When Did That Number Last Change?

Auditors from the Government Accountability Office spent seventeen months inside the acquisition machinery of U.S. Special Operations Command, and one piece of the work was mechanical. Open the online portal that the command's own policy names as the place where current program information lives. Look up what the costliest programs expect to spend. Nine programs were in scope and eight of them were required to keep that figure on file. GAO reported that the portal held no total cost estimate data for any of the eight.¹ The program offices produced the numbers when asked directly, and two of them took several months to do it.¹

While that field sat empty, the command was reporting that eight of the nine programs were meeting their cost goals.¹ Nobody falsified anything. The policy required program managers to keep current data in the portal, and it never specified how often, which GAO named as part of the reason the data was missing.¹ An instruction to stay current with no interval attached is an instruction to do it eventually.

A number nobody was obliged to touch

There is a real difference between a figure that has held steady because the work is under control and a figure that has held steady because no one has had a reason to open it. On a status page they are the same picture. GAO's own cost estimating guide sets the expectation without much ambiguity, saying that a program should be monitored continually by comparing planned against actual performance, and that the estimate itself should be updated with actual costs so that it remains relevant.² The rule existed. What was absent was anything that made the rule fire on a particular Tuesday.

Something did fill the space. The officers responsible for oversight told GAO they do not routinely require programs to report cost performance against baseline in the portal, and that they concentrate instead on whether programs obligate and expend their funds in a timely way and whether future funding needs have been planned out.¹ That measures something genuine. It measures whether money is leaving the building on the schedule somebody wrote down. A program can spend exactly what it forecast, exactly when it forecast it, and still be assembling a thing that will need half again as much before anyone can use it.

The yardstick moves and the delay stays where it is

Five of the nine programs were behind, with delays running from roughly seven months to almost five years.¹ Officials from four of the delayed programs told GAO they were on track to achieve their cost goals regardless, and three of those four credited fixed-price contracts for holding the line.¹ Then there is a note under one of GAO's tables that repays a second reading. Armed Overwatch was late by two years and ten months against its full-rate-production date. The SEAL Delivery Vehicle Mk 11 was late by four years and nine months against its date for full operational capability. Command officials said both programs established new dates, and against those new dates both are currently on time.¹

An unchanged number and a controlled number look identical on a dashboard. What separates them is a question about when the figure was last touched and who touched it, and most executive review rituals have nowhere to put that question.

Where fresh numbers actually come from

GAO also asked each program how many of twelve leading practices for iterative product development it had fully implemented, and most of the programs running late reported fewer than the programs that were not late.¹ Four of the five delayed programs reported using three of the twelve. Three of the four programs that came in early reported nine, ten, and eleven.¹

I would be careful about reading that as an argument for a methodology, because the more useful reading is about where evidence comes from. A program that hands something usable over every few months is forced to restate what the thing costs and what it does, every time, because somebody is about to take delivery and will ask. The restatement is a byproduct of the handover rather than an additional chore. A program that plans once and delivers at the end has nothing compelling a restatement until the end arrives, and by then the estimate has aged past the point where refreshing it could change a decision anybody still has open. The cadence of delivery sets the cadence of what you know.

A couple of weeks ago I wrote about safety leaders blaming clumsy software for the near misses their people were not filing, and I have no interest in relitigating that argument here, because this one has a different shape entirely. In the SOCOM case the duty to report was real, the system was built and running, and the missing piece was an interval.

So the question to carry into your next program review is small and a little rude. Ask when the number in front of you last changed, who changed it, and what would have had to happen for it to move. If nothing has touched it since the program was approved, what you are looking at is a memory with a date printed on it. Take a minute with whatever you review most often and see how much of it has not moved in a year, and whether you would have noticed.