finds out what the project actually changed, the correction that closes the gap, and the evidence that goes back into the system.

Delivered is not realized

The first lesson is the distance, and the distance is easy to underestimate because it is invisible at the moment of closure. At the moment the corridor handed the operation its acceptance certificate, everything looked finished: the stations open, the services running, the records archived, the team released, the closure report signed. The benefit, on that day, existed only in the arithmetic of the business case: 180,000 trips a day at six minutes a trip at 40 units an hour, about 216 million units a year of travel-time value. That arithmetic is 18,000 hours a day, 5.4 million hours a year over 300 operating days, 216 million at 40 units an hour. It was true the day it was written, and it will be true in month forty, and it is true the way a recipe is true: it describes a meal that has not been cooked. The trips must happen. The behavior must change. The value must be realized.

The output-to-benefit chain from chapter 1 and the benefit dependency network of chapter 7 did the structural work here, and the chain deserves its readback because the failure sits in its middle. The project produces outputs: the corridor, the stations, the ticketing, the signal priority, the trader-support program. The outputs enable capabilities: a frequent, reliable bus service; an integrated fare; access routes that exist. The capabilities produce outcomes only when behavior changes: commuters shift from cars, traders use the delivery windows, residents walk to the stations. And the outcomes become benefits only when someone values them: time saved, emissions avoided, crashes prevented, access gained, shops kept viable. The chain breaks at the behavior link, and nothing downstream of the break can compensate. A corridor that nobody rides is not a deliverable that fell short; it is a cost with no benefit attached — the capital, the subsidy, the land, the disruption, the trust spent.

Chapter 2 built the four levels of success, and they are the vocabulary for the distance. Delivery success is the corridor delivered, accepted, and closed, the chapter 43 achievement, real and necessary and insufficient. Product success is the corridor working as designed: the trains on time, the stations clean, the counters counting. Adoption success is the behavior: the trips taken, the fares integrated, the access used, the behavior that chapter 11 made the unit of delivery. And benefit success is the value, the time saved, the emissions avoided, the access realized, the value the business case was built on. The four can disagree, and the disagreement is the project’s real status report. At BlueLine, delivery success and product success were complete, and adoption and benefit success were partial — and the partial was not a failure of the build. It was the normal state of a promise that still had work to do. The project that reports delivery success as if it were benefit success has made the book’s oldest category error: the confusion that the completed artifact is the achieved value.

The distance has a clock, and the clock deserves a name because it changes the question. Call it the benefit clock: the time that runs from handover to the moment the promised value appears, or the case is revised. Every benefit row carries a clock, and the clocks have different lengths, which is why the benefit cannot be judged at a single date. The travel-time benefit at BlueLine had a clock of months: the first operating year was the window in which the ridership would either approach the 180,000 band or reveal that it would not. The emissions benefit had a clock of years: the annual reduction can only be read against the baseline after a full year of operation. The access benefit had a clock that began the day the first station opened and will never fully close, because access is a standing condition, not an event. And the economic benefits — the shop viability, the neighborhood economy — had the longest clocks, the ones that make the shortest budgets nervous. The benefit clock is why the evaluation has a calendar, why the trigger is written before the data arrives, and why the organization that evaluates benefits only once has judged the whole project on the first month of a ten-year curve.

The close of chapter 43 handed this chapter its subject in a sentence worth keeping: the certificate records the transfer, and the transfer was made on evidence. The benefit is the same discipline on a longer clock. The acceptance certificate said the corridor could be operated at the agreed performance. The benefit evaluation asks whether it was operated, by whom, for whom, and what changed. The two questions are sequential, and the project that closes without planning the second has planned the ceremony and not the value.

The one-page benefit profile

The minimum viable instrument is the benefit profile, one page per benefit row, and the one-page constraint is the discipline: a benefit that cannot be written on one page has not been thought through, and a benefit that has not been thought through will be defended by hope. The profile is the benefit row of the business case turned into an accountability, and it has seven lines.

The owner. Every benefit row has a named person accountable for its realization — the operational owner with authority over the levers, transferred at the gate in chapter 37, not the project manager who built the thing and not the analyst who counted it. At BlueLine the ridership row’s owner was the operator’s commercial director, the person who could change the timetable, the fare, the marketing, the feeder contract. The access row’s owner was the authority’s head of inclusive mobility, the seat Grace Njoroge’s federation had been arguing with since the stairs render of chapter 9, now with the measure beside it. The shop-viability row’s owner was the displacement-program lead, the person who had carried the 1,400 shops through the relocation. Ownership obeys the rule chapter 8 built for every accountability: the owner must be able to affect the outcome, and the owner must be answerable for it, and both conditions are checked at the transfer, not discovered at the review. The owner who owns the number but not the lever is the failure pattern wearing an org chart.

The baseline. The dated snapshot of the world before the change: the 12-minute crossing, the 50,000 vehicle-trips a day on the corridor roads, the modal share, the crash counts, the shop revenues, the share of carless households within a safe walk. Chapter 7 wrote the baseline because a case without a baseline is a wish, and the evaluation rediscovers why: without the dated before, the realized value cannot be measured, and the benefit row becomes a claim with no past. A baseline captured after the opening is not a baseline; it is a reconstruction, and the reconstruction will be flattered.

The target. The measurable advantage the benefit was promised to deliver, in units with the counting rule attached: 180,000 weekday trips, six minutes saved, 3,400 tonnes of CO2-equivalent avoided a year, 12 fewer serious crashes a year, 65 percent of carless households within a safe ten-minute walk, supported shops back to pre-construction revenue within 12 months. The target inherits the counting discipline of chapter 37 — what this number counts, measured how, on what evidence, excluding what — because the corridor’s four progress numbers taught the whole book that a number without a counting rule is a statement of faith.

The dependency chain. The benefit’s chain from chapter 7 in one line, outputs to capabilities to behavior to outcome to benefit, because the chain is the diagnostic map when the benefit fails. The travel-time row’s chain ran from the corridor and the feeder contracts to the frequent service and the integrated fare to the mode shift to the shorter journeys to the time value. When the row failed in the no-feeder neighborhoods, the chain showed where: the capability existed, the behavior did not, because the feeder link was missing. The chain is the row’s theory of change, and the evaluation’s first act is to test the theory at each link, not to stare at the gap at the end.

The timing. When the benefit is expected to appear, with the shape of the curve — the adoption delay, the rise, the plateau — and the review dates at which the row is judged. The ridership row reviewed monthly against the band with the trigger; the emissions row annually against the baseline; the shop-viability row quarterly; the access row on a standing cadence. The timing is a commitment, because a benefit with no timing cannot be judged, and an evaluation with no timing will be postponed until it is too late to act.

The assumptions and the risks. What would have to be true for the benefit to appear, and what would make it fail: the modal-shift split, the feeder availability, the fare level, the fuel price, the employment base. The assumptions are not decoration. They are the sensitivity table of chapter 7 in profile form, and the evaluation’s honest claim will be judged on how well the profile anticipated what the field showed.

The review cadence. Who reviews the row, when, on what evidence, and what the trigger is — the reconvening clause that chapter 7 wrote and the corridor’s month-thirty-six meeting executed. The cadence is the row’s governance, the connection between the benefit and the decision. A profile that ends at the target without the cadence has built a dashboard with no one obligated to read it.

One page, seven lines. The sophisticated variant is the benefit register, the table of profiles across a project or program, which chapter 45 will need when the benefits span several projects; the register is the profile’s aggregation, not its replacement.

The failure modes are the ones the corridor’s review was written to catch. The profile written after the benefit failed, when the baseline has to be reconstructed. The owner who owns the number but not the lever. The target that moved with the meter: the ridership band redefined after the miss, the access target softened after the audit. The measure that counts the training instead of the behavior, the 95-percent-trained 35-percent-adopted gap of chapter 11 restated as a benefit. And the row that was archived with the team, the profile the closure report listed and the operation never received.

The clock and the curve

The benefit realization curve is the primary visual of this chapter, and it is worth drawing because the drawing is the argument. The horizontal axis is time from handover; the vertical axis is benefit realized as a share of target. The curve has three parts. The first is the adoption delay: for the first weeks and months the realized benefit is low, often below zero on the disbenefit side, because the new behavior is expensive, the staff are learning, the users are changing habits — the implementation dip that chapter 11 named through Michael Fullan’s work. Performance declines before it rises, and the decline is not failure; it is the shape of change. The second part is the rise, the adoption curve climbing as the behavior takes hold, the ridership building, the trips shifting, the outcomes appearing. The third is the fork: the curve either sustains, flattening toward the target and holding because the reinforcement continues, or it erodes, rising partway and falling back because the reinforcement stopped — the service cut, the wayfinding decayed, the marketing ended, the champions left — and the behavior reverted toward the old equilibrium.

Figure 44.1: The benefit realization curve. Time from handover on the horizontal axis, benefit realized as a share of target on the vertical axis; the adoption delay and the dip before the rise, the climbing curve, and the fork into the sustained path where reinforcement continues and the eroding path where it stops, with the intervention markers placed where corrections lift the curve. The corridor’s three neighborhoods are three curves wearing one corridor.

The curve teaches three things that a single benefit number hides. First, the timing: a benefit judged too early is judged during the dip. The corridor’s ridership in its first operating month, below the band, would have misled a committee that evaluated then; the month-thirty-six trigger, eleven months in, was not too early, because the dip had passed and the curve had shown its shape. The evaluation date is a decision, made on the curve’s logic, not on the calendar’s convenience. Second, the erosion: the curve that rises and then falls is the silent benefit failure, the one no trigger catches, because the trigger was written for the rise, and the fall happens after the reviews stopped, the dashboard archived, the evaluation budget spent. That is why the realization dashboard has a standing cadence and the reinforcement is a budgeted line, not a hope. Third, the intervention: the curve is not a weather report, it is a control chart, and the intervention markers sit on it — the correction that lifts the curve, the feeder program, the fare integration, the access works, the marketing — each an owned action with a budget and a date, placed where the evidence says the gap is.

The corridor’s curve at month thirty-eight showed the pattern in full: the adoption delay in the first months, the central and northern segments climbing, the eastern segment joining late, and then the fork — the feeder-served neighborhoods continuing to climb toward their forecast, the partial neighborhoods flattening below theirs, the no-feeder neighborhoods flattening early and low. The curve split into three, which is the shape the corridor-level number had hidden. The 161,000 against 180,000 was not one gap; it was three gaps wearing a single number, and the correction was not one action, it was three.

The numbers on the curve carry the chapter’s arithmetic, and the arithmetic is worth doing once. At the forecast, 180,000 trips at six minutes at 40 units an hour is 216 million units a year, the working shown above. Realized at 161,000 trips, the same arithmetic gives about 193 million units a year — a gap of about 23 million units on the travel-time row alone, before the emissions, access, and shop-viability rows are counted. The gap is not a rounding error; it is the corridor’s annual interest on the promise, and it is the number the correction must beat. The arithmetic is simple because it is designed to be simple, and the simplicity is the point: a benefit claim that cannot be reproduced in one line will be defended in a courtroom instead of a review.

The clock and the curve wear the delivery styles differently, and the difference decides where the benefit review lives. At BlueLine, the register is predictive and the benefit clock is a formal calendar: the operating year, the annual evaluation, the gate-style review the trigger can call early. At KijaniPay, the register is adaptive and the clock is continuous: the benefit is the product’s outcome metric, read in the product loop of chapter 32, tested by the experiments that are the adaptive register’s contribution analysis. At Meridian, the register is hybrid and the clock is the grant’s: the access outcomes measured on the wave cadence, the grant window at month thirty-six the evaluation’s deadline. At Northstar, the register is compressed and the clock runs during the project: the interim benefit counted while the response is still running, with the sustained-outcome question for the evaluation that follows. Four registers, four clocks, one discipline: the benefit is measured on the cadence the delivery style already set, and the evaluation is designed on the same clock, not on the anniversary of the close.

What moves first

The realization curve has an instrument problem the evaluation must solve before it can read the curve: the benefits are lagging, and the project needs leading, and the leading measures are adoption measures. Chapter 11 taught the pair at the point of work — the 95-percent-trained and the 35-percent-adopted, the leading indicator that moves first and the lagging outcome that follows — and chapter 37 built the metric families for the corridor: ridership against the band, journey time against the baseline, access findings open against closed, incidents, the counters’ calibration, the operator’s readiness items. The benefit evaluation extends the same discipline to the benefit horizon. The leading measures are the adoption behaviors: the trips, the fare integrations, the station accesses, the wayfinding use. The lagging measures are the outcomes: the journey times, the emissions, the crashes, the shop revenues, the access shares.

The leading measures matter because they are the early warning, and the early warning is the only intervention that is cheap. At the corridor, the ridership counter gave the signal in month thirty-six, eleven months into operation, while the journey-time surveys and the emissions baseline would not have spoken for another year. The signal was late relative to the first operating month, when the no-feeder neighborhoods’ low boarding counts were already visible on the counters. The lesson is the lesson of chapter 41 wearing a benefit’s clothes: the project read the counters as construction data, the calibration item, instead of as benefit evidence, the forecast of the fork. The organization that measures the lagging outcome and not the leading behavior will learn what happened a year after it could have done something about it.

The RE-AIM lens that chapter 11 borrowed from Russell Glasgow and colleagues — reach, effectiveness, adoption, implementation, maintenance — extends naturally to the evaluation, and the corridor’s rows show why the last three terms are the ones that matter at realization time. Effectiveness, the time saving measured where the pilot worked, was proven in the feeder-served neighborhoods, and the evaluation had to resist generalizing it. Adoption, whether the people who should use the corridor actually do, was the row the counters carried, and it split by neighborhood. Implementation, whether the corridor was delivered as designed, was the eastern segment’s story: the seams, the re-run, the month-twenty-eight opening. And maintenance, whether the benefits hold, is the curve’s fork — the reinforcement that must be budgeted after the adoption is achieved. The evaluation that measures only the first two terms has evaluated the corridor’s promise, not its practice.

The instrument for the pair is the realization dashboard, and it inherits the minimal dashboard discipline of chapter 37: six numbers, each with an owner, a source, a refresh cadence, a threshold, and a decision attached. The corridor’s dashboard at month thirty-eight carried the six rows that mattered: weekday ridership against the band, with the trigger that had fired and the decision the meeting was making; realized door-to-door journey-time saving by neighborhood cluster, the three-part row that showed the fork; the park-and-ride car counts at the three outlying stations, the disbenefit that was eating the emissions row; the forecourt conflict counts at the three stations, the safety row that had moved in the wrong direction; shop-viability revenue recovery in the three trading estates; and the carless-access share by neighborhood, Grace Njoroge’s row. Six numbers, six owners, six sources, six thresholds, six decisions. The discipline is the discipline of chapter 37: the number without the decision is decoration, and the row that nobody owns will be the row that fails.

The disbenefits nobody bids for

The benefit profile has a second page, and the second page is the disbenefit profile: the costs the project imposes on people who did not ask for them, and the value that moves from one pocket to another. Chapter 7 built the disbenefit branches into the benefit dependency network, the displacement and the subsidy sitting beside the chain, and the corridor’s operating years have made the branches real.

The subsidy is the cleanest disbenefit: 90 million units a year, the number chapter 18 put in the same conversation as the 2,400 million-unit capital envelope and chapter 36 carried into the service transition. The evaluation’s job is not to mourn it; it is to measure the value it buys. A subsidy is a purchase, and the purchase has a price per unit of realized benefit. The corridor’s evaluation was the first time anyone asked what the 90 million bought in realized travel-time value, emissions avoided, and access gained, and whether the neighborhoods that pay the subsidy are the neighborhoods that receive the benefit. At month thirty-eight the subsidy was paying for a service whose benefits landed disproportionately in the feeder-served neighborhoods, and that distribution, not the subsidy’s size, was the finding.

The displacement is the disbenefit with the longest memory. The 1,400 shops survived the relocation of chapter 9 — the staged moves, the footfall-mitigation measures, the association’s co-designed plan — and the opening brought the corridor’s customers to the estates. The revenue recovery is real and uneven: two of the three trading estates above pre-construction revenue, the third, beside the no-feeder neighborhood, still below, with closures continuing. The displacement-support program was measured against shop viability, the row Councillor Kamau’s constituents had carried since the petition of chapter 9, and the evaluation’s finding is that the support program’s outcomes track the corridor’s benefit geography: the estates that gained the corridor’s access gained the customers, and the estate that could not reach the corridor kept the loss. The disbenefit row has an owner, a baseline, a target, and a correction, like any benefit, because the displacement that is not owned is the debt that becomes a political crisis at the next budget cycle.

The unintended consequences are the rows nobody wrote, and the evaluation’s obligation is to find them before the newspapers do. The park-and-ride traffic, about 4,100 cars a day at the three outlying stations, moved the vehicle-kilometres from the corridor to the feeder roads and the surface streets: the emissions gain eaten back at the margin, the congestion moved, not removed. The forecourt conflicts at three stations, where the mixed traffic meets the footways: crash counts down on the corridor and up in the interfaces, the safety benefit and the safety cost sharing a station. The crowding at the feeder-served stations, the success’s own disbenefit: the buses full, the platforms busy, the wait longer than the timetable promised, the benefit realized so well in one neighborhood that it degrades itself. And the second-order displacement: the rents and values shifting, the neighborhood that gains the corridor’s access becoming less affordable to the people the access was built to serve — the gentrification that chapter 48 will meet again under sustainability and equity.

The rule of the disbenefit profile is the rule of the benefit profile: an owner, a baseline, a target, a measure, a timing, and a correction. The disbenefit that is measured is a problem to be managed; the disbenefit that is not is a crisis to be discovered. The evaluation’s honest claim is not the corridor’s net benefit; it is the distribution — who gained, who paid, who bore the risk, and who had a voice, the four questions that chapter 48 will make the book’s sustainability lens. The corridor’s month-thirty-eight meeting was the first time the four questions had a room.

Who keeps the clock after the closure

The benefit clock has an owner problem that the project creates and the evaluation inherits: the project that built the thing is gone — the team released in chapter 43, the budget closed, the dashboard archived — and the clock is still running, and somebody must keep it. The discipline of chapter 37 was to build the measurement system that survives the project: percent complete archived, the operating measures taking their place at the opening gate. The discipline of this chapter is the same discipline on a longer clock. The realization dashboard is not the project’s dashboard with a new name. It is the operation’s dashboard, funded by the operation or by the evaluation budget, owned by the operational owners whose names sit on the benefit profiles, and reviewed on the cadence the profiles set.

The funding is the quietest failure, and it deserves its own decision. The evaluation and the realization measurement are part of the project’s cost, and the honest business case of chapter 7 writes the evaluation budget into the case at authorization, not after the closure when the money is gone. At the corridor, the counters were a capital item, funded and installed; the journey-time surveys, the modal-shift counts, the shop-revenue data, the carless-access audits were operating items, and the operating budget did not carry them, because nobody had written the evaluation budget into the second-half authorization. The corridor got lucky: the authority’s evaluation unit had a standing line, and the month-thirty-six trigger gave the committee the authority to spend it. The organization with no evaluation line and no trigger has no way to learn what its projects changed — and the learning is not a luxury, it is the only thing that makes the next business case better than the last.

The failure pattern is the dashboard that stopped updating, and it deserves its description because it is the most common end of a benefit program. The dashboard is built at the opening gate with six rows and six owners. The first reviews happen, the operator reports, the committee reads. Then the cadence loosens: the refresh slips from monthly to quarterly to when-someone-remembers; the owner changes roles and the row changes hands; the source disappears with the vendor contract; the threshold is renegotiated after the miss. The dashboard becomes the artifact chapter 37 named the dashboard of convenience, the number that always lands in target because the target moved. The tell is the meeting that stops asking questions: the review that reads the numbers and moves on, the row nobody challenges, the trigger that fires and goes unanswered. The chapter 41 detection discipline applies to the benefit register exactly as it applied to the project: the register that stops changing is the register that has stopped being true.

The minimum viable practice is the standing review, and its shape is the shape of the chapter 43 close with a longer clock: the realization review on the profile cadence, the six rows on the page with their owners in the room, the threshold crossed, the decision made, the record kept. The corridor’s month-thirty-eight meeting was such a review, called by a trigger written in the business case of chapter 7 before the corridor was half built. The trigger worked because it was specific, checkable, and written before the pressure arrived — the same properties chapter 23 taught for the contingency trigger and chapter 43 taught for the rollback criteria. The benefit trigger is the rollback criteria of the value: the condition, the decider, the restored state, and the window, applied to the promise.

An evaluation with a plan

The realization dashboard tells the committee that the value is short. The evaluation tells it why, and for whom, and what to do. The two are different instruments, and the confusion between them is the source of the evaluation’s bad reputation. Monitoring is the standing measurement: the counters, the dashboard, the cadence. Evaluation is the periodic deep question: the study that asks what changed, at what cost, for whom, and whether the change would have happened anyway. Monitoring is continuous and cheap and shallow by design; evaluation is periodic and expensive and deep by design. The project that substitutes one for the other has either bought a study that will not be read or run a dashboard that will not be believed.

The evaluation plan is the one-page instrument that turns the deep question into a commitment, and it has six lines. The purpose: why the evaluation is being done. The corridor’s purpose was written by the trigger: to decide what to do about the benefit gap, with the city’s next investment decision, the feeder program, waiting on the answer. The questions: the five that every post-project evaluation must answer in its own terms — did the intended change happen, for whom did it happen, at what cost, what would have happened without the project, and what explains the gap — plus the sixth, what do we do now. The methods: how the questions will be answered. The corridor’s methods were the comparison neighborhoods, the dose-response across the twenty-four stations, the door-to-door journey-time surveys, the counter data, the shop-revenue records, the crash and conflict counts, and the interviews with the people closest to the change: the operators, the traders, the residents, the advocates. The timing: when the evaluation happens and how long it takes — the first read at the corridor’s first operating year, the deeper read at the second, the benefit clocks dictating the calendar. The budget: what the evaluation costs and who pays — the evaluation unit’s line, the surveys, the analysis, the review, written into the case. And the independence: who can say no, and who can remove them if they do, the chapter 24 and chapter 41 test applied to the evaluation. The operator has an interest in the ridership row, the consortium in the benefit claim, the city in the subsidy, and the evaluation that cannot be removed by the people it reviews is the only evaluation that will be believed by the people it serves.

The evaluation has a machine’s help and a human’s accountability, and the pair deserves its boundary, because the evaluation’s product — the contribution claim — is exactly the kind of fluent output a machine produces convincingly and cannot verify. The machine’s legitimate work is the aggregation of the counter data, the anomaly detection that flags the counter that drifted, the summarization of the survey responses, the clustering of the interview notes, the drafting of the findings narrative from evidence the evaluators provide, the scenario generation that tests the dose-response assumptions. Each use carries the discipline of chapter 40: the source data named, the sensitivity handled, the output treated as a draft, a hypothesis, a signal, until a named person verifies it against the evidence, with the verification record that says what was generated, from what, checked by whom, and decided by whom. And the boundary is absolute where the chapter’s subject lives: the contribution claim, the range, the caveat, the recommendation, and the signature are human decisions, because the machine cannot hold an interest, cannot be cross-examined, and cannot be removed. The corridor’s evaluation used the machine for the aggregation and the drafting and the anomaly detection, and every number on Abena’s board carried the verification record beside it — the provenance of chapter 40 applied to the evaluation, because the benefit claim that cannot be traced to its evidence is the claim that will not survive the council chamber.

Abena’s evaluation was the same instrument the readiness review had been: the independence, the evidence standard, the triangulation of chapter 41 — what people say, what the documents show, and what the field shows, with the disagreement as the diagnosis. The evaluation found the disagreement in the no-feeder neighborhoods: the operator’s ridership data said the trips were not coming, the trader interviews said the customers were not coming, and the field walk said why — the buses did not come, the walk was long and unsafe, the parking was scarce, the corridor was six minutes away from a neighborhood that could not reach it.

The post-program evaluation is the larger form of the same instrument, and it belongs in this chapter because the corridor is the boundary case: the benefits span the segments, the feeder network, the displacement program, the subsidy, and the question of what the whole corridor program changed is a program question. Chapter 45 will build the program governance that carries it. The distinction matters at the design stage, because the evaluation’s questions decide its methods: the project evaluation asks whether the project’s outputs produced the intended outcomes; the program evaluation asks whether the coordinated change — the projects together, the benefits sequenced, the operations absorbing the result — produced the value the strategy promised. Abena’s plan named the boundary: the evaluation would answer the project question, whether the corridor delivered its promised time, access, and safety, and it would flag the program question, whether the corridor plus the feeder network plus the displacement program plus the subsidy regime was worth what it cost, and the flag would go to the council already planning the next corridor.

Contribution, not attribution

The evaluation’s hardest question is the one the corridor’s numbers cannot answer by themselves: would the ridership have been different without the corridor, and how much of the change that happened would have happened anyway. The before-and-after comparison is the seductive answer and the wrong one, because the world moved while the corridor was built. The fuel price rose in the second operating year, which pulls people toward transit and inflates the corridor’s apparent success. A major employer relocated its headquarters to the feeder-served side of the city, which moved 4,000 commuters and their trips with it. The city opened a competing park-and-ride on the eastern edge, which drains the eastern stations. The before-and-after number — the crossing that fell from 12 minutes to 6 — credits the corridor with every minute of the fall and every trip of the rise, and the credit is false. That is not the corridor’s fault and not the evaluation’s either. It is the counterfactual’s: the world that did not happen, the corridor’s twin without the corridor, and the twin does not exist.

The discipline of contribution analysis, named to John Mayne’s work in the notes, is the honest substitute for the impossible claim, and its structure is the structure of every defensible benefit evaluation. The claim is not that the project caused the change. The claim is that the project contributed to it, in a specific way, supported by evidence of the chain, and bounded by the plausible alternatives. The corridor’s contribution claim is built in three steps. First, the chain: the outputs exist, the corridor, the stations, the service; the capabilities exist, the frequent reliable service, the integrated fare, the access routes; and the behaviors the chain depends on are observed in the field — the trips, the mode shift, the access use — with the counts and the surveys as evidence. Second, the alternatives: the evaluation tests what else could explain the observed change — the fuel price, the employer relocation, the competing park-and-ride, the city’s general modal shift — and it builds the comparison that controls for them: the neighborhoods without the corridor’s access, the corridors without the feeder service, the stations with and without the integrated fare. Third, the bounded claim: the evaluation states what the corridor contributed, in a range, with the caveats named, in the form the committee can defend: a reduction in corridor journey time of about six minutes, measured against the baseline and sustained in the operating data, with the door-to-door benefit varying by neighborhood from about one to about eleven minutes depending on the feeder access; a share of the observed ridership, estimated against the comparison areas at between 80 and 90 percent of the forecast band’s trips, with the balance explained by external factors; and the emissions and crash rows stated the same way, with the range and the caveat.

The dose-response design was the corridor’s most powerful instrument, and it deserves its place in the method because it uses the project’s own geography as the experiment. The twenty-four stations split naturally into the feeder-served and the not, the integrated-fare and the not, the accessible and the not, and the ridership and the realized savings vary with the treatment: the stations with the feeder service and the integrated fare realizing the full saving, the stations without realizing a fraction, the dose matching the response. The dose-response evidence is stronger than the before-and-after because it compares like with like inside the same city — the same corridor, the same weather, the same fuel price, the same employer relocation — and it attributes to the treatment only the difference the treatment explains. The corridor’s finding was the dose-response’s finding: the realized benefit tracks the feeder access, not the corridor, and the corridor’s contribution to the travel-time row is a function of the network that carries people to it. That is exactly the finding the correction needs, because it names the lever.

The reverse errors deserve their names because the evaluation must resist them on both sides. The first is the credit error: claiming the corridor caused the ridership that the employer relocation produced, the before-and-after claim that flatters the project and misleads the next investment. The second is the blame error: accepting that the no-feeder neighborhoods’ low ridership is the corridor’s failure, when the corridor was never the constraint — the feeder network was — and the corridor’s own rows, the time, the service, the safety, were delivered as promised. The blame error is the subtler of the two because it feels like accountability, and the evaluation that accepts the blame will misallocate the correction, funding the corridor’s fixes that will not move the neighborhoods’ numbers, when the correction belongs to the network. The honest claim says both halves: the corridor delivered its promised contribution, and the benefit gap lives in the rows the corridor does not own.

The bounded claim has a discipline the reader can take away: the evaluation never says the project caused the benefit. It says the project contributed the benefit within a stated range, on stated evidence, against stated alternatives, and the claim is written so that the reader can see what would change it. The claim is a hypothesis with its test attached, and the test is why the evidence standard matters: every finding carries its source, its counting rule, and its caveat.

Value erosion and the correction

The evaluation’s finding is not the end of the work; it is the beginning of the correction, and the correction is where the benefit clock either gets wound or stops. The curve’s fork is the choice, and the corridor’s month-thirty-eight meeting sat exactly at it, with the trigger fired and the evidence on the board. The options were the options every realization review faces: reinforce, revise, or stop.

The reinforce option is the correction, and the correction is an owned intervention with a budget, a date, a trigger, and an owner, placed on the curve where the evidence says the gap is. The corridor’s correction was the feeder improvement program, and its arithmetic is the chapter’s numbers-that-matter, worth working because the decision turned on it. The gap is about 19,000 weekday trips, the 180,000 forecast against the 161,000 realized. The travel-time value of the gap follows from the case’s own arithmetic: 19,000 trips a day at six minutes is 1,900 hours a day; 570,000 hours a year over 300 operating days; at 40 units an hour, about 23 million units a year. The feeder program — connecting the two no-feeder neighborhoods with a fifteen-minute service and lifting the partial neighborhoods to twenty — cost about 40 million units of capital, new vehicles, stops, the integrated-fare extension, and about 8 million units a year of operating cost. The evaluation’s estimate, from the dose-response, was that the program would recover about 15,000 of the missing trips a day, about 80 percent of the gap, worth about 18 million units a year on the travel-time row alone, before the emissions, access, and shop-viability rows were counted — a net of about 10 million units a year against the 8 million operating cost, with the capital of 40 million paying back in about four years on the travel-time row alone, positive at the corridor’s 5 percent discount rate over the operating horizon. The arithmetic is simple because it is the case’s own arithmetic — the same 40 units an hour, the same 300 days, the same six minutes — and the correction’s claim is judged by the same reproducibility the benefit claim was judged by.

The arithmetic is honest about its limits, and the honesty is the discipline. The 15,000-trip recovery is an estimate from the dose-response, not a certainty, and the evaluation said so: the range, 12,000 to 17,000 trips recovered; the assumption that the feeder service’s uptake follows the pattern of the feeder-served neighborhoods; and the trigger that would test it, the counters in the two neighborhoods read monthly, the recovery on track by month forty-two or the program re-examined. The correction is funded on the evidence and tested on the evidence, the same trigger discipline that called the meeting.

The revise option is the honest renegotiation, and it is the option the committee had to keep alive because the reinforce option could fail. The revised benefit forecast is the benefit row rewritten on the evidence: if the feeder program lands, the travel-time row realizes about 96 percent of the forecast by the third operating year; if it does not, the row settles at about 90 percent. The case — the 230 million units a year of net benefit from the corrected case of chapter 37, the subsidy, the capital — holds at the revised level or it does not, and the finding goes to the funders with the numbers and the caveat. The corridor’s own sensitivity arithmetic had tested the case at a 25 percent ridership shortfall in chapter 7 and found it still positive, so the revised forecast was not the rescue of a broken case; it was the case’s own swing analysis landing on the evidence, the discipline that chapter 15 taught for the estimate applied to the benefit. The revise option is not the retreat it looks like. It is the chapter 7 discipline on the far side of the project: the business case re-tested on the evidence, the option analysis re-run with the field’s numbers, the recommendation made on the revised case rather than on the original hope. The organization that revises its benefit forecast honestly and tells its funders spends its credibility on the truth and keeps the rest; the organization that holds the original forecast past the evidence has built the green-report culture of chapter 40 around its own benefits.

The stop option is the correction’s boundary, and it belongs in the set because the corridor’s case made it real. If the feeder program’s 40 million units of capital and 8 million a year cannot be funded, and the revised case at about 90 percent of forecast still justifies the subsidy, the stop decision is the decision not to spend the 40 million. The evaluation’s contribution is to make the decision visible as a decision: the opportunity cost named, the value foregone counted, the choice made by the council on the evidence rather than by the budget cycle by default. The chapter 42 lesson — the recovery option is chosen by what the value requires, not by what the plan promised — applies to the benefit program with the same force. The reinforcement that cannot pay for itself should not be funded, and the project that realizes its benefits is the project that knows when the benefits stop being worth buying.

The erosion mechanisms are the failure pattern this section is written against, and they deserve their names because they are how the curve falls after the reviews stop: the service cut that the subsidy pressure forces, the frequency reduced, the headways lengthened, the six-minute saving still measured and the twenty-minute wait added outside it; the schedule padding that chapter 37’s drill caught, the operator padding the timetable to make the on-time number, the padding eroding the rider’s realized saving by its own measure; the maintenance deferral, the escalator down, the lighting out, the lift waiting, the access benefit quietly receding; the wayfinding decay, the signs faded, the maps out of date, the visitor’s first trip taking the wrong turn; and the fare change, the integrated fare un-integrated, the surcharge that prices the marginal rider off the marginal trip. Each erosion is a benefit row losing its value without a single headline event, and the counter-discipline is the reinforcement budget, the maintenance line, the wayfinding audit, the fare review — the rows the realization dashboard must now carry, because the benefit that is not maintained is the benefit that was never realized. It was borrowed.

The evidence goes back into the system

The last leg of the benefit loop is the one the corridor’s meeting had to decide as its final act, and it is the leg that makes the whole discipline worth its cost: the evaluation’s evidence goes back into the system — into the strategy, the portfolio, and the next business case. The project that evaluates but does not feed the evidence back has paid for the lesson and refused it.

The portfolio connection is the nearest one, and the corridor’s meeting carried it in the room. The city was already planning the next corridor, the council had a finite budget, and the evaluation’s findings — the feeder gap, the dose-response, the park-and-ride leak, the forecourt conflicts — were the reference-class data of chapter 15 for the next case: the documented evidence the next business case would use to size its ridership forecast, name its feeder dependency, price its evaluation budget, and write its trigger. The benefit rows corrected by the evaluation become the calibration for the next estimate, and the organization that feeds them forward is the organization whose forecasts get better with each project — the only way forecasts ever get better.

The strategy connection is the longer one, and it is the one that tests the thesis itself. The corridor’s evaluation asked whether the city’s mobility strategy — the corridor plus the feeder network plus the displacement program plus the subsidy — produced the value the strategy promised, and the answer, partial, distributional, with the feeder gap named, was a finding about the strategy, not about the project: the strategy’s value depended on the network that the strategy’s other parts had not delivered. The finding goes to the strategy review, and the strategy review responds with the feeder program — a new project with its own business case, its own trigger, and its own evaluation — and the loop is the learning loop of the whole book: the evidence the project’s outcomes produce feeding the selection of chapter 5 and the business case of chapter 7 for the next round, the Project Mastery equation’s Learning lens closing the loop around the Value lens.

The benefit register is the instrument of the feed-back: the profile table grown into organizational memory, the rows, the owners, the baselines, the targets, the realized values, the corrections, the lessons, kept across projects, owned by the portfolio or the PMO that chapter 47 will build, reviewed at the portfolio cadence, and consulted by every new business case. The register is the difference between the organization that remembers its projects and the organization that repeats them. The chapter 43 lesson — the lesson is a claim with an owner and an application — applies to the benefit register with the same force: the register row without an owner and a use is a remark wearing a lesson’s clothes.

The corridor’s meeting closed with the three rows on the board becoming three decisions, and the three decisions are the chapter’s summary in action. First, the reinforce decision: the feeder improvement program authorized at about 40 million units of capital and 8 million a year, funded on the evaluation’s dose-response evidence, with the counters in the two neighborhoods read monthly and the recovery trigger at month forty-two. Second, the revise decision: the benefit forecast updated, the travel-time row restated at about 96 percent of target if the program lands and about 90 percent if it does not, the corrected case taken to the council with the numbers and the caveats, the subsidy’s value-for-money stated per unit of realized benefit. Third, the feed-back decision: the evaluation’s findings added to the benefit register, the reference-class rows for the next corridor, the feeder dependency written into the next business case’s assumptions, the trigger written into the next case before the next project begins.

The loop is the chapter’s close, and it is the book’s close in miniature: the four lenses holding the corridor together from month zero to month forty and beyond. The Value lens: the 216 million units a year that became 193 and will become about 210 if the correction lands. The People lens: the owners, the traders, the residents, Grace Njoroge’s federation, Councillor Kamau’s constituents — the people the evaluation found gained and the people it found waiting. The Delivery lens: the corridor that was delivered, closed, and realized, the acceptance certificate and the benefit curve as the two bookends of the same responsibility. And the Context lens: the neighborhoods, the feeder network, the fuel prices, the employer relocation — the conditions that decided the realized value as surely as the construction did. The Learning lens is the whole chapter: the evidence fed back into the system so the next corridor starts with the truth this one paid for.

The most common next failure, after a chapter like this one, is the evaluation that is commissioned and filed: the report delivered, the findings accepted, the corrections funded, and the loop closed, because the organization believed that the evaluation’s purpose was the report. The report is not the purpose. The purpose is the three decisions, and the three decisions are only real when they are owned, funded, dated, and reviewed: the feeder program with its trigger, the revised forecast with its funders, the register with its readers. The project that realizes and evaluates has done the whole work of the book’s first forty-four chapters, and the next chapter takes the work one level up: the benefit that belongs to no single project but to the program, the coordinated change that realizes value across projects, sequences the benefits, and absorbs them into operations. The corridor’s evaluation ended with a flag for the program question, and the program question is where the value’s real story lives.

Practice

One. A quick check: name the level and the measure. For each scene, name the level of success it reports — delivery, product, adoption, or benefit — and the measure that would be leading or lagging for it. (a) The clinic platform release passed integration tests and was accepted. (b) The clinic’s clinicians chart 82 percent of their patient encounters on the platform, up from 35 percent at the December opening. (c) The corridor’s crossing time fell from 12 minutes to 6. (d) The supported shops in the trading estates recovered to pre-construction revenue. (e) The mobile app’s checkout completion rose after the reorder, and the merchant cohort’s settlement delays fell by a day. (f) The response’s health posts reported no stockout of critical medicines for the third consecutive week.

(a) is product success, the deliverable works as designed, and its measure, the acceptance evidence, is a lagging measure of the build; the leading measure for what matters next is the adoption behavior the acceptance does not count. (b) is adoption success, the behavior change, and it is the leading indicator for the access outcomes that follow by months; the 82 percent is the number the benefit evaluation must watch, because the behavior that is not happening is the benefit that is not coming. (c) is the outcome, the changed state, and the journey time is the lagging measure that ridership and mode shift lead; the six minutes is real, and the door-to-door saving is the realization question, the chapter’s neighborhood finding. (d) is the benefit, the measurable advantage valued by the traders and the program’s funders, the lagging outcome of the displacement program, led by footfall and revenue share. (e) is a product-plus-adoption pair: checkout completion is the adoption behavior, leading, and the settlement delay is the outcome, lagging — the KijaniPay benefit row the product loop measures continuously rather than at a project gate. (f) is a benefit during delivery, the humanitarian response’s interim benefit, measured weekly on the field’s clock — the reminder that some benefits realize while the project is still running, and the evaluation’s job there is the sustained-outcome question, whether the stockouts stay away after the response hands the routes and the records to the coordination cell. The common discipline: name the level before the number, because the level decides what the number means and what it cannot mean.

Two. A field drill: write the one-page benefit profile. Take a delivered project in your own organization, one whose outputs are already in service, and write the benefit profile for its most important promised benefit: the owner, the named person with authority over the levers; the baseline, the dated snapshot of the world before the change; the target with its counting rule; the dependency chain in one line; the timing, the clock, and the shape of the curve; the assumptions and the risks; and the review cadence with its trigger. Then write the disbenefit profile on the second page: the subsidy, the displacement, the unintended consequence, each with an owner, a baseline, and a correction. Then name the row that is currently reported as green and is actually unmeasured.

The drill passes when the profile is one page, the owner is a person with a lever, the baseline is dated, the counting rule is written, and the disbenefits are on the page. The most common failure is the profile written from the business case without the field — the target quoted and the baseline reconstructed, the owner named from the org chart without the lever test; the repair is the two questions, can this person affect the outcome, and is this person answerable for it, and the profile is rewritten until both answers are yes. The second failure is the profile that ends at the target, no timing, no cadence, no trigger; the repair is the clock and the review, because the benefit without a clock cannot be judged and the row without a cadence will be the row nobody challenges. The third failure is the disbenefit page left blank, the subsidy unowned, the displacement unmeasured, the unintended consequence undiscovered; the repair is the walk, the field visit that finds the park-and-ride, the forecourt, the estate that lost the customers, because the disbenefit that is not written down will be discovered in the budget cycle or the newspaper. The green-but-unmeasured row is the drill’s whole point: name it, and write the measure that would make it accountable.

Three. A field drill: build the realization dashboard. For the same delivered project, build the realization dashboard in the discipline of chapter 37: six numbers, each with an owner, a source, a refresh cadence, a threshold, and a decision attached — at least two of them leading adoption measures, at least two lagging outcome measures, and at least one disbenefit row. Then write the trigger for the dashboard’s most important row: the specific condition, the decider, and the window. Then name the row the trigger would have caught in month one if anyone had been reading it.

The drill passes when every row answers the chapter 37 test — what decision does this number serve — and the dashboard fits on a page. The most common failure is the dashboard that mirrors the delivery report, percent complete and budget variance and nothing about the behavior; the repair is the leading-lagging pair, the adoption measure that moves first and the outcome measure that follows, because the dashboard with no leading row learns the benefit failed a year after the counters knew. The second failure is the row without an owner or a decision, the number measured because it can be; the repair is the chapter 37 discipline, the number without the decision is decoration. The third failure is the trigger written in the mood, if things are going badly, which is not a trigger because it cannot be checked; the repair is the specific condition, the band, the threshold, the two consecutive months — the same specificity that called the corridor’s committee back to the table. The month-one row is the drill’s lesson: every project has one, the counter calibrated as a readiness item and read as construction data, the adoption number collected and not owned, the early warning that was there and unread; the drill’s value is naming it, because the next project’s month one is the month the trigger should have been watching.

Four. A decision room: what does the committee do with the neighborhood split? It is month thirty-eight at BlueLine, and the evaluation is on the table: the corridor-level ridership about 11 percent below the band, the realized travel-time saving nine to eleven minutes in the feeder-served neighborhoods, three to four in the partial, about one in the no-feeder two, the park-and-ride traffic eating the emissions row, the third trading estate still below pre-construction revenue, the carless-access share at 52 percent against the 65 percent target, and the council’s budget cycle six weeks away, with the next corridor’s business case waiting on the corridor’s evidence. The options: (a) report the corridor-level benefit — the six-minute saving and the ridership within 90 percent of forecast — as the realization claim, and take the feeder program to the next budget cycle with a modest proposal; (b) report the neighborhood split, recommend the feeder improvement program at about 40 million units of capital and 8 million a year with the recovery trigger and the revised benefit forecast at 96 percent or 90 percent of target, and take the corrected case to the council; (c) renegotiate the ridership band with the funders now, on the argument that the 180,000 forecast was always the case’s optimism and the realized 161,000 is the honest number; (d) defer the decision to the next committee, on the argument that the evaluation needs a considered response and the budget cycle can absorb a placeholder. Decide the move, defend the trade, and say what the record must carry.

The defensible answer is (b), and the reasoning is the chapter’s whole discipline: the corridor-level number is not the story, the neighborhood split is the diagnosis, and the correction is funded on the diagnosis — the feeder program with its arithmetic, its trigger, and its revised forecast, taken to the council as a corrected case rather than as a promise. (a) is the reasonable-but-risky answer, and its risk is the report that hides the fork: the committee reports the average and the council funds on the average, the no-feeder neighborhoods stay unserved, the third estate keeps losing its shops, and the next budget cycle discovers the row the report concealed — the green-report culture of chapter 40 applied to the benefit claim; it is defensible only if the feeder program genuinely cannot be funded this cycle and the corridor-level claim is stated with the split named. (c) is the wrong answer at this point because it renegotiates before the correction is tested: the 161,000 is not the ceiling, the dose-response says the feeder program can lift it, and the committee that resets the band now will fund the next corridor on a forecast it never had to defend; the revise option belongs in the set, but it is the consequence of the reinforce option’s failure, not the alternative to trying it. (d) is the unsafe answer because it lets the budget cycle decide: the placeholder becomes the default, the feeder program dies in the cycle’s arithmetic, and the decision visible in month thirty-eight becomes the decision made by circumstances in month forty-four — the chapter 43 lesson that the decision postponed is the decision made by circumstances. The record must carry the three decisions: the feeder program with its budget, its trigger, its owner, and its review date; the revised benefit forecast with its range and its assumptions; and the benefit register rows updated with the evaluation’s findings — the reference-class evidence for the next corridor, because the record is the feed-back, and the feed-back is the only thing that makes the next case better than this one.

Five. A decision room: what can the evaluation honestly claim? The city council has asked for the evaluation’s headline for the next budget cycle: “the corridor cut journey times by 6 minutes and attracted 180,000 trips a day,” the sentence the mayor’s office would like to publish. The evaluation’s evidence: the station-to-station saving of about six minutes measured against the baseline, the door-to-door saving varying from about one to about eleven minutes by neighborhood, the ridership at about 161,000 against the 180,000 forecast with the fuel-price rise and the employer relocation in the operating period, and the comparison areas showing the corridor’s contribution between 80 and 90 percent of the observed trips. The options: (a) publish the mayor’s sentence, on the argument that the corridor’s saving is real and the forecast was the case’s target; (b) publish the bounded claim — the six minutes with the neighborhood range, the ridership at 161,000 with the comparison evidence and the contribution stated as a range; (c) publish no numbers, only the narrative of the corridor’s opening and the feeder program’s case; (d) publish the bounded claim with the trigger and the revised forecast, including what the corridor has not yet achieved. Decide the move, defend the trade, and say what the public record must carry.

The defensible answer is (d), and the reasoning is the contribution discipline: the claim that the corridor cut journey times is true within its definition — the station-to-station six minutes measured against the baseline — and false beyond it, the door-to-door range that the corridor alone does not control; and the claim that the corridor attracted 180,000 trips is not true at all, it is the target, and the difference between the target and the claim is the difference between the promise and the evidence. The bounded claim carries the honest sentence: the corridor contributed a reduction in corridor journey time of about six minutes, with the realized door-to-door saving varying by neighborhood from about one to about eleven minutes depending on the feeder access, and the corridor’s contribution to the observed ridership of about 161,000 trips a day estimated at between 80 and 90 percent against the comparison areas, with the feeder program authorized to close the gap and the revised forecast at about 96 percent of target if it lands. The sentence is longer, and the mayor’s office will resist it, and the resistance is the evidence that it is true. (a) is the unsafe answer, the credit error: the fuel price and the relocation helped, and the corridor took the credit, the claim that the next corridor will be funded on a forecast the public was told was already achieved; the claim’s cost is not the correction, it is the trust — the trust ledger of chapter 9 charged for the benefit claim. (b) is the reasonable-but-risky answer, the bounded claim without the forward view, honest about the past and silent about the decision; the repair is the trigger and the revised forecast, because the public that reads the bounded claim in (b) will read the revision in (d) as a retreat, while the public that reads the revision with the claim will read it as the management it is. (c) is the wrong answer because the evaluation’s evidence is the city’s money — the subsidy, the capital, the trust — and the narrative without the numbers is the report that hides the range, the chapter 40 failure wearing an evaluation’s clothes. The public record must carry the claim, the range, the contribution evidence, the correction, and the trigger, because the record is the accountability, and the accountability is the chapter’s whole argument: the benefit realized as the only proof of the project.

Six. The mastery drill: design an evaluation that is honest about contribution. It is month ten after the go-live of a project you led or worked on, or one you can observe closely, and your organization’s leadership has asked for an evaluation that answers whether the project worked, to decide whether to fund the next phase. The project delivered a new service or system, the outcome of interest is real, the leadership wants a number they can publish, and the field has confounders: other initiatives, market changes, policy shifts, people’s behavior changing for their own reasons. Design the evaluation: the purpose, the five questions, the methods with the comparison design, the measures with the leading and the lagging, the timing, the budget, and the independence; then write the contribution claim the evaluation will be able to defend — with its range, its evidence, and its caveats — and the claim the leadership must not be allowed to publish instead.

The drill passes when the evaluation is designed to answer what the project contributed, not what happened, and the claim is written as a bounded contribution with the evidence named. The method design is the test of the chapter: the comparison areas, the neighborhoods, sites, or cohorts without the treatment, chosen for similarity on the measured confounders; the dose-response, the units with more treatment versus less, the stations with the feeder service versus without, the cohorts with the onboarding versus without; the baseline captured before or reconstructed from records with the caveat named; and the chain tested link by link, the outputs observed, the capabilities observed, the behaviors observed, before the outcome is credited to the treatment. The claim’s shape is the chapter’s shape: the project contributed an improvement of X, measured against the baseline, with the comparison evidence showing the contribution between A and B percent of the observed change, the variation by condition named, and the confounders stated; the claim is a hypothesis with its test attached, and the reader can see what would change it. The claim the leadership must not publish is the credit error in its two forms: the before-and-after number that credits the project with the market’s rise, and the target restated as an achievement — the 180,000 forecast published as the 180,000 realized, the claim that will be funded forward on evidence it never had. The unsafe design is the evaluation with no independence, the review commissioned by the team that built the thing and removable by them; the independence test of chapter 41 applies unchanged — who can say no, and who can remove them if they do. And the drill’s honest close is the decision the evaluation is for: the next-phase funding made on the contribution claim with its range, the correction funded on the diagnosis, the revised forecast taken to the funders, and the evidence fed into the register, because the evaluation designed to be honest about contribution lets the next project start with the truth this one paid for.

Seven. The transfer question. On the delivered project you chose for the drills, or the one you work on now: where do the benefit rows live — in the business case’s appendix, archived with the team, or on the profile pages with owners, baselines, and clocks? Who owns the row that matters most, and does that person own the lever — the timetable, the fare, the contract, the budget — or only the number? What is the baseline, and was it dated before the change, or will it be reconstructed at the evaluation, with the reconstruction’s flattery? What does the clock read, and is the reading on a dashboard with a refresh cadence, an owner, and a trigger, or in a report that was produced and filed? What is the curve doing in your project — still in the adoption dip, climbing, sustained, or eroding — and what would the evidence be that told you which? What are the disbenefit rows — the subsidy, the displacement, the unintended consequence — and who owns them, or are they the rows nobody wrote and the crisis that will be discovered? Who pays for the measurement and the evaluation after the closure, and was the evaluation budget written into the business case at authorization, or is it the operating budget’s discovery? What is the leading measure that would have warned you in month one, and who was reading it? What are the confounders in your field — the fuel prices and the employer relocations and the other initiatives — and what is the comparison that would separate your contribution from theirs? What claim could your organization defend in public today, and what claim would it like to publish instead, and what is the distance between them? And the question underneath them all: what will the evaluation’s evidence change — the next business case, the next forecast, the next trigger, the next budget, or nothing at all? The project that realizes and evaluates and feeds the evidence back has closed the whole loop: the benefit owned, the clock kept, the curve held, the contribution claimed with evidence and caveat, the evidence fed back into the system, the learning the multiplier that makes the next project’s judgment better than this one’s — the mastery of realization as the accountability that outlives the team, and the question of what the project changed answered by the value, not by the certificate. Chapter 45 will ask it at the scale of the program, where the benefit belongs to no single project, the coordinated change that realizes the value across projects, the benefits sequenced, the operations absorbing the result, and the evaluation that follows the value beyond the boundary of any one project’s close.

Notes

• The composite cases remain author-created illustrative material. The BlueLine month-thirty-eight evaluation review, the ridership trigger fired in month thirty-six after two consecutive months below the band, the counters at all twenty-four stations calibrated, the weekday ridership at about 161,000 against the 180,000 forecast, the six evaluation areas in three pairs, the realized door-to-door savings of nine to eleven, three to four, and about one minute by neighborhood cluster, the feeder frequencies of eight to ten minutes, thirty minutes, and none, the park-and-ride counts of about 4,100 cars a day at the three outlying stations, the forecourt conflict counts at three stations, the shop-viability revenue recovery in the three trading estates with the third below pre-construction, the carless-access share of 52 percent against the 65 percent target with 31 percent in the no-feeder neighborhoods, the modal-shift observation of 55/45 against the model’s 70/30, the feeder improvement program’s estimate of about 40 million units of capital and 8 million a year of operating cost recovering about 15,000 of the missing trips a day with the range of 12,000 to 17,000 and the recovery trigger at month forty-two, the revised benefit forecast at about 96 percent of target if the program lands and about 90 percent if it does not, and all named characters and roles are teaching constructions consistent with the facts established in earlier chapters: the corridor’s four segments, twenty-four stations, the phased opening of the central and northern segments at month twenty-five, the eastern segment’s opening at month twenty-eight with the re-run of the seams, the month-twenty-six intervention, the rebaseline and rephase, the independent readiness review, and the reviewer named Abena, from chapters 41 and 42; the corrected business case with the travel-time savings of 216 million units a year from 180,000 passenger trips a day saving six minutes, about 5.4 million hours a year at 40 units an hour, the net annual benefit of about 230 million units, the 90 million-unit operating subsidy, the ridership trigger written before opening with the committee reconvening on materially lower ridership, the benefit ownership transferred at the gate to named operational owners, the displacement-support program measured against shop viability with Councillor Diana Kamau’s access row, and the benefit dependency network with its disbenefit branches, from chapters 7 and 37; the model’s 70/30 split versus the field’s 55/45, the 1,400 shops, the trader relocation, the market association, Yusuf, Aisha Bello, and the petition’s growth, from chapter 9; the baseline of 50,000 vehicle-trips a day, the 12-minute crossing rising to 22 in a decade, the modal share, and the corridor’s case arithmetic, from chapter 7; the adoption corridor, the trained-versus-adopted gap, the RE-AIM framework, and the implementation dip, from chapter 11; the counting-rule discipline, the data-quality log, the minimal dashboard, the metric families, the six operating measures at the opening gate, the benefit-owner rows, the schedule-padding drill, and the measurement transition, from chapter 37; the reporting provenance, the green-report culture, and the verification record, from chapter 40; the readiness review’s independence test and the three-lines framing, from chapters 24 and 41; the eastern segment’s forecast and the month-twenty-eight risk carried publicly, from chapter 38; the closure transfer, the acceptance certificate, and the lesson-with-an-owner discipline, from chapter 43; the grant’s access outcomes, the adoption measures at the point of work, Nora Kariuki’s operational ownership with the grant targets, and the month-thirty-six grant window at Meridian, from chapters 8, 11, and 36; the KijaniPay benefit clock, the 350 applications by September, the run-rate readings, the settlement-visibility prototype, the fraud-loss and settlement-promise guardrails, and the product-loop measurement, from chapters 32, 38, and 40; and the Northstar response’s interim benefits, the health-post stockouts, the handover to the coordination cell, and the sustained-outcome question, from chapters 29, 35, and 42. The teaching numbers introduced here are fully reproducible from the text where the arithmetic is shown: the realized travel-time value of about 193 million units a year follows from 161,000 trips against the 180,000 forecast on the same 6-minute and 40-units-an-hour basis, and the gap of about 19,000 trips at about 23 million units a year follows from 19,000 trips a day times 6 minutes, 1,900 hours a day, 570,000 hours a year over 300 operating days, 22.8 million units a year at 40 units an hour, stated with its assumptions; the corridor’s modal-shift observation, the neighborhood ridership shares, the park-and-ride counts, the emissions offset, the forecourt conflicts, the shop-revenue recovery, the carless-access shares, the feeder program’s capital, operating cost, recovery estimate and range, the month-forty-two trigger, and the revised forecast percentages are author-created teaching constructions consistent with the established facts and stated with their assumptions, not measurements the earlier chapters established; the exact ridership of any operating month and the precise recovery of any correction are presented as teaching patterns, not as data from the case’s earlier chapters. The standards and sources are described in the book’s own words: the general frames for benefit realization, monitoring, and evaluation follow ISO 21502:2020, Project, programme and portfolio management: Guidance on project management, and the PMBOK Guide, Eighth Edition, Project Management Institute, November 2025, per the book’s reference baseline of 1 August 2026, which treat value, benefits, monitoring, and evaluation among their general practices and performance domains, all described here in the book’s own words as general frames rather than quoted; per the same reference baseline, the ISO/TC 258 committee’s published project list includes supporting guidance for post-project and post-program evaluation issued in 2026, described here only as a general frame for the evaluation discipline, with the instruction that the current committee project list be consulted at publication; the contribution-analysis discipline, the shift from claiming attribution to bounding contribution through the testing of the assumed chain and the assessment of plausible alternatives, follows John Mayne, “Addressing Attribution Through Contribution Analysis: Using Performance Measures Sensibly,” The Canadian Journal of Program Evaluation 16, no. 1 (2001), pages 1-24, described here in the book’s own words; the leading-and-lagging adoption and outcome measures and the reach, effectiveness, adoption, implementation, and maintenance framework follow Russell E. Glasgow, Thomas M. Vogt, and Shawn M. Boles, “Evaluating the Public Health Impact of Health Promotion Interventions: The RE-AIM Framework,” American Journal of Public Health 89, no. 9 (1999), pages 1322-1327, as introduced in chapter 11 and extended here, described here in the book’s own words; the implementation dip, the performance decline that precedes the rise when a new practice is adopted, follows Michael Fullan as cited in chapter 11, described here in the book’s own words; the UK Treasury’s Green Book, cited in chapter 7 for the social time preference rate, likewise frames public-sector appraisal and evaluation as a continuing discipline in which projects are monitored and evaluated after the decision, described here in the book’s own words as a general frame rather than quoted; the benefit profile with its owner, baseline, target, dependency chain, timing, assumptions, and cadence, the benefit clock, the benefit realization curve with its adoption delay, erosion, and reinforcement interventions, the realization dashboard, the evaluation plan with its purpose, questions, methods, timing, budget, and independence, the contribution claim in its bounded form, the revised benefit forecast, and the benefit register are the author’s own method-neutral instruments, named and described in this book’s own words. This book remains independent of PMI, ISO, the UK government, and all standards and framework bodies, and no proprietary certification manual, commercial text, or framework guide is reproduced or paraphrased here.