1. Scope of the Baseline
This is an earned value system for $36,008,000 of a $243,040,000 program — the internal labor envelope, and nothing else. That is a decision, not an omission, and it is recorded in §6 of the project management plan rather than discovered here.
| In this baseline | Not in this baseline | |
|---|---|---|
| Scope | Internal labor: $36,008,000, 16.6% of base | External: $180,992,000, 83.4% of base |
| What drives the cost | Hours worked, which correlate with effort | Participants dosed, sites activated, batches made — which do not |
| Who controls it | The Program Director, through the functions | Contracts, protocols and the pace of 260 independent institutions |
| Better measured by | This document | Enrolment dashboard and the tranche release mechanism |
If the program had run EVM across all $217,000,000, the largest single driver of every variance would have been how many participants had been randomized — a number the enrolment dashboard reports directly, daily, in participants rather than dollars. The EVM version would have been the same information, converted into currency, delayed by a monthly close, and harder to act on.
A control that restates another control in a worse unit is not a second control. It is the first one with more work attached, and it dilutes attention from the measure that actually moves.
2. Position at the Status Date
Baselined at Gate 4 (30 June 2026). Measured at 15 October 2026.
| Measure | Value | Reading |
|---|---|---|
| BAC — budget at completion | $36,008,000 | The internal labor envelope for the whole program |
| PV — planned value | $15,886,000 | 44.1% of the envelope should have been consumed by now |
| EV — earned value | $15,487,000 | 43.0% of the envelope's work has actually been done |
| AC — actual cost | $16,500,000 | What has actually been spent on it |
| SV — schedule variance | $-399,000 | Work worth this much has not happened when it was planned to |
| CV — cost variance | $-1,013,000 | The work that HAS been done cost this much more than it was budgeted to |
| SPI | 0.975 | Slightly behind plan — but read §4 before trusting this number |
| CPI | 0.939 | Every dollar spent has bought 0.94 dollars of budgeted work |
The cost variance is the one that matters here. $1,013,000 of overspend on $15,487,000 of delivered work is a 6.5% efficiency loss, and unlike the schedule variance it is concentrated rather than spread.
3. By Function
| Function | PV | EV | AC | SV | CV | CPI |
|---|---|---|---|---|---|---|
| Clinical Operations | $3,318,000 | $3,120,000 | $3,290,000 | $-198,000 | $-170,000 | 0.95 |
| Technical Operations / CMC | $2,867,000 | $2,810,000 | $3,480,000 | $-57,000 | $-670,000 | 0.81 |
| Regulatory Affairs | $1,866,000 | $1,880,000 | $1,845,000 | $14,000 | $35,000 | 1.02 |
| Clinical Development | $1,642,000 | $1,610,000 | $1,660,000 | $-32,000 | $-50,000 | 0.97 |
| Nonclinical Safety & Pharmacology | $1,611,000 | $1,611,000 | $1,585,000 | $0 | $26,000 | 1.02 |
| Biostatistics & Data Management | $1,510,000 | $1,455,000 | $1,520,000 | $-55,000 | $-65,000 | 0.96 |
| Quality Assurance (GxP) | $1,030,000 | $1,015,000 | $1,105,000 | $-15,000 | $-90,000 | 0.92 |
| Pharmacovigilance & Drug Safety | $968,000 | $975,000 | $960,000 | $7,000 | $15,000 | 1.02 |
| Program Management Office | $677,000 | $660,000 | $700,000 | $-17,000 | $-40,000 | 0.94 |
| Development Committee (oversight only) | $211,000 | $211,000 | $205,000 | $0 | $6,000 | 1.03 |
| Market Access & Commercial | $186,000 | $140,000 | $150,000 | $-46,000 | $-10,000 | 0.93 |
| Total | $15,886,000 | $15,487,000 | $16,500,000 | $-399,000 | $-1,013,000 | 0.94 |
The analytical method transfer to Aldergate failed acceptance criteria on the first pass (issue I-02). The remediation is visible in the budget as a contingency draw for the external cost — but the internal cost is here, in CPI 0.81, and it is the larger number of the two. Embedding a CMC team on site for months does not appear on any invoice.
This is the case for maintaining an internal-labor PMB at all. Without it, the method transfer failure looks like a $1,850,000 contingency draw. With it, the same event also cost roughly $670,000 of internal effort that was planned to be doing something else.
Two other rows are worth reading. Market Access has earned $140,000 against $186,000 planned — the largest proportional schedule variance in the table, on the smallest budget, in the function carrying the assumption that determines the business case. And Nonclinical is the only account that is complete: earned equals planned exactly, and it came in slightly under cost, which is what a finished account looks like.
4. Why SPI Is the Weakest Number Here
The schedule performance index reads 0.975, which sounds like a program very nearly on schedule. The program is 76 participants behind an enrolment curve that determines the filing date. Both are true, and the reason they are both true is the most useful thing in this document.
| Why SPI understates it | The mechanism |
|---|---|
| SPI is denominated in dollars, not time | A schedule variance of $399,000 says that work worth that much has not happened yet. It does not say how late anything is. Two months of slip on a cheap activity and two weeks on an expensive one produce the same number. |
| It only sees the 16.6% | Enrolment is external. The sponsor's own monitoring and data management effort continues at close to plan whether participants are randomized on curve or not — so the internal envelope barely registers the thing that is actually slipping. |
| SPI returns to 1.000 at completion, always | When the last work package is done, EV equals PV by definition, however late it is. SPI cannot report a late finish; it reports zero variance on the day you finish, whenever that is. |
| The critical path is invisible to it | Earned value is indifferent to sequence. A program can be ahead on SPI and still have its single binding constraint slipping, because the constraint is a date, not a dollar. |
SPI is included here because a performance measurement baseline that omitted it would look like it was hiding something, and because the calculation is instructive. But a program that manages its schedule through SPI has substituted a lagging, currency-denominated proxy for a leading, time-denominated measure it already has.
The honest summary of this section: the cost half of earned value works and the schedule half does not, and that is not a peculiarity of this program — it is true wherever the critical path is set by duration rather than by effort. In drug development it always is.
5. Forecast to Completion
| Measure | Value | What it assumes |
|---|---|---|
| EAC — estimate at completion | $38,363,000 | BAC ÷ CPI. Assumes the 0.939 efficiency seen so far is TYPICAL and continues for the remaining work. |
| ETC — estimate to complete | $21,863,000 | EAC less what has been spent |
| VAC — variance at completion | $-2,355,000 | The forecast overrun on internal labor if nothing changes |
| TCPI — to-complete performance index | 1.052 | The efficiency the REMAINING work must achieve to finish at BAC |
The program has been running at 0.939 efficiency for 43% of the internal envelope. To land on budget it now has to run the remaining 57% at 1.052 — better than it has ever managed, starting immediately, with no stated plan for how.
When TCPI exceeds CPI by more than a few points, the budget is not a forecast any more. It is an aspiration, and the honest response is to change the estimate rather than to keep reporting against a number the arithmetic has already ruled out.
The recommended position is therefore the EAC, not the BAC: internal labor is forecast to finish about $2,355,000 over its envelope. That is 1.0% of the authorized program, it is absorbable within contingency, and it is a materially better thing to say at a gate than “on budget” followed by a surprise eighteen months later.
One caveat on the EAC method. Dividing BAC by CPI assumes the variance is typical — that whatever caused it keeps causing it. Most of this variance is the method transfer remediation, which is a discrete event that is largely behind the program. An atypical EAC would forecast much closer to BAC. Both are defensible and they differ by roughly $2,355,000, which is why the assumption is stated in the table rather than buried in the arithmetic.
6. Variance Thresholds — What Actually Triggers What
A performance baseline that reports variance without defining a threshold produces a monthly number nobody acts on. These thresholds are set at Gate 4 with the baseline, and they map onto the change authority tiers in the project management plan rather than inventing a parallel escalation path.
| Trigger | Threshold | Response | Who |
|---|---|---|---|
| Routine | CPI 0.95–1.05 at account level | Reported. No action required — this is measurement noise on a 0/100 rule with long work packages. | Control account owner |
| Explain | CPI below 0.95 at account level, or CV over $250,000 | Written variance analysis: cause, whether it is typical or one-off, and the ETC impact. Not a plan — an explanation. | Control account owner to the PMO |
| Correct | CPI below 0.90, or forecast VAC over $1,000,000 | Corrective action plan with a named owner and a date by which the index should recover. | Program Director |
| Re-estimate | TCPI exceeds CPI by more than 0.05 | The BAC is no longer achievable on current performance. Table a revised EAC at the next gate rather than continue reporting against a number the arithmetic has ruled out. | Development Committee |
TCPI 1.052 against CPI 0.939 is a gap of 0.113 — past the threshold. The threshold was set at Gate 4, before anybody knew which direction the variance would run, which is the only time it is possible to set one honestly.
That is the entire value of writing thresholds down in advance. A program that decides after the fact what counts as a significant variance will always find that the current variance falls just short of it. Every threshold above was agreed by people who did not yet know whose account would breach it.
Note what the thresholds are denominated in. Three of the four are CPI, and the fourth is TCPI — all cost. There is deliberately no SPI threshold, for the reasons in §4: a schedule trigger on this program reads from the enrolment curve and the critical path, not from earned value. Building an SPI threshold here would have created an escalation route that fires on the wrong evidence and stays silent on the right evidence.
7. How This Baseline Is Maintained
| Rule | Detail |
|---|---|
| Earned value is claimed by rule, not by opinion | Work packages under 0/100: nothing is earned until the package is complete against its dictionary acceptance statement. No partial credit, for the same reason milestones are binary. |
| The baseline moves only with the scope baseline | CR-01 and CR-02 added budget; both were re-baselined into the PMB at the point the Committee approved them, and not before. |
| Actuals come from the finance system, not from the program | The Program Director does not supply the AC line. A program that reports its own actuals is marking its own homework. |
| Reported monthly, tabled at every gate | Alongside the budget and the schedule, so the three baselines are always read together. |
A package that is genuinely three-quarters done earns nothing until it is finished, so EV lags reality and SPI reads low on any program with long work packages. The alternative — percent-complete estimated by the person doing the work — produces a smoother curve and a number nobody can audit.
A measurement system that flatters the program is not measuring the program. The conservative rule is worth its cosmetic cost, particularly in an environment where the same evidentiary discipline is applied to everything else.