Implementation feasibility: capacity, cost, and rollout

A policy on paper is a promise; a working program is staff, systems, money, and a rollout that lets you correct a wrong bet.

The idea

Getting a policy adopted is only half the job. Whether it actually works depends on unglamorous things: is there legal authority to act, enough people to run it, funding that lasts the year, data and IT to track it, a realistic enforcement and compliance burden, and a timeline that isn’t fantasy?

The second move is de-risking. You rarely know demand exactly, so you don’t bet the whole program on day one. A pilot tests one slice cheaply, a phased rollout stages the load so you can add capacity, and a sunset clause makes a wrong bet expire by default instead of entrenching.

How it works

Each lever has to serve the demand the rollout creates. A pilot carries a small cohort, so modest funding is plenty; a big-bang launch drops the whole population on you at once, so every lever must be fully resourced from day one — with no slack for surprises. Here is the arithmetic the planner runs:

Fixed budget: 100 units.  Share of demand each lever must cover:
  staff .35    IT .25    enforcement .20    outreach .20

Peak demand D set by the rollout you choose:
  pilot  D = 30      phased  D = 60      big-bang  D = 100

  required_i   = D x weight_i          # capacity that lever needs
  funded_i     = min(1, spent_i / required_i)
  feasibility  = 100 x sum(weight_i x funded_i)  -  2 x overspend

Big-bang, spending [staff 28, IT 18, enf 12, out 12] = 80 units:
  required = [35, 25, 20, 20]                 (D = 100)
  funded   = [.80, .72, .60, .60]             every lever short
  feasibility = 100 x (.35*.80 + .25*.72 + .20*.60 + .20*.60)
             = 100 x (.280 + .180 + .120 + .120) = 70

Same 80 units, phased:
  required = [21, 15, 12, 12]                 (D = 60)
  funded   = [1, 1, 1, 1]  -> feasibility = 100, and you deliver
  60% now, add capacity, then finish. The budget was never the problem;
  trying to serve everyone on day one was.

To make a same-day big-bang feasible here you’d have to fund exactly [35, 25, 20, 20] — the full 100 with zero margin. That is the tell: if a plan only works when nothing goes wrong, it isn’t a plan.

When to use it

Watch out for

• Assuming authority you don’t have. If the agency can’t compel the action or spend the money without new statutory or regulatory authority, the elegant design is dead on arrival. Confirm legal authority first, not last.
• Starving the invisible levers. Staff and IT get budgeted; enforcement and outreach get cut. The result is silent non-compliance (no one checks) and low take-up (the people it’s for never hear about it) — the program “exists” but doesn’t work.
• One-time money for an ongoing program. A launch grant that funds year one and vanishes leaves a cliff. Match the funding horizon to the commitment, or plan the wind-down.
• Ignoring the compliance burden on others. Feasibility isn’t just your capacity — it’s the paperwork and cost you push onto applicants, businesses, or front-line workers. Too heavy and they route around it.
• A pilot with no exit. A pilot only de-risks if you pre-commit to what would make you stop or redesign. Without kill criteria and a sunset clause, “pilot” is just a slow, unaccountable full launch.

Worked example

An interviewer says: “The state just passed a new childcare subsidy. Walk me through whether it will actually work.” A strong answer doesn’t re-argue the policy — it stress-tests delivery. Legal authority to disburse funds and set eligibility: confirmed. Administrative capacity: who takes applications, and does the caseload math work? Data and IT: can the existing benefits system verify income, or is that a year-long build? Funding: is it recurring or a one-time appropriation with a cliff? Enforcement and take-up: how do eligible families even find out? Then de-risk: run a six-month pilot in two counties, phase the statewide rollout by region as capacity comes online, and attach a two-year sunset so the legislature has to look at the evidence before it becomes permanent. That answer shows you know a bill is not a program.