Research Finance

Research Finance

The money side of research operations: building a study budget that funds what the study will actually consume, tracking spend against delivery rather than against the calendar, and deciding which research to fund when the portfolio asks for more than the budget holds.

When to use this skill

• Costing a study before a funding request, where the budget will be scrutinised line by line and a missing line item becomes unfunded work
• Answering "what does this cost per participant" — the number every funder, finance partner, and sponsor asks first
• Tracking a programme mid-flight and needing to know whether the spend is buying delivery or just buying time
• Forecasting an overrun early enough to descope rather than late enough to need supplementary funding
• Prioritising a research portfolio that asks for more than the budget available
• Defending a research budget against a finance partner who sees a cost centre and needs to see decision value

Inputs the skill expects

• Unit costs: per participant, per site, per site-month, and fixed programme costs
• Enrolment target, screen failure rate, site count, and duration
• Contingency and overhead rates your organisation applies
• For tracking: budget at completion, weighted milestones with completion, and planned versus actual spend by period
• For portfolio work: each study's cost, the value of the decision it informs, and the probability it changes that decision
• Decision deadlines — a study that lands after the decision is worth nothing

Clarify First

Before generating, confirm these inputs. If any is unknown or vague, ASK — do not assume:

• Screen failure or recruitment yield rate — budgets built on enrolled participants alone systematically under-fund screening, which is real work on real people
• Whether overhead and contingency are inside or outside the quoted number — the same study is quoted at wildly different totals depending on this, and the mismatch surfaces after the award
• The decision each study informs and its value — without it a portfolio can only be ranked by cost, which funds the cheap studies rather than the valuable ones
• Who holds the budget and what triggers a change request — determines how much contingency you need and how granular the tracking must be

Stop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.

Workflows

Workflow 1 — Build a study budget from unit costs

1. Separate costs into four blocks: per participant, per site, per site-month, and fixed. Most under-budgeting comes from costs sitting in the wrong block — site coordination billed as fixed rather than per site-month is the classic.
2. Mark which per-participant costs are incurred on screened rather than enrolled participants. Screening is charged on everyone screened.
3. Set contingency and overhead. [RECOMMENDED] 10% contingency minimum; 8% is the floor below which every protocol amendment becomes a change request.
4. Run the builder. Read the cost per enrolled participant and the cost per insight — those are the two numbers the funding conversation turns on.
5. Clear every fail. They are all cases of work the study will do and the budget does not fund.

python3 research-ops/research-finance/scripts/study_budget_builder.py \
  --input research-ops/research-finance/assets/sample_budget.json \
  --format text

Workflow 2 — Track burn against milestone delivery

1. Weight the milestones by share of total work, not by how visible they are. Enrolment usually carries 40-50% of the weight; setup milestones feel important and are cheap.
2. Record planned and actual spend by period.
3. Run the tracker. It computes earned value, cost and schedule performance, and the estimate at completion.
4. Act on the spend-versus-delivery gap, not on the spend-versus-calendar view. A programme at 36% spent and 24% delivered is heading for a 50% overrun, and the monthly finance report showing "on budget" will not tell you that.

python3 research-ops/research-finance/scripts/burn_vs_milestone_tracker.py \
  --input research-ops/research-finance/assets/sample_burn.json \
  --format text

Workflow 3 — Prioritise the portfolio by decision value

1. For each candidate study, state the decision it informs, the cost of getting that decision wrong, and the probability the study changes the choice.
2. Record decision reversibility and the decision deadline.
3. Run the prioritiser. It computes expected decision value, discounts for reversibility, zeroes out studies that arrive too late or inform decisions already made, and allocates the budget greedily by value per unit cost.
4. Take the fail findings to the portfolio review directly. "This study informs a decision that has already been made" is a conversation worth having out loud, and the ranking alone will not force it.

python3 research-ops/research-finance/scripts/portfolio_prioritizer.py \
  --input research-ops/research-finance/assets/sample_portfolio.json \
  --budget 400000 --format text

Decision frameworks

Where research budgets under-fund themselves

Interpreting the cost performance index

[PROVEN] The single most useful number in research finance is the gap between percent spent and percent delivered. A gap above 20 points is the reliable early signal of a supplementary funding request, and it appears months before the calendar-based view shows anything wrong.

Value of information

A study is worth funding to the extent that it changes a decision, and a decision is worth informing to the extent that getting it wrong is expensive.

expected decision value = decision value × P(research changes the decision)
                          × reversibility multiplier

Two studies are automatically worth zero regardless of their inputs: one informing a decision already made, and one answering after the decision deadline. Both are common, and both survive portfolio review because nobody asks the question directly.