Why windows leak when it rains

Your window is not a seal. It is a drain, water-tested to about 33 mph and no further — and the wet patch is almost never where the water actually got in.

The call always comes on a Monday.

There was a line of storms through central Texas on Saturday afternoon — nothing anybody would put on the news, gusts in the fifties, over in forty minutes. And now there is a damp patch on the drywall under the dining room window, or a puddle on the sill that came back twice, or a stain on the trim that wasn't there in May. The window is nine years old. It has never done this. Something has obviously broken.

Almost nothing has broken. The window is doing exactly what it was designed and certified to do, and what it was designed to do is not what you think.

Your window is not a seal. It's a drain.

Here is the sentence that reorganises everything else on this page.

Water gets into every window. Not defective ones. Every one. It gets between the sash and the frame, past the weatherstrip, down the inside of the jamb, into the channel at the bottom. The assembly is built on the assumption that this will happen, and its actual job is to catch that water in a trough and put it back outside before it reaches anything that minds.

You do not have to take my word for this, and you should not, because the people who make windows have written it down. The governing installation practice in the United States is ASTM E2112, and it defines the sill pan — the tray that sits under the window in the rough opening — as a piece of flashing whose purpose is to collect water and drain it to the exterior, explicitly "including water that may enter through the window unit". That is the standard's own parenthetical. A document written to tell professionals how to install a window takes it as read that the window will pass water, and instructs them to build a tray for it.

So the correct mental model is not a boat hull. It is a roof. A roof is not waterproof either; a roof is a set of overlapping surfaces that gives water a shorter path down and out than any path inward. A window works the same way, and a window that leaks into your house is almost never a window that has lost a seal. It is a window that has lost a drain.

Thirty-three miles an hour

Now the number, which I have never once seen on a homeowner-facing page about window leaks, and which is sitting in a public document published by the window industry's own trade association.

Windows sold in the United States are rated under NAFS — the North American Fenestration Standard, referenced by both the residential and the commercial building codes. A unit gets a design pressure: the wind load it can take structurally. It is separately tested for water, and the water test is run at a fraction of the structural number. For ordinary residential and light commercial classes that fraction is 15% of the design pressure, with a floor of 2.86 psf and a ceiling of 12 psf.

AAMA publishes the conversion.¹ Here is what those pressures are in wind:

Design pressure	Water tested at	Roughly equivalent wind
15 psf	2.86 psf	33 mph
30 psf	4.50 psf	42 mph
50 psf	7.50 psf	54 mph
80 psf	12.00 psf	68 mph

Read the top row again. A code-compliant residential window at the lowest common grade is tested for water resistance at a wind of about 33 miles an hour. Not a hurricane. Not a named storm. A brisk spring afternoon in Kansas. Above that speed the standard makes no claim whatsoever about whether your window keeps water out, because nobody tested it there and nobody said they had.

The upper end is barely better. The test pressure is capped at 12 psf in the US no matter how strong the window is — about 68 mph. AAMA's own footnote points out, drily, that a category one hurricane starts at 74.

The second thing worth knowing is how the test runs.² It is four cycles of five minutes with the pressure on and one minute with it off, water spraying the whole time — and that one minute off is drainage time, deliberately built into the procedure. The test is not asking whether water enters. It is asking whether the unit sheds it fast enough between gusts.

This is not a scandal, and I want to be careful here, because it would be easy to write it as one. A window has to open. It has to let light through, cost something a person will pay, and meet egress rules. It is a hole in your wall that you have negotiated down to a manageable size. Thirty-three miles an hour is not negligence; it is the honest edge of what a moving assembly in a $400 price class will do. The scandal is only that nobody tells you, so the first time the edge gets crossed you assume a failure that never occurred.

Who wrote that number, and why it's still the honest one

The table above comes from a white paper titled Storm-Driven Rain Penetration of Windows, Skylights and Doors. It was released in 2005, updated in 2018, and it names its intended readers in the opening paragraph: homeowners, distributors, builders — and insurance adjusters.

Sit with that last one for a second. This is a document published by the trade association of the companies that make the windows, written to be handed to the person deciding whether a claim gets paid, and its argument is that a window which leaked during a storm is not necessarily a defective window.

Every word of it is true, and that is what makes it interesting rather than cynical. The physics is right, the arithmetic checks out against its own table, and the conclusion follows honestly from both: a product tested at 33 mph cannot fairly be called defective for admitting water at 60. But you should know what you are holding. It is a liability document that happens to be correct, and its correctness is doing work for somebody.

So here is the part worth taking from it. The same paper that protects the manufacturer is the only place that will tell you the real number. No salesperson opens with tested for water at less than half a hurricane. No brochure prints that table. The industry's defensive document is, by a distance, the most useful consumer document on this subject — because in order to defend the product it first had to describe it accurately, and nothing else in the entire literature does.

We don't build windows here and we don't underwrite houses, so there is nothing in it for us in either direction. The number is real. So is the reason you have never seen it.

Where the water appears is not where it got in

This is the part that will save you the most money, and it comes straight out of the same industry paper.

AAMA's list of potential sources of storm water at a window opens by pointing away from the window entirely: the attic, the roof, the soffit, a wall penetration like an exhaust fan. Water enters up there, runs down the inside of the wall or along a roof member, and exits at the first interruption in the cavity it meets — which is very often the rough opening of a window, because a rough opening is the biggest hole in that wall.

The window is not the leak. The window is the drip point.

Look at the search results for this problem and the single most common phrasing people use is some version of water is coming in at the top of my window. And a leak at the head of a window is the one place where the odds most strongly favour the water having come from somewhere else. Water does not enter a window at the head and run uphill to the top of the frame. It falls. Something above it put it there — flashing lapped the wrong way, a failed kick-out at a roof-wall junction, a gutter overshooting in heavy rain, a soffit vent with a clear path in. All of those present at the window, because gravity presents everything at the window.

The test costs nothing: go outside during the next real rain and look up. Not at the window — above it. Then, on a dry day, run a hose on the wall a metre above the head and see if you can reproduce it without wetting the unit at all. If you can, no amount of work on the window will fix anything.

Reading your own leak

Four patterns, and they separate cleanly.

Water in the track that stands there for a day, then overtops the inside lip. This is a blocked weep, and it is the most common thing on this list by a distance. The channel is doing its job — it caught the water — and the drain at the bottom is packed with the compacted mineral-and-insect paste that accumulates in every window track over a few years. In ordinary rain it never fills, so you never know. In a storm it fills in twenty minutes.

Water at the sill corners, after wind, with a clean track. Suspect the frame joints or the pan under the unit. Wind-driven rain gets pushed into mitred corners, and if there is no pan below — or the pan has no upstand at the back — there is nothing to stop it running inward.

Water at the head. See above. Look up.

Water only on one elevation, always the same one, only in driving rain. That is your prevailing storm wall, and what you are seeing is the pressure difference doing what pressure differences do. It is the elevation to spend money on and the only one worth testing.

There is a fifth pattern people rarely consider, and it is a genuine one: the pane fogged and the sill is wet because they are the same event. A failed insulating-glass seal puts water inside the unit, and a saturated one weeps at the glazing bead. If the glass is hazed between the panes and no amount of cleaning touches it, you are not looking at a rain leak.

And a sixth that is not a leak at all, though people ask about it in the same words: cold air at the window, and the sill bone dry. Leaking gets said about air as readily as about rain. If there is no water in it, a drafty window is the other diagnosis — four causes, only one of which caulk touches, and an incense stick settles which one in about thirty seconds.

The caulk gun is how a leak becomes rot

The near-universal advice for a leaking window is to seal around it, and it is close enough to sensible to be dangerous.³

Caulk the head. Caulk the jambs. Do not caulk the sill. That is not my preference; it is the instruction in the US Department of Energy's Building America guidance, which states it about as plainly as it can be stated — "Do not caulk across the sill" — and then says to seal the top and the sides of the opening and leave the bottom open. The reason is the whole of this article: the bottom is the way out. Seal it and the tray you have built under your window stops being a tray and becomes a bathtub with your wall framing at the bottom of it.

Manufacturers say the same thing at the level of the unit itself. Pella's installation instructions tell the installer to interrupt the exterior sealant bead with a two-inch gap at every weep hole, and to keep sealant off weep hoods entirely. JELD-WEN's install sheet goes further and ties it to money: they do not endorse installing into a barrier-type wall unless there is a sill pan, integrated with through-wall flashing, that can drain to the exterior — and failing that, a warranty claim may be denied.

So the advice from the internet is to seal the drain, and the advice from the people who built the thing, printed in the box it came in, is that sealing the drain voids the warranty.

If you're the one holding the squeegee

Two reasons this article is on a window cleaning site rather than a contractor's.

The first is that keeping the drain open is a cleaning task, and the manufacturers say so in the cleaning section. Pella's care instructions run through what to use on the frame, then the tracks, and then — in the same breath, in the same list — "Keep weep holes open and clear of obstructions." AAMA's storm paper names clogged drainage and weep holes in its maintenance paragraph as a contributor to storm leaks, alongside the age of the unit and the state of the weatherstrip. Nobody else is going to do it. The roofer is not looking at your weep holes. The window cleaner is the only trade that touches every window on the building on a schedule, and clearing a weep takes four seconds with a piece of trimmer line.

The second is that we cause some of these. Water at pressure will go places rain never would: blasting a window with a pressure washer drives water through the weatherstrip and past the drainage plane at a pressure the assembly was never rated for — the water test tops out at 12 psf, and a consumer machine at the nozzle is roughly two hundred times that. The unit had no chance. And then the leak shows up on Monday, and the invoice is dated Saturday, and you get the call.

Which is why the diagnosis matters to you more than it does to the homeowner. Being able to walk somebody through this at the door is the difference between a callback and a claim — the same argument as with scratches, for the same reason.

What actually fixes it

Clear the weeps first. It's free, it's four seconds a window, and it is the single most likely cause on the list. Then wait for the next storm.

Look up before you spend anything. A head leak that a hose can reproduce with the window dry is not a window job at all.

Get the caulk off the sill if somebody put it there. Cutting a bead out of the bottom of a window frame is a ten-minute job that reverses years of slow damage.

Stop pressure-washing windows. There is no version of this that is worth what it risks.

And treat a genuine storm leak as information, not as a defect. If water comes in at 33 mph, the drain is off. If it only comes in at 60, the window is outside its rating and always was — and the honest options are a higher-grade unit, a better-flashed opening, or living with it.

The reason I know the number by heart is that my territory prints it on the calendar. Central Texas is outside 33 mph with rain behind it several times a spring, which means every window I service spends a few days a year past the last point anybody made a promise about it. That is not a Texas problem. It's just that Texas tells you sooner. The window in Ohio is rated the same; it's simply waiting for a worse Saturday.

Notes

Where the 33 mph comes from, and why it is not a scare number. AAMA's white paper on storm-driven rain carries a conversion table: design pressure, the water test pressure derived from it, and the wind speed that pressure corresponds to. The conversions run on the standard engineering relationship between wind velocity and the pressure it exerts — pressure in pounds per square foot is roughly 0.00256 times the square of the speed in miles per hour. Check it against their own rows and it reproduces them: 2.86 psf returns 33.4 mph, 7.50 psf returns 54.1, 15.00 psf returns 76.5, and the table prints 33, 54 and 77. The paper is careful to say the conversions are for understanding rather than code compliance, and that is fair. But the number is not a rhetorical device. It is the industry's own arithmetic, in the industry's own document, and it is the reason a window can be entirely code-compliant, entirely undamaged, and still let water in on a Tuesday.
The four-cycle test, and the one minute that gives the game away. Water penetration for most residential grades is run to ASTM E547, which sprays the outside of the window at a fixed rate while a fan holds a pressure difference across it. What is interesting is the shape of the cycle: five minutes with the pressure on, then one minute with the pressure released — and the water still spraying. That one minute is not a rest. It is drainage time. The test is built around the assumption that the unit will take on water during the pressurised phase and needs an interval to shed it before the next one, and a unit passes if nothing crosses the innermost plane by the end. Read that sequence and the whole philosophy of the product falls out of it. Nobody is testing whether water gets in. They are testing whether it gets back out fast enough that you never find out.
Why the caulk advice is so durable despite being wrong. It is worth being fair to it. Sealing the head and the jambs of a window is correct — that is where wind-driven water is pushed into the gap between frame and opening, and a sealant bead there is doing real work. The instruction is not "never caulk a window." It is "the bottom is not the same as the sides," and that distinction is invisible to anyone reasoning from the general principle that gaps are bad. It also survives because the failure is slow and silent: caulk the sill, and for two years nothing happens, because in ordinary rain the assembly never takes on more water than the frame can hold. The bill arrives later, in a different form, usually as soft trim or a stain on drywall, and by then nobody connects it to a tube of sealant from three summers ago.
A note on old windows, which mostly don't do this. A pre-war double-hung has no weep holes, no sill pan and no drainage plane, and it is not leaking into the wall — because it was never sealed in the first place. Wood, putty, a drip kerf under the sill and a frame that breathes: water that gets in leaves again through the same generous gaps it arrived by, and the assembly dries because air moves through it constantly. That is also why the house was cold. The modern window is a far better product by every measure a homeowner cares about, and the price of it is that a sealed assembly has no tolerance for a blocked drain. Tightening a building removes its margin for error along with its draughts, which is the same trade documented on this site in the context of interior condensation.

Sources

AAMA SER-1-18, Storm-Driven Rain Penetration of Windows, Skylights and Doors — White Paper — Table 1's water-test-pressure / wind-speed equivalents; the rule that R, LC and CW performance-class windows and doors and all unit skylights are water-tested at 15% of design pressure with a 2.86 psf minimum and a 12 psf maximum (AW class at 20%); the document's own stated audience of homeowners, distributors, builders and insurance adjusters; and the exclusion of storm-surge and flood water from its scope. Cited under AAMA, the publisher in 2018; AAMA became part of FGIA in 2020.
FGIA, NAFS Overview — AAMA/WDMA/CSA 101/I.S.2/A440 — the material-neutral fenestration standard that serves as the basis for product certification required by the IBC and IRC; current edition NAFS-22
ASTM International, E2112 Standard Practice for Installation of Exterior Windows, Doors and Skylights — the drainage plane and sill-pan detail that decides where water goes
ASTM International, E331 — Standard Test Method for Water Penetration of Exterior Windows, Skylights, Doors, and Curtain Walls by Uniform Static Air Pressure Difference — the uniform-pressure leak test, and the standard's own caution that the result depends on installation, adjustment and the rigidity of the supporting construction, not on the window alone
ASTM International, E547 — Standard Test Method for Water Penetration of Exterior Windows, Skylights, Doors, and Curtain Walls by Cyclic Static Air Pressure Difference — the cyclic counterpart to E331, for assemblies whose leakage only shows up under repeated pressure
US DOE Building America Solution Center (PNNL), Windows and Doors are Fully Flashed — “Caulk the outside edges of the head and side jambs and install the window … Do not caulk across the sill.” Pan flashing with a back dam at the sill, side flashing lapped over the pan, top flashing over the side.
JELD-WEN, Installation Instructions for Vinyl Windows without Nailing Fin (JII011) — carries the line in capitals: “Ensure weep holes/channels are clear of debris for proper water drainage. DO NOT seal weep holes/channels.” It also tells the installer to notch any sill adapter under each weep hole and to leave gaps in the exterior sealant bead at the weeps, and warns that failing to follow the instructions may result in denial of warranty claims.