Every Builder Promises Six Months. The Data Says 6.7. Nobody Told You About the Missing 13 Days.

Three weeks from substantial completion, my plumber tells me the rough-in is 95% done, which is what he told me last week and the week before that, while the actual completion percentage crept from 85% to 90% to the 95% where it has now parked itself like a car with a dead battery in the driveway of a house that was supposed to close eleven days ago. The percentages climb in smaller and smaller increments, like Zeno's paradox applied to copper pipe, and I write the same note in my project log that I have written on every residential build for two decades: "Last 20% of plumbing. Allow extra."

I never had a number for "extra," because nobody in this industry had ever measured the gap between "almost done" and done with enough rigor to turn a gut feeling into a dollar figure that you could put on a spreadsheet and hand to a homeowner who deserves to know what they are actually paying for when their builder says "two more weeks" for the third consecutive month. Now somebody has.

The First Benchmarks That Don't Come From a Vendor's Slide Deck

On June 25, a company called Buildots launched the Intelligence Lab, billed as the construction industry's first AI-powered research hub producing freely available benchmarks from aggregated, anonymized project data. Buildots makes hard-hat-mounted cameras that capture 360-degree images of construction progress and compare them against BIM models, and their client list reads like a general contractor's hall of fame: Turner Construction, JE Dunn, Intel, HOCHTIEF, Bouygues, all Fortune 500 contractors who have given Buildots access to the kind of schedule-versus-actual data that most project managers keep locked in a filing cabinet labeled "do not share with the owner."

I am skeptical of construction tech companies bearing data (Buildots sells the cameras that detect schedule slippage, so alarming schedule findings are not exactly against their interests), but I have spent twenty years watching software vendors cherry-pick the numbers that sell licenses while burying the ones that reveal how rarely their tools get used after the first ninety days. Buildots did something different: they published findings with methodology notes, and they published the findings that make their own clients' performance look mediocre, which is either brave or reckless depending on how many of those clients renew their contracts after reading the report.

One finding stopped me cold, and I have been running the math on it ever since.

Anybody who has run a residential project just read that and exhaled, because they have known this in their bones for years without ever having a number to attach to the feeling, and the feeling has always been that the last stretch of every build operates under a different physics than the first 80%, a physics in which time dilates and tasks multiply and the finish line recedes at exactly the speed you approach it.

The Math Nobody Wanted to Do

NAHB's 2026 survey puts the average single-family construction timeline at 6.3 months from start to completion for homes built for sale, with construction loan rates averaging 7.5% APR. If the long-tail effect adds seven excess percentage points across all activities in the build — 27% of duration consumed minus the 20% of work remaining — that extends the overall schedule by roughly 7% of 6.3 months, which works out to 0.44 months, or about 13 days, stretching the average build to 6.7 months rather than 6.3. That does not sound catastrophic until you run the carrying cost on an active construction draw.

On a $400,000 draw at 7.5% APR, the interest alone is $82 a day. Multiply by 13 and you get $1,069 per home that vanishes into the gap between "almost done" and actually done, a sum that nobody writes on a change order and nobody bills for and nobody discloses to the buyer whose rate lock is sweating because closing slipped two weeks past the date the builder promised with a handshake and a confidence that was, per the data, structurally unjustified from the day the schedule was written.

Census data from January 2026 shows a seasonally adjusted annual rate of 935,000 single-family starts. Multiply 935,000 homes by $1,069 and you arrive at roughly $1 billion per year, industry-wide, from one structural inefficiency that had never been measured until a company mounted cameras on hard hats and let the AI count the days that everyone else had been rounding down to zero. (Buildots measured this in commercial construction; whether the exact ratio holds for residential is unproven, which I address below, but the mechanism is the same.)

Schedule Adherence: The Number Your Builder Won't Give You

Buildots published schedule adherence rates by project type, and the numbers read like a descending staircase of broken promises:

Healthcare hits milestones 65% of the time, data centers 57%, commercial and industrial land in the low-to-mid 40s, and education falls below 39%.

Residential construction is not on the list, not because residential is better but because nobody collects the data. No production homebuilder publishes schedule adherence metrics, no custom builder I have worked with in twenty years has ever calculated the percentage of milestones they hit on time, and if you ask your GC what their schedule adherence rate is, they will look at you the way a dog looks at a ceiling fan.

If residential falls in the commercial range, and I would bet good money it does given that residential crews face the same weather, the same inspection bottlenecks, the same trade labor shortages, and a version of the same MEP coordination problems that drive commercial schedules sideways, then your 6.3-month build has somewhere between a 40% and 55% chance of hitting its planned milestones on time. You are not told this when you sign the contract. You are told "six months, weather permitting," and then weather happens, and inspections happen, and the tile guy's appendix happens, and by month eight you are wondering why you ever believed a timeline that the data, had it existed, would have told you was optimistic from the moment it was printed.

Why the Last 20% Takes 27%

Buildots identified a 20% to 50% gap between planned weekly MEP output and actual delivery, with top-performing mechanical, electrical, and plumbing teams moving three times faster than average ones, a disparity that compounds as trades stack up on each other in the final months of a residential build where framing delays push insulation which delays drywall which delays trim which delays paint which delays the final clean which delays the certificate of occupancy which delays the closing that was supposed to happen on the date your buyer's mortgage rate lock expires.

Every activity inherits the long tail of every activity before it. By the time you reach substantial completion, you are not dealing with one 13-day delay but with the statistical residue of forty overlapping long tails compressed into whatever time remains before the buyer's lender starts charging rate-extension fees that nobody budgeted for because nobody knew the delay was structural rather than accidental.

Twenty years of running projects taught me this intuitively, but intuition does not survive a budget meeting, and "I've seen this before" does not persuade a homeowner whose carrying costs are climbing while their builder insists that substantial completion is two weeks away for the fourth consecutive time. What the industry lacked was proof that this was not laziness or bad scheduling or a particular crew having a bad week, but a structural feature of how construction work finishes, documented across enough projects and enough geographies that the pattern is no longer deniable by anyone who bothers to read the data.

Tasks accelerate in the middle and decelerate at the end because the easy work gets done first and the hard work, the work that requires coordination between trades who have already moved their best crews to the next job, concentrates at the finish. Inspections introduce waiting periods that pile up hardest in the last few percent because you cannot close a wall until the inspector signs off and the inspector has thirty-seven other projects requesting the same visit on the same day. Punch list items are individually trivial and collectively enormous, each one requiring a phone call to a subcontractor who has mentally moved on, a visit to verify, and a callback when the fix reveals a second problem that was hiding behind the first.

What a Homeowner Should Actually Do With This

If you are building a home, or about to sign a contract with someone who promises to build one in a timeframe that the data suggests is structurally optimistic, here is what this information is worth to you in practical terms.

Ask your builder to add 13 days to the construction schedule, not as a vague contingency buffer that everyone pretends does not exist and nobody defends when the owner asks why the timeline is "padded," but as a named line item called completion drag, backed by the Buildots data, explained in a sentence: "The final 20% of construction activities takes 27% of their duration, per the first AI-powered benchmark study of construction schedules published in June 2026." If your builder refuses to discuss it, or has never heard of the data, you have learned something useful about how they plan their projects, which is to say they plan them the way everyone has always planned them, on optimism and precedent and the assumption that this time will be different even though it never is.

Budget the carrying cost explicitly: on a $400,000 draw at current rates, 13 days costs $1,069 in interest alone; on a $600,000 draw, $1,604; on a million-dollar custom build, $2,671. These are not catastrophic numbers relative to the total project cost, but they are real numbers that belong in your contingency calculation rather than hiding in the gap between the schedule your builder gave you and the schedule the data says you will actually experience.

Ask for milestone adherence data from previous projects. Your builder has this information even if they have never assembled it, because every project has a planned completion date and an actual completion date, and comparing those two numbers across the last five homes they delivered will tell you more about their scheduling accuracy than any conversation about process or experience or how many years they have been building houses in your county.

The Counterargument, Stated Honestly

Buildots collected this data from commercial construction: data centers, hospitals, office buildings, projects where fifty-person MEP teams coordinate across building systems that would fill a residential house three times over. Residential is different in ways that matter. Shorter timelines, smaller crews, fewer mechanical interdependencies. A three-person plumbing crew roughing in a 2,400-square-foot house does not face the coordination overhead of a hospital wing where ductwork, fire suppression, low-voltage, and medical gas all compete for the same ceiling cavity, and the 20% to 50% MEP productivity gap that Buildots measured may genuinely overstate the residential version of the problem.

But the long-tail effect is not about project scale. It is about how human beings finish tasks, how inspections create mandatory waiting periods that disproportionately affect the final stages, and how the last trade on any job site is always competing for labor with every other project in the county that is also in its final 20% and also promising its buyer that substantial completion is two weeks away. I have watched this pattern repeat on a $180,000 starter home in a subdivision and a $2.4 million custom build on a hillside lot with a cantilever that required three engineering revisions. Dollar amounts changed, but the shape of the delay was identical.

What We Still Do Not Know

Buildots has not published residential-specific data. My $1 billion industry estimate assumes every single-family home experiences the average long-tail effect, which overstates aggregate impact because some builds run ahead and some builders are genuinely better at scheduling. Schedule adherence rates for residential construction remain unknown. Nobody has independently verified Buildots' data aggregation methodology. And the three-times productivity gap between top and average MEP teams may partly reflect project complexity rather than crew quality.

I am not asking you to take these numbers as gospel. I am asking you to notice that, for the first time, the numbers exist at all.

For twenty years I have written "allow extra" in my project logs. I can finally write a number next to it. Whether that number is $1,069 or $800 or $1,400 matters less than the fact that the industry's most persistent schedule pathology now has a measurement, published by a company that sells the cameras detecting the problem (and knows it), available for free to anyone who wants to check it.

The last 20% takes 27%. Plan accordingly.