Chapter 7: Build a Business Case and Benefits Hypothesis

The case that was already done

Two months after the framing session, the BlueLine steering committee meets to review the second half of the corridor program. Daniel Osei, the consortium’s finance director, has asked for the business case. What arrives is a consultant deck from month zero: forty slides, corridor renders, a benefits list of eleven items, and a single headline number, a net present value of 1,200 million units, with the note “prepared in support of the approved business case.”

Lena Voss turns to the options section. There is none. She turns to the sensitivity analysis: one slide, three bars, all green. She turns back to the headline.

Daniel speaks first. “I asked for the case because we are about to commit the second half of the money. I want to know what we are buying, what it costs, who realizes it, and what could change the answer. This deck tells me the number was decided in month zero and has not been touched since.”

Miguel Alvarez, the modeler, is already running the corrected model. “The case was built on a seventy-thirty mode split. The field counts killed that assumption in month eight. The deck has not noticed.”

“Then run it again,” Daniel says. “With the model we now trust, and with the subsidy, the displacement, and the things the corridor actually does to the neighborhoods. If the case survives that, I will sign the second half. If it does not, we talk about what it would take.”

This chapter is about that rerun. The business case is the project’s economic, social, operational, or public-value rationale, written before authorization and renewed through delivery. Done well, it is a decision instrument: it lets a sponsor answer, on evidence, whether the work should exist, which option should be taken, and how the organization will know when the value has arrived. Done badly, it is a filing: persuasive, filed, and never reopened. The corridor’s original case was the second kind, and the difference between the two is the difference between committing money to a number and committing money to a decision.

A decision instrument, not a filing

A business case exists to make three decisions possible, and everything in it should serve one of the three.

First, is the change justified: does the value it produces exceed the full cost of producing and running it, for the people whose value counts? Second, is this the best of the realistic options: among the genuinely available ways to produce that value, including doing nothing and doing the minimum, does this one earn its extra cost? Third, can the value be realized and kept: who changes their behavior, who owns the benefit after handover, and how will the organization measure whether the change happened?

Every weakness in a case is a failure on one of those three questions. The case with a single number fails the first: it has a conclusion and no argument. The case with a benefits list and no owners fails the third: it promises outcomes nobody will be accountable for. The case with one column fails the second: it never asked whether the money had a better home.

The word that matters in the chapter title is “hypothesis.” The business case is the authorization-time statement of the value hypothesis from chapter 6, in economic form. Each benefit is a claim: a specific measurable change will occur for a named stakeholder group, resting on a stated chain of dependencies, with an owner and an evidence plan. Nothing in the case is true because it is written down; the case states what the project expects the world to do, so the project can check whether the world complied. A case written as a promise invites celebration at handover. A case written as a hypothesis invites measurement.

The case for change is a baseline argument

The case for change opens with a baseline, not a vision: the current state, why it is untenable or improvable, and the gap between where the organization is and where it needs to be. The baseline is a dated snapshot with numbers: journey times, modal share, congestion hours, access counts, emissions, operating costs. At BlueLine the baseline is the corridor’s plainest argument: fifty thousand vehicle-trips a day, growing, on roads that cannot be widened; the city’s own model projects the crossing that now takes twelve minutes taking twenty-two in a decade.

The drivers of change must be real whether or not the proposed solution is built. Population growth, employment patterns, policy commitments, and binding constraints are the case’s load-bearing walls. The discipline test: remove the solution from the case for change and see what remains. If nothing remains, the case is the solution’s echo. “The corridor exists because congestion is rising” survives the test; “the corridor exists because a corridor would be nice” does not. This is the same discipline as chapter 6’s un-solving of the request, applied to economics: the case must stand on the problem before it can carry the solution.

The gap statement closes the section: the difference between where the baseline trends and where the stated objectives sit. It should be one sentence long and it should embarrass the organization a little. “At current trends the city misses its accessibility and emissions targets by 2040” is a case for change. “The corridor will modernize the city’s image” is a brochure.

Options, including the one nobody proposes

The second decision, best option among the realistic alternatives, is where most cases quietly die: the proposal must compete with things it would rather ignore. The minimum set of options has four members: do nothing, do minimum, do something, do more.

Do nothing is the reference, and it needs careful definition. It is not “the world stands still”; it is “the world without this project,” and that world still changes: congestion grows, policy moves, competitors act. The do-nothing baseline is the zero from which every other option is measured, and it is not free — a city that does nothing still pays the congestion cost, just outside any project’s accounts. Do minimum is the smallest intervention that changes the situation materially: signal timing, bus lanes, demand management, a smaller scope — and its entire purpose is to force the proposed option to justify its extra cost. The question is never “is the corridor better than nothing,” which a straw man always loses; it is “is the corridor worth its price over the do-minimum,” and that is a real question.

Do something is the proposed option; do more is the ambitious variant, the superscope, included for a different reason — to expose the option’s upper boundary and test whether the incremental cost of ambition earns its incremental benefit. Most projects never analyze do more, and then wonder why scope creeps upward without an economic argument; the row was always missing.

The comparison needs a discipline check of its own: options must be analyzed while they are open, or the retrospective becomes an audit of the authorization logic. The failure pattern to name is the one-option case — the deck with a single column, the alternatives invisible — whose tell is simple: nobody in the room can answer what they would buy instead with this money, and what they would lose by not buying it.

The comparison the renewal builds has the shape the original deck should have carried: every row populated, the do-nothing baseline as the zero, the non-monetized rows beside the money. The NPV column follows from the discounting worked later in this chapter.

Read the money rows and the trap appears: on net present value alone, the do-minimum beats the corridor. The corridor’s defense is the row below the money: 62,000 households within a ten-minute walk of frequent service, more than twice the do-minimum’s 28,000, plus the emissions and the city-scale access the stated objectives name. The table is the argument in miniature. The extra capital buys the objective, and the honest question is whether the objective is worth its price — a judgment, not a spreadsheet output.

Costs that outlive the project

The third element is the full cost, which means the cost across the whole life of the asset, not the cost until handover. Five buckets: capital, to build; operating, to run; maintenance, to keep; renewal, to replace at end of life; and decommissioning, to close. The classic failure is comparing options on capital alone, or stopping the clock at handover, which is exactly the moment operating cost and benefit both begin. For the corridor: 2,400 million units of capital, and then an operating subsidy of 90 million units a year that belongs in the same table, because the corridor’s economics are the corridor’s economics whether the cost appears in year zero or year twenty.

Two cost disciplines deserve their own sentences. Sunk costs are gone: money already committed cannot be recovered by continuing, so it never appears in a forward-looking decision, a point the renewal gate will lean on hard. And the price basis must be stated: real figures (constant purchasing power) or nominal figures (as spent), because mixing them inflates the future and hides the present. A real case states its basis on every page that carries a number.

Opportunity cost completes the picture. The money the corridor absorbs has alternatives, and the do-minimum row is the corridor’s most concrete opportunity cost: 2,250 million units less capital, at the price of fewer households served. Opportunity cost is not an abstraction; it is the row you refused to compare.

Benefits are measured changes, not deliverables

Benefits are the fourth element, and they are the most commonly faked, because the words are easy to confuse. An output is what the project produces: the corridor, the stations, the ticketing system. An outcome is a changed state or behavior: journeys that take six minutes less. A benefit is a measurable advantage valued by a stakeholder: the value of those six minutes to the people who save them. The case must count benefits, and benefits are never outputs relabeled. “The corridor will deliver twenty-four stations” is a bill of materials, not a benefit. “Commute time falls by six minutes for a hundred and eighty thousand daily trips” is a benefit waiting for its valuation.

Tangible benefits can be measured and often monetized: time savings, cost savings, revenue, avoided expenditure. Intangible benefits resist pricing but not measurement: trust, wellbeing, cohesion, brand, learning. The discipline is to measure what can be measured and name what cannot, and never to let “intangible” mean “imagined” — a benefit that cannot be measured is a hope, and hopes belong in the assumptions list with an owner and a test date, not in the benefits list. The organization that puts “improved community trust” in its benefits list without a measurement plan has written a slogan in a ledger.

The spine that holds benefits together is the benefit dependency network, the chapter’s map of how value actually arrives. It runs in six layers, from enabling outputs through capabilities and behavior changes to outcomes, benefits, and strategic objectives, and every benefit is a hypothesis about a chain through these layers; every chain is an audit trail.

Figure 7.1: The benefit dependency network for the BlueLine renewal. Enabling outputs feed capabilities, which enable behavior changes, which produce outcomes, which become benefits, which serve strategic objectives; the disbenefit branches, displacement and subsidy, sit beside the chain.

The network earns its name from its arrows. A benefit with no dependency chain is a wish: nothing connects it to an output anyone will build. A benefit whose chain crosses an ownership boundary with nobody responsible on the far side is a risk: the project builds the output, the operation does not create the capability, and the benefit silently fails. The map makes both failures visible before the money moves.

The arithmetic of time

The fifth element is the numbers, and they all rest on one idea: a unit of currency now is worth more than the same unit later, because it can be invested and because the future is uncertain. Discounting translates future amounts into present terms: the present value of an amount received n years from now is that amount divided by (1 + r) raised to the n-th power, where r is the discount rate. The rate is a decision, not an input. A commercial organization uses its cost of capital; a public body uses a social time preference rate, which the United Kingdom’s Treasury Green Book sets at 3.5 percent for most appraisals. The consortium running BlueLine uses 5 percent, its cost of capital, and the choice matters: a higher rate discounts the future more heavily and tilts the comparison toward cheaper, faster options.

The discount factor is the arithmetic’s workhorse: at 5 percent, one unit arriving one year from now is worth 0.9524 today, and one arriving in year twenty is worth 0.3769. A benefit of 230 million units in year twenty is worth about 87 million units today: the corridor’s rewards arrive slowly, and the arithmetic insists on charging them for the wait.

Here is the corrected case, the one the committee asked Miguel to run. All figures are author-created teaching numbers in real local-currency units, with the arithmetic shown so any reader can reproduce it.

The travel time savings deserve their own line of arithmetic, because they are the case’s engine. A hundred and eighty thousand passenger trips a day, saving six minutes each, is eighteen thousand hours a day, about 5.4 million hours a year over three hundred operating days, and at 40 units an hour, 216 million units a year. Every one of those inputs is a claim, and the sensitivity section will treat them accordingly.

Discount at 5 percent. The present value of the capital is 800 times (1.0000 + 0.9524 + 0.9070), which is 2,287.5 million units. The present-value factors for the eighteen benefit years run from 0.8638 down to 0.3769 and sum to 10.60, so the net benefits are worth 230 times 10.60, which is 2,438.6 million units. Net present value (NPV): 2,438.6 minus 2,287.5, about +151 million units. The same arithmetic across the options yields about +224 million units for the do-minimum and about +230 for the mitigated corridor, the numbers the comparison table carries; any reader can reproduce them from the factors in this section.

Now the interpretation, because the number is not the argument. The case is positive: it covers its cost of money with a margin. The margin is thin. Simple payback, the years until cumulative benefits cover cost, is 2,400 divided by 230, about 10.4 operating years, or roughly twelve to thirteen years from the start of construction. Payback is blind to everything after it and to the time value inside it; treat it as a liquidity and risk-exposure constraint, never as a decision rule. The internal rate of return (IRR), the rate at which the NPV reaches zero, comes out at about 5.7 percent, barely above the 5 percent discount rate. The sentence worth rereading is the one before this: the case clears its hurdle by less than one point. A case that clears by less than one point lives or dies on its assumptions.

For benefits that resist monetization, the same discipline takes the form of cost-effectiveness: cost per unit of outcome, compared against a threshold or against other programs. The corridor’s emissions line is the example. At 750 units per tonne, the 30 million units a year of monetized emissions represents about 40,000 tonnes avoided a year, about 720,000 tonnes over the eighteen-year horizon, which makes the corridor cost about 3,300 units per tonne avoided, more than four times the value input. As a carbon project the corridor is a poor one, and the honest conclusion is that it is not a carbon project: the emissions line is a co-benefit. The discipline of cost-effectiveness exists precisely to stop a project from claiming a co-benefit as its reason.

The number is a range wearing a costume

Every input in the corrected case is a guess with a pedigree, and the single number hides the guesses. The sixth element is honesty about that, and its instrument is sensitivity: asking, one assumption at a time, what would have to be different for the recommendation to change. The output is a table, and the table’s purpose is to find the swing factor: the assumption whose plausible range moves the decision across the line.

Read the table as a sentence: ridership dominates. It enters three benefit components, time, emissions, and safety, and the corrected case breaks at roughly a 5 percent ridership shortfall, because the case carries almost no headroom on its own engine. (The vehicle-operating-cost saving was held fixed for this pass; most of it comes from congestion relief for all traffic, not from corridor ridership.) A case whose most important assumption can flip the recommendation at five percent below its own forecast has no headroom; it is not a case that survives contact with reality, it is a case that must be designed around.

The design response to a fragile case is not better formatting. It is structure. Stage the funding so the next tranche is earned by evidence. Test the swing assumption early: for ridership, the first months of operation are the experiment, so the trigger must be written before opening, not discovered after. Create the option to stop or re-scope cheaply, because a decision that hinges on one assumption is really a series of smaller decisions, and staging is how the series is bought. The economists’ name for this is the real options view of investment: flexibility has value, and a project that can be slowed, shrunk, or stopped without ruin is worth more than one that cannot.

Optimism bias is the reason sensitivity analysis cannot be decorative. Human beings systematically understate costs and overstate benefits, and the pattern is documented, not anecdotal: a widely cited 2002 study of 258 transport projects found rail projects averaging about 45 percent cost overrun and roads about 20 percent. The United Kingdom’s Green Book requires appraisers to adjust for optimism bias using reference-class data where available: locate the project in the class of similar completed projects, forecast from what that class actually achieved, and disclose the adjustment. The opposite practice, a case built on the team’s best hopes, is not optimism; it is misrepresentation wearing a spreadsheet.

Who owns the benefit after handover

The seventh element is the one most cases skip, because it is about people who are not in the room. Every benefit needs an owner on the operations side, named before handover: the person accountable for the measured outcome after the project ends. The benefit register carries the full set: benefit, dependency chain, baseline, target, measurement method, review cadence, owner. Its purpose is to make chapter 1’s ownership change concrete: the project produces the output, the operation must create the capability and the behavior change, and someone must be accountable for the difference. A benefit whose owner is the project manager is a benefit with an owner who will be gone when the value was supposed to arrive.

Disbenefits are the mirror, and they are as real as benefits. Negative effects on stakeholders need the same treatment: named, measured, owned. The corridor’s displacement of about 1,400 shops, the delivery delay that chapter 3 priced at roughly 5.9 million units over construction, the noise, and the subsidy burden all belong in the case, with owners on the far side. The subsidy has one: Aisha, the transit operator, owns the 90 million units a year because she owns the operating result it funds. The displacement has one: the mitigation owner, who owns the trader-support program and its measured outcome, shop viability, not just its budget.

Unintended effects are the disbenefits nobody listed, and they travel outside the delivery boundary. Traffic diverted onto local streets, rents rising near stations, the school crossing that suddenly sits beside a busway. Chapter 3’s lesson — that the project is a system and the case must look beyond the construction boundary — applies to the accounts with the same force it applies to the context map. A case that counts only what happens inside its own fence is a case that will be surprised by its own consequences, and the surprise will be expensive.

One more row belongs in the register: distribution. NPV aggregates; it adds gains and losses across people and reports the total. A positive total can hide a project whose benefits flow to commuters while its costs land on traders. That is not an economic detail; it is the political and ethical shape of the project. The case must show who pays and who benefits with the same seriousness as the total. The corridor’s equity row is why its options table carries household access and shop impact next to the money.

The gate that renews the case

The eighth element is time. The case is a living instrument with an expiry date, re-opened at gates, when an assumption changes materially, when a new option appears, when strategy moves. The renewal is not a ceremony; it is the full method again: baseline updated, costs and benefits re-run, sensitivity re-done, owners re-confirmed, the decision re-stated. And the renewal can kill the project. Chapter 2’s stop-or-pivot conditions live here: a case that dies at a gate is a success, because the money that would have been spent after the evidence died is saved.

The BlueLine gate is the method’s hardest lesson, and it turns on sunk cost. About a thousand of the 2,400 million units are already committed; they are gone, and no future decision should treat them as reclaimable. The forward question is the only real one: what remains costs about 1,540 million units, 1,400 to finish as engineered plus 140 for the mitigations, against a net annual benefit of 250 million units a year with the mitigations in place. Discounted the same way, the forward NPV is about +1,111 million units, and it stays positive even at a 25 percent ridership shortfall, about +406 million units. The project was a fragile bet in month zero and a sound continuation in month ten, and the two questions must never be merged. The committee that refuses to continue because the original case was weak has made the sunk-cost error in reverse: it has let past money, or past reputation, contaminate a forward decision. The committee that continues without re-examining the case has made the original error again.

The recommendation the committee reaches is the case in miniature. Continue with the mitigations, on three conditions. First, the ridership trigger: if early operating ridership is materially below the forecast band, the committee reconvenes and re-scopes, and the trigger is written before opening, not after. Second, benefit ownership transfers now, at the gate, to named operational owners, not at handover, when it is too late. Third, the displacement-support program is measured against shop viability, and its result is reported in the same table as the NPV, because a number that ignores the traders is a number the traders will eventually veto.

Councillor Diana Kamau adds the last row herself. “The number is positive,” she says. “The traders it displaces do not get a vote in the number. I want the access row and the displacement row signed before I sign the money.”

Daniel signs the money with the conditions attached. The case, at last, is a decision instrument.

What the delivery approach changes

The delivery approach changes where the case lives, not whether it exists. In predictive delivery, the case is a formal gate artifact: full options analysis, whole-life costs, defined horizon, sensitivity, and a gate that must be able to reject, or the ceremony is empty. In adaptive delivery, the case is thin and fast. The benefits hypothesis is the product hypothesis; each increment is a small options analysis of its own — build against not-build and change — and funding is renewed on evidence rather than on promise. In hybrid delivery, which is BlueLine’s shape, the case is staged: monetized rows sharpen as evidence arrives, judgment rows stay open, and the gates re-run the arithmetic with the same discipline each time. What does not change across the styles is the question set — is it justified, is it the best option, who owns the value — and the honesty rule: name the swing assumption.

A word about the assistant that will draft your case narrative, propose an options list, and structure the sensitivity table. A language model does all three quickly and usefully — and, with equal fluency, produces a business case whose arithmetic is wrong and whose intangibles are confident, because fluency is not evidence. Three rules. Reproduce every calculation in a spreadsheet and check it; a model’s numbers are drafts, not data. Treat the model’s option lists as prompts for the do-nothing and do-minimum rows you would rather skip, not as substitutes for the stakeholders who will live with the option. Keep financial, personal, and regulated data out of any tool not approved for it. The judgment of what counts as a benefit, who owns it, and which assumption is allowed to carry the decision is human, and it is the entire point of the exercise.

Practice

1. A quick classification. Label each statement as output, outcome, benefit, disbenefit, or cost: (a) “The corridor opens with twenty-four stations and a shared ticketing system.” (b) “Average commute time on the corridor falls from forty-six to forty minutes.” (c) “Traders on the corridor carry about 5.9 million units of delivery delay during construction.” (d) “The operator carries an annual subsidy of 90 million units.” (e) “Households within a ten-minute walk of frequent service rise from 15,000 to 62,000, which the city values as an equity gain.”

(a) is an output: produced by the project. (b) is an outcome: a changed state, not yet a benefit, because nobody has yet said who values it and by how much. (c) is a disbenefit: a negative effect that must be owned and counted, or it will arrive unowned. (d) is a cost: an ongoing financial burden that belongs in the same table as capital. (e) is a benefit: a measurable advantage valued by a stakeholder, and it becomes one only because the city states its valuation; without the valuation sentence, it would still be an outcome. The trap in the set is (b): outcomes and benefits are the most commonly confused pair in real cases.

2. A benefits map. Take an initiative you know from work, community, or study, and draw its dependency network in six layers: enabling outputs, capabilities, behavior changes, outcomes, benefits, strategic objectives. Then take the two benefits at the top and write, for each: baseline, target, measurement method, and a named owner who is not on the project team.

The drill succeeds when every benefit has a chain that reaches a behavior change, because a benefit with no behavior change is an output no one will use, and when every top benefit has an owner outside the project, because that person will be accountable after handover. The red flags are a benefits layer that relabels outputs, and a register whose owner column contains the project manager’s name.

3. A numbers drill. A proposed system costs 550 million units, spent now, and is expected to produce 150 million units a year for five years. Using the discount factors at 8 percent, 0.93, 0.86, 0.79, 0.74, and 0.68 for years one through five, compute the present value of the benefits, the net present value, the simple payback, and state whether the internal rate of return is above or below 8 percent.

The present value of the benefits is 150 times the sum of the factors, 4.00, which is 600 million units. The NPV is 600 minus 550, plus 50 million units. The simple payback is 550 divided by 150, about 3.7 years, so the investment is repaid early in year four. Because the NPV is positive at 8 percent, the IRR is above 8 percent; solving the series puts it at about 11 percent. The interpretation: the project clears its cost of capital, but payback misses the years after five and the time value inside the payback period, so the NPV is the decision rule; the IRR is best used as a hurdle comparison, with care, because it can mislead on scale and on projects with changing cash-flow signs.

4. A decision room. A board must choose between two community programs. Option A: NPV +120 million units, serves 3,000 households, no subsidy. Option B: NPV +90 million units, serves 12,000 households, annual subsidy of 30 million units. The board’s stated objective is access for low-income households. Choose and defend, and name the evidence that would change your answer.

A defensible answer chooses B, and the defense is the chapter’s argument in miniature: NPV is the monetized summary, not the decision, and the objective is access. Cost per household served: A costs 40,000 units per household, B costs 7,500, so B is the cheaper way to buy the objective; the subsidy row is not a veto but a line with an owner and a review date. The unsafe choices are choosing A on NPV alone while the objective is access, and choosing B while leaving the subsidy unowned. The evidence that would change the answer: measured evidence that served households do not actually adopt, or that the subsidy cannot be sustained, in which case the honest response is to re-scope B rather than silently fall back to A.

5. The mastery drill. The corridor’s corrected case carries the sensitivity table from this chapter: base +151; ridership 25 percent below forecast, −554; value of time 25 percent below, −421; capital cost 25 percent above, −421; operating subsidy 50 percent above, −326; opening delayed two years, −76. Identify the assumption that most changes the build recommendation, explain why the table says so, and describe the response you would put into the plan this week.

The answer is ridership, the case’s swing factor. It enters three benefit components — travel time, emissions, and safety — and the corrected case breaks at roughly a 5 percent ridership shortfall: the build recommendation survives only inside a narrow band of its own most important forecast. The response is structural, not presentational: stage the remaining funding and tie the next tranche to a measured ridership trigger in the first operating months, write the re-scope trigger before opening, and transfer benefit ownership to named operational owners now, so the measurement exists before the question matters. Capital cost is the defensible alternative if procurement is the live risk, but it must be argued from its own sensitivity logic. The unsafe responses are “the number is positive, we proceed,” and a decorative one, more slides, no trigger, no owner. Note the drill’s quiet trick: the continuation decision is robust, about +1,111 million units forward and positive even at the 25 percent shortfall, which is why the committee continues, but the fragility of the original case is exactly what tells the committee what to protect while it does.

6. Transfer question. What is the most consequential number in the last business case you saw? Where did it come from, what would have to be true for it to be wrong, and who owns the benefit it promises after handover?

The durable principle: the business case is a decision instrument built from a baseline, an options comparison, whole-life costs, benefits with owners, and honest sensitivity, and the number is a summary, never the argument. The most common next failure is the one the deck in the opening committed: approving on the summary, then filing the case until the next gate, letting ownership and measurement questions wait until handover, when they are already too late. The decisions the case produced, the conditions, the triggers, and the owners, must now be carried into the project’s authorization, which is the subject of the next chapter, “Authorize with a Charter and Govern with Clarity.”

Notes

• BlueLine Urban Mobility Program is a composite case created for this book; no real city, corridor, or company is depicted. The corridor figures in this chapter, including the corrected cash flow, the sensitivity table, and the options comparison, are author-created teaching numbers in real local-currency units, with the arithmetic shown so the reader can reproduce it.
• The United Kingdom’s appraisal guidance is HM Treasury, The Green Book: Central Government Guidance on Appraisal and Evaluation, 2022 edition. It sets the standard social time preference rate at 3.5 percent for most appraisals and requires adjustment for optimism bias, using reference-class data where available. Readers should consult the current edition for the official text.
• The cost-overrun evidence is Bent Flyvbjerg, Mette Skamris Holm, and Soren Buhl, “Underestimating Costs in Public Works Projects: Error or Lie?” Journal of the American Planning Association 68, no. 3 (2002): 279-295, a study of 258 transport projects that found rail averaging about 45 percent cost overrun and roads about 20 percent. The reference-class idea and the phrase “reference class forecasting” are from Flyvbjerg’s later work on planning under uncertainty.
• The benefit dependency network is from John Ward and Elizabeth Daniel, Benefits Management: Delivering Value from IS & IT Investments (Wiley, 2006). The layer names here are the author’s summary of the network’s logic, not a reproduction of the source’s diagrams.
• The mechanics of discounting, net present value, payback, and internal rate of return follow standard corporate finance treatment; see Richard A. Brealey, Stewart C. Myers, and Franklin Allen, Principles of Corporate Finance, 13th edition (McGraw-Hill, 2020). The caution that IRR can mislead on scale and on cash-flow sign is standard in that literature.
• Social cost-benefit practice, including the treatment of intangibles and distributional effects, follows the general treatment in Anthony E. Boardman, David H. Greenberg, Aidan R. Vining, and David L. Weimer, Cost-Benefit Analysis: Concepts and Practice, 5th edition (Cambridge University Press, 2018).
• The practice of valuing travel time in transport appraisal, which this chapter simplifies to a single per-hour figure, is standard in the field; in the United Kingdom it is set out in the Department for Transport’s Transport Analysis Guidance (WebTAG), which readers should consult in its current version.
• The value of staging and flexibility, treated here as the response to a fragile case, draws on the real-options literature; see Avinash K. Dixit and Robert S. Pindyck, Investment under Uncertainty (Princeton University Press, 1994).
• ISO 21502:2020, the general ISO guidance on project management, lists the development of a business case among the initiating activities of a project life cycle; readers should consult the official publisher for the current edition. This book is independent of ISO, PMI, and all named framework owners.