The Vogtle nuclear expansion in Georgia was supposed to bring two new reactors online in 2016 and 2017 for roughly $14 billion. Unit 3 reached commercial operation in the summer of 2023. Unit 4 followed in 2024. The final bill cleared thirty billion dollars, and the project ran more than seven years past its original date — a schedule that, according to staff testimony filed with the Georgia Public Service Commission, had been called “overly aggressive and unachievable” from inception in 2009.
The interesting number is not the seven years. It is how the seven years accumulated. Not in a single announcement. In a long sequence of small refinements — a month here, a rebaseline there, a July schedule superseded by an October schedule superseded by the next quarter’s reforecast. Each move looked defensible on its own. Together they described a schedule that had stopped being the schedule anyone in the field was actually working from.
Most construction schedules lose credibility this way. Slowly, then all at once. The float that quietly drains each month is the leading indicator. The tense conversation between the operator, the financier, and the contractor about which date is real is the lagging one. The distance between those two moments — sometimes twelve months, sometimes twenty-four — is where the credibility gap does its work, and where the outcome the owner cares about is either protected or lost.
How a Schedule Loses Belief
A construction schedule is a promise made in the language of logic ties, durations, and float. It is also a story three parties agree to tell each other — the owner, the general contractor, and the financier — about when the plant runs, when the interest clock stops, when the operator starts making product. On day one the three parties tend to believe the same story. Somewhere between day one and mechanical completion, they usually stop.
The loss of belief rarely announces itself. It shows up in the float report. A schedule that opened with sixty working days of float on the critical path loses four days in the first month, six in the second, another five in the third. Nothing has been formally delayed. The end date on the cover page has not moved. The float has just quietly gone somewhere — absorbed by a permit that arrived late, a rebar submittal that took three cycles, a switchgear delivery that slipped a week, a weather day that was not in the plan. By month six the critical path is running on zero float and every downstream activity is one bad week away from a push.
This is the pattern we watch develop on the programs that eventually crack. The monthly reports still say the date is achievable. The narrative in the executive summary still uses the same three verbs it used at award. The number under the surface — the trend line on float consumption — is already telling a different story to anyone who is reading it. The field knows first. The superintendents know. The area leads know. The scheduler running the weekly update knows. The gap between what the field is working from and what the executive report is describing is the schedule’s credibility deficit, expressed in weeks.
The Rebaseline That Buys Nothing
When the gap gets wide enough that it cannot be talked around, the standard move is a rebaseline. The logic gets reshuffled. Some activities move in parallel that were sequential. A crew count assumption gets raised. A working-hours change is written in. The new schedule prints, the end date is close to the last one, and everyone signs it. Within two months the same trend line reappears, running against the new baseline, drifting at roughly the same rate.
The reason is that most rebaselines are resequencing exercises dressed up as recovery plans. They rearrange the arrows on a page without changing anything about the field conditions producing the drift. If the permit authority is running eight weeks slow on the same class of submittal, rebaselining the submittal to a shorter cycle produces a schedule the permit authority will not actually match. If the switchgear vendor is running four weeks late on delivery and has no path to accelerate, a schedule that assumes on-time delivery is a hope, not a plan. If the trades on the floor are producing 78 percent of the productivity in the estimate, a rebaseline that assumes 100 percent going forward is fiction the moment it prints.
The Vogtle case is a public example of this pattern at scale. Testimony before the Georgia commission described a project that had rebaselined so often that the “dynamic” nature of the schedule masked the true drift — delays that looked like weeks in each update turned out to be years when the sequence was totaled up. California’s high-speed rail Central Valley segment tells a similar story from a different corner of the industry. In its 2024 business plan, the authority reported that the ongoing Central Valley contract packages had run between 2.5 and 6.75 years past their original completion dates, a weighted average of six years and three months, and the full Phase 1 route no longer had a projected completion date at all. In both cases the rebaselines were technically defensible. They did not restore credibility because they did not change the field conditions producing the drift.
One Schedule, Not Three
The first move in restoring credibility is boring and unpopular. It is to insist, at the level of the operating rhythm, that there is one schedule — and that the schedule the crews are working from in the field is the same schedule the owner reviews on Thursday and the financier receives on the fifteenth. In practice, on large programs, this is not what happens. There is a contract schedule that sits in the legal file and gets referenced only in claims discussions. There is an executive schedule with a smoothed narrative and a milestone view. There is a field schedule the superintendents actually run against, kept in the same tool but with sequences and durations the executive view has not been told about. The three drift apart, and the drift itself becomes the failure mode.
A single schedule sounds like a trivial requirement until an owner tries to enforce it. The general contractor has good reasons to keep the executive view smoother than the field view — the request for information log is uncomfortable, the productivity trend is uncomfortable, the crew ramp is uncomfortable. The owner has good reasons to accept the smoother view — the financier is on the distribution list, the board deck is due, the operator’s launch marketing is already booked against the date. The financier has good reasons to look only at the milestone view — the covenant test is on a specific date and the rest is noise. Everyone is behaving rationally inside their own frame. The schedule loses credibility anyway.
The programs that regain credibility, in our experience, do so by collapsing the three views back into one. Not by rewriting the schedule. By opening the field schedule to the owner and to the financier, on the same cadence, with the same trend lines and the same exceptions visible. The first month of that discipline is uncomfortable for everyone. By the third month the conversations at the review table are different. The argument stops being whether the date is achievable and starts being which specific field condition is producing the drift and what has to change to close it. That is the conversation that recovers schedule credibility. It cannot be had with three versions of the truth on the table.
The Leading Indicators That Matter
A credible schedule is one that reports on itself in a small number of measures the operator and the contractor both trust. The list is shorter than most program dashboards. Float consumption on the critical path, measured monthly and trended, not just reported as a snapshot. Actual crew productivity against the estimate, by area, with a trend line long enough to distinguish a bad week from a bad trajectory. Submittal cycle time against plan, because a project running two review cycles when it was priced for one is losing weeks in a place the schedule bar chart does not display. Long-lead delivery status against the promise made at award, with the vendor’s current forecast rather than the contracted date. And the number of critical-path activities running with less than a specified minimum of float, which is the earliest visible sign that the schedule is about to lose its resilience.
These indicators share a property that matters. They move before the milestone dates move. By the time a milestone slips on the cover page, the underlying trend has been visible in the leading indicators for three to six months. A program that reads the leading indicators can act while the schedule still has room to absorb the change — adding a second crew on the pacing area, resequencing a work package, expediting a submittal, activating a second source on a long lead. A program that reads only the milestone dates is reading yesterday’s news, and by the time the news arrives the room to maneuver has already been consumed.
This is the discipline the credible schedules run on. Not more indicators. Fewer of them, read more seriously, tied to a specific decision the owner is prepared to make when the trend crosses a line. Most programs collect the data. Far fewer decide, in advance, what number triggers what action. The decision made ahead of the trend is the decision made calmly. The decision made after the milestone slips is a rescue.
Regaining Credibility Mid-Build
Schedules can be recovered mid-build. Not often, and not easily, and never by producing a new bar chart with the same activities in a shorter arrangement. Recovery is a discipline, and it has a shape.
We watched a heavy industrial program run through this shape last year. Twelve months into a thirty-month build, the float had run down to zero on two critical-path chains and the executive report was still describing the end date as firm. The owner called for an outside review of the schedule against the field. The review did not produce a resequencing. It produced a list — twenty-two specific field conditions that were driving the drift, ranked by weeks-lost impact. Six of them were vendor issues. Five were submittal-cycle issues in the owner’s own review chain. Four were crew productivity issues concentrated in one area with a specific supervision gap. Three were permit and inspection cycle issues. Four were sequencing choices that had made sense at award and no longer did.
The recovery plan addressed the twenty-two items directly. The owner accelerated its own submittal review by adding a second reviewer on the two pacing packages. The general contractor moved a superintendent from an area running on rate to the area that was not. Two long-lead vendors were put on weekly rather than monthly reporting, with the owner in the room. One work package was resequenced — but only after the field conditions producing the sequence problem had been addressed, not before. The schedule was rebaselined once, six weeks later, against the new conditions rather than the old assumptions, and it held. The program landed within six weeks of its original date, and the leading indicators had returned to trend by the time it did.
The recovery worked because the conversation stopped being about the schedule and started being about the field. When the paper and the field disagree, the paper is wrong every time. Programs that recover credibility do so by chasing the field until the paper catches up, and then rebaselining to a set of conditions that actually exist. Programs that rebaseline first, and then hope the field will meet it, produce another rebaseline in ninety days.
The Schedule Everyone Believes
A schedule the operator, the financier, and the contractor all believe is not a schedule with a longer or shorter end date. It is a schedule with a shorter distance between the field and the paper. That distance — the credibility gap — is the actual variable an owner can manage. The end date is downstream of it.
Closing the gap requires three things that are unglamorous and mostly non-technical. It requires one schedule, opened to all three parties on the same cadence, with the field conditions visible rather than smoothed. It requires a small set of leading indicators, tied in advance to specific decisions the owner is prepared to make when the trend crosses a line. And it requires the discipline to rebaseline only after the field conditions producing the drift have been addressed, not before — because a rebaseline against unchanged conditions is a schedule that will fail on the same day, in a different way, ninety days later.
The operators who look back on a large build with a schedule that held are almost never the ones who ran the fewest rebaselines. They are the ones who ran the rebaselines that mattered, at the moments the field had actually changed, and who spent the months in between making sure the paper on the executive’s desk and the paper the superintendent carried into the morning huddle were the same paper. The date on the cover was the outcome. The credibility of the schedule was the program.
