What does a stalled program actually cost per month?
Running cost climbs whether anything ships or not. The interesting number is not the overspend — it is the gap between what you are paying for and what anybody outside the program can actually use.
The number everybody reports
Spend against budget. It is on the front of every steering pack, it is the number the finance business partner has prepared, and it is nearly useless.
It is a measure of consumption. A program consuming exactly what it planned to consume, on schedule, within tolerance, can be delivering nothing at all — and will report green while doing it, because green means we are spending what we said we would. Nothing in that sentence is about a working system.
The failure mode is not that people lie. It is that everybody in the room is answering the question honestly and the question is wrong.
The number nobody reports
What is live.
Not what is built. Not what is signed off, not what is code-complete, not what is in UAT, not what is behind a feature flag awaiting a business decision. What a person outside the program is using today, in anger, without being asked to.
It is uncomfortable for an obvious reason: on most stalled programs the honest answer is nothing, and has been nothing for a while. A number that is zero for eleven consecutive months is not a metric anybody volunteers. But it is the only one that distinguishes a program that is slow from a program that has stopped, and those two need completely different responses.
The chart above carries no figures on the vertical axis, deliberately. The shape is the argument, and a number there would be an invented statistic dressed up as research.
Where the money actually goes
Four places, roughly in order of size and inversely in order of how often they appear on a slide.
The parallel run. Two systems, two support arrangements, two sets of reconciliation, and a data-sync job that somebody now owns full time. This is almost always the largest line and it is almost never on the program's own budget, because it landed on operations.
Coordination overhead, which grows with age. A program in month four has one weekly meeting. The same program in month thirty has a steering group, a design authority, three workstream stand-ups and a change board, all of which exist because something went wrong once. None of them will be removed, because removing governance requires somebody to say the risk has passed.
The old system, still billing. License, support, the hosting, and the two people who know it. Nobody has switched it off because the new one is not finished, and it is not finished partly because those two people are busy keeping the old one alive.
Option value on the decisions being deferred. The hardest to see and often the largest. Every month the program runs is a month the organization does not make the decision the program is standing in for.
A program that has been green for twenty-four months has not been measured for twenty-four months.
>
— Delphine Ostrowski, Founder and Chief Executive
What to do on Monday
Three things, each doable inside a fortnight, none of which needs a consultant.
Count what is live. One number, one owner, reported weekly, defined as a person outside this program used it this week without being asked. Do not let it be redefined. It will be zero at first and that is the finding, not a failure of the metric.
Price the parallel run separately. Take it off operations' budget and put it on the program's. It changes the arithmetic on every remaining decision, immediately, because suddenly finishing has a return.
Put a switch-off date on the old system and work backward from it. Not a target — a date, owned by a named person, with the retention question answered first. Almost everything a stalled program needs to do next falls out of that one commitment.
None of this makes a stalled program move. It makes it visible, which is the part that has usually been missing for a year and a half.
