Choose what to fund. Make the trade-offs clear.

Suppose a city has $40 million over three years to improve access to jobs. It is considering more frequent buses, fare support and new feeder routes.

“Improve transport” is too broad to compare those options. In this case, define the main measure as additional residents able to reach a relevant employment centre within 45 minutes, at the times they need to travel and at an affordable fare. The city also has minimum access obligations for low-income riders.

That definition changes the analysis. Cheaper tickets do little for someone whose bus does not run before their shift. A new route may not help if an unreliable connection pushes the total journey beyond 45 minutes.

Ask how the outcome will be measured and which obligations are non-negotiable. A project with a high average benefit cannot compensate for failing a required service condition unless the decision-maker explicitly changes that condition.

Use the same three-year period for costs and benefits. Clarify whether the $40 million covers operating spending only, whether vehicles and infrastructure are already available, and what funding is committed beyond the period.

For this example, assume all figures below include the additional operating costs over three years and that required vehicles are available. Do not compare one route's annual cost with another route's three-year commitment.

Also check delivery capacity. A programme can fit the cash budget while exceeding the supply of drivers, maintenance staff or available depot space. Record those requirements alongside each proposal before adding projects together.

This is the public-service version of the constraint check in a growth strategy: an attractive opportunity is useful only if the organisation can deliver it.

Suppose the proposals have these estimates relative to the existing service:

The frequency proposal reaches 1,000 additional residents per $1 million. Broad fare discounts reach 400. The feeder route reaches about 833 when considered alone.

Those ratios provide a first comparison. They do not tell you to allocate every available dollar to frequency. The $20 million proposal may cover all suitable corridors at that service level; the next dollar may earn much less. The feeder route may also reach a population with an access requirement that the corridors do not serve.

HM Treasury's Green Book treats the preferred option as a judgment across objectives, costs, benefits, risks and distributional effects. A single ranking is not the whole decision.

Suppose 4,000 of the feeder route's 10,000 residents are already counted in the frequency proposal. Funding both adds 26,000 unique residents, not 30,000.

The feeder therefore adds 6,000 residents to that combination, making its incremental reach 500 per $1 million once the corridor programme is selected. That remains above the broad fare proposal's standalone ratio, but it is lower than the feeder's initial 833.

Now ask whether the estimates capture connections correctly. Some residents may gain useful access only when both a feeder and a more frequent main route operate. That is an interaction to model, not a reason to subtract all shared passengers automatically.

Define the unit carefully. A person benefiting from both projects can experience two improvements, but should be counted once in a measure of unique residents newly achieving the access threshold. If the objective instead includes total travel-time savings, their additional time benefit may still matter.

Assume the feeder needs the full $12 million to run at a useful frequency over the three years. Giving it $6 million does not necessarily deliver half of the access benefit. It may create a service too infrequent to connect with work shifts.

Compare a properly designed smaller version if one exists, with its own costs and outcomes. Do not invent proportional benefits to make the budget fit.

The same issue appears in commercial cases. Half-funding a product integration may leave the product unusable; adding one employee to a team that needs three specialists may not unlock the project. Dependencies and minimum commitments belong in the allocation itself.

Suppose targeted fare support costing $4 million is needed to meet the city's stated access obligation. A feasible initial proposal is:

The broad fare discount remains unfunded. Explain why: it adds less reach per dollar under the initial estimates and is less targeted than the support required by the access obligation.

Frequency and the feeder provide an estimated 26,000 unique additional residents with access. Do not add a further reach number for the targeted support until you know whom it serves and how it interacts with those residents. It may make the assumed access affordable rather than adding an entirely separate group.

The reserve also needs a purpose. Identify the costs or uncertainties it covers, who can release it and what evidence would justify spending it. Keeping money unallocated is sensible only when it improves the overall decision compared with the best available use today.

If driver costs rise, does the portfolio still fund the promised service for all three years? If the feeder reaches only half the expected additional residents, would another option become preferable? If the access gap is mainly outside the selected corridors, does the portfolio still meet the objective?

These are practical sensitivities. Start with the assumptions most likely to change the choice, rather than varying every number by an arbitrary percentage.

Also ask what happens in year four. A three-year grant can fund a trial, but it cannot support a permanent service promise without a continuation plan. Compare recurring commitments with the realistic funding available after the grant.

Set a review point and measures tied to the objective: actual journey times, reliable access at relevant hours, affordability for priority groups and total delivery cost. Ridership helps explain demand, but a full bus alone does not demonstrate that the previously underserved population can reach work.

If evidence differs from the forecast, assess whether the cause is poor implementation or a weak premise. Fixing an unreliable connection is different from abandoning a route with no meaningful demand.

Let the desired outcome guide resource choices and the evidence you collect afterwards. Our public-sector framework develops the outcome question further.

For a commercial version, try Bank Branch Rationalization. Explain which locations to retain or change, including overlapping coverage, customer access and costs that genuinely disappear.