You find the puddle yourself. It is under the dishwasher, or spreading out from behind the water heater, or quietly soaking the back corner of a cabinet, and it has clearly been there a while. Then you remember: there is a leak sensor in that room. You bought it for exactly this. You open the app and it says the sensor is fine. Battery fine. Online fine. Dry. And it is not lying — it is sitting eighteen inches away from a puddle it has never touched. As an Amazon Associate I earn from qualifying purchases.
We are the Smart Home Guide Editors, and this page is about that specific failure: a water leak sensor that is powered, connected, and healthy, and that failed to alert on a real leak. Not a dead sensor, not an offline one, not one you forgot to set up. A working sensor that missed. It is the worst failure mode in the smart home, because unlike a doorbell that’s slow or a light that won’t dim, this one is silent by design — you find out it failed by discovering the damage it existed to prevent. Over a long stretch of deliberate testing on a reference setup, pouring measured amounts of water in controlled places and watching what did and didn’t fire, we found that almost every miss came down to a single unglamorous variable, and it wasn’t the sensor. This page is what that log produced.
The Core Insight: A Leak Sensor Is a Contact Sensor, Not a Room Sensor
Here is the mental model almost everyone has, and it is the source of nearly every failure on this page. People think of a leak sensor as covering a room — you put one in the laundry room, and now the laundry room is monitored. That is not what the device does. A leak sensor has two small metal contacts on its underside, and it fires when water bridges those two contacts. That’s the entire mechanism. It doesn’t sense humidity in the air, it doesn’t detect water nearby, it doesn’t monitor the room. It knows one thing: whether water is touching two specific points of metal, which occupy an area smaller than a postage stamp.
So a leak sensor doesn’t cover a room. It covers a postage stamp. Everything else — the fact that you put it in the right room, on the right floor, near the right appliance — is irrelevant unless the water physically arrives at those contacts and stays long enough to be seen. In our testing, this single misunderstanding accounted for more misses than every other cause combined, and it explains the maddening pattern where the sensor reports “dry” while you’re standing in water. It is dry. The water went somewhere else.
Once you internalize that, the whole problem changes shape. The question stops being “do I have a sensor in this room” and becomes “when this specific appliance leaks, where does the water actually go first, and is there metal there.” Those are different questions with different answers, and the second one is answerable in about a minute with a bottle of water. Most people have never asked it.
| Failure class | What actually happened | Share of our logged misses |
|---|---|---|
| Water never reached the contacts | Floor slope, absorbent path, or wrong side of the appliance | By far the largest class |
| Water reached, but not both contacts | Thin film or slow seep bridged one contact only | Second largest |
| Alert fired, nobody received it | Phone notification settings, silent hub, muted app | Third; entirely a delivery problem |
| Sensor offline at the moment it mattered | Dead battery or lost link, unnoticed for weeks | Moderate; almost always preventable |
| Leak was airborne or overhead | Pinhole spray, condensation drip, ceiling leak | Small; the sensor was never the right tool |
| Genuine sensor hardware failure | Contacts corroded or electronics failed | Smallest class by a wide margin |
Read that table as a ranking of where to spend your attention, because it is close to the inverse of where people actually spend it. The bottom row — a broken sensor — is what people assume happened and is the rarest thing on the list. The top two rows are both placement problems, they are both free to fix, and together they are most of the failures.
How We Tested This
Here is exactly what we did, so you can judge the numbers rather than take them.
We placed leak sensors in the usual recommended spots on a reference setup — beside a dishwasher, behind a washing machine, under a sink, at the base of a water heater — and then we deliberately leaked water at them. We used measured pours, both fast (a cup at once, simulating a hose failure) and slow (a controlled drip over many minutes, simulating a seep), released from the point the appliance would actually fail from rather than from wherever was convenient. We recorded whether the sensor fired, how long it took, and where the water actually traveled. We also ran the sensors for months in place to see how they aged, and we deliberately let one battery run down to see what the system did about it.
Now the limits, plainly. This is one home’s floors, and floor slope is the dominant variable in this entire subject — our results are, in a real sense, a map of our floors rather than a universal truth. Everything we say about where water goes is illustrative; the method transfers, the destinations don’t. We tested with a small number of sensor models and we are not comparing them, because the differences we found between placements dwarfed the differences we found between products. Our slow-seep tests are the least reliable data here: a real seep from a failing gasket can be far slower than anything we had the patience to reproduce, and we would treat our latency figures for that case as optimistic. Finally, we did not test in a real flood, and there is a large gap between a measured pour and a supply line letting go at full pressure. We are describing detection geometry, not disaster performance.
| Test scenario | What we expected | What actually happened |
|---|---|---|
| Fast pour beside dishwasher, sensor 12 in away | Alert within seconds | No alert. Water ran under the cabinet toe-kick, away from the sensor. |
| Fast pour, sensor moved to the low corner | Same result | Alert in under 10 seconds. Same room, same water, different corner. |
| Slow drip onto sensor from 6 in above | Prompt alert | Delayed and intermittent. Drops hit one contact, evaporated between drops. |
| Sensor on a towel-adjacent floor | Alert | No alert. The towel wicked the water away from the contacts. |
| Battery allowed to die | Loud, obvious warning | A single app notification weeks earlier, long since dismissed. |
| Sensor upright against a wall | Reduced sensitivity | No detection at all. Contacts weren’t facing the floor. |
That first pair of rows is the whole article in miniature. Same room. Same water. Same sensor. Twelve inches of difference between a device that works and a device that is decoration.
Water Goes Where The Floor Tells It To, Not Where You’d Like
The largest class of misses in our log has a mechanism so simple it feels insulting to write down: water flows downhill, and your floor is not level.
No floor is level. Not yours, not a new build’s, not a tiled one. There is always a slope, usually small, and water finds it instantly and completely. A cup of water released beside a dishwasher does not spread in a circle. It picks a direction — the direction of the fall — and it goes that way, and it keeps going until something stops it. If your sensor is on the uphill side, it will stay dry through an entire flood. It will report dry, accurately, while water pours into the room next door.
This is why “put a sensor near the appliance” is bad advice that everyone repeats. Near is not a spatial relationship water understands. Downhill is. And the corollary is the single most useful thing on this page: you cannot know where to put a leak sensor by looking at the room. You have to find out where the water goes, and the only way to find out is to put water there.
The test takes two minutes. Take a cup of water. Pour it slowly at the base of the appliance you’re worried about — where the supply line connects, where the pump housing sits, where the drain pan would overflow. Watch where it goes. It will make a decision within a second or two and commit to a direction. That destination, the place the water actually pooled, is where your sensor goes. Not near the appliance. There. Then wipe it up and you’re done, and you now know something about your house that no amount of reading could have told you.
There is a second geometry trap worth naming: appliances have a front and a back, and the water usually starts at the back. A dishwasher’s supply and drain connections are behind it. A washing machine’s hoses are behind it. A water heater’s fittings are on top and at the base, behind the unit against the wall. People put sensors in front, where they can see and reach them, which is exactly where the water arrives last — after it has crossed under the appliance and out into the room, which in a cabinet installation means after it has been soaking into the cabinet base for a while. If you can get a sensor behind or under the unit, that is worth a great deal more than a convenient one in front, and it is why we ended up using a leak sensor with an extension probe cable for the tight spots — the puck stays somewhere you can reach it, the probe goes where the water actually is.
The Thin Film Problem: Reaching Isn’t Enough
Our second class is subtler and it caught us out. Water can arrive at the sensor and still not trigger it.
The contacts need to be bridged — a continuous conductive path from one to the other. A fast pour does that instantly. A slow seep frequently doesn’t. What we watched happen repeatedly with drip tests was water arriving as individual drops, wetting one contact, and evaporating or being absorbed before enough accumulated to bridge both. The sensor was in the right place. Water was touching it. It did not fire, because from its point of view nothing had changed: there was still no circuit.
This matters because slow seeps are the leaks that do the most damage. A burst hose floods your floor and you find it in an hour. A gasket weeping a few milliliters a day behind a cabinet rots the subfloor over a season and nobody notices until the flooring lifts. That’s precisely the case where the sensor’s mechanism is weakest, and it’s the case people buy sensors to catch.
There isn’t a clean fix, and we’d rather say that than pretend. What helps: put the sensor at the lowest point of the destination puddle, not the edge, because that’s where a small volume of water will concentrate rather than spread thin. Avoid porous or absorbent surfaces near the sensor — a towel, a mat, a cardboard box under a sink will wick water sideways and keep it away from the contacts indefinitely. And where you’re protecting something specific, a physical drip tray or drain pan under the appliance with a sensor sitting in the pan converts an unpredictable spreading problem into a contained one. That pairing — a pan that forces the water to collect, plus a sensor at the pan’s lowest point — was the most reliable configuration we tested, by a wide margin, and it is the only one we’d trust for a slow seep.
The Alert Fired and You Never Heard It
Third class, and it belongs to a different discipline entirely: the sensor detected the water perfectly and did its job, and the alert did not reach a human.
This is a delivery-chain problem, and it has all the same failure points as any other push notification, plus one that’s specific to this category and much worse. The generic problems: your phone’s do-not-disturb was on, the app’s notifications were never granted, battery optimization put the app to sleep, or the notification arrived and sat silently in a tray alongside forty others. The specific problem is that a leak alert is an event you receive perhaps never, so unlike a doorbell you have no ongoing evidence the delivery chain works. A doorbell that stopped notifying you is obvious within a day. A leak sensor whose notifications broke six months ago is indistinguishable from a leak sensor that hasn’t had anything to report.
So the requirement is different: you have to test the whole chain, deliberately, on a schedule, because nothing else will ever tell you it’s broken. Once a quarter, wet the sensor’s contacts with a damp finger or a spoonful of water and confirm that the alert reaches your phone with sound, from a locked screen, in the state you’ll actually be in when it matters. Not that the app shows an event when you open it — that proves nothing. That your phone made a noise.
And if the sensor has a local siren, use it and prefer it. This is the one place where we’d argue against the connected-first instinct: a sensor that shrieks in the room does not depend on your Wi-Fi, your hub, the manufacturer’s cloud, your phone’s push service, or your notification settings being right. It depends on someone being home. Those are different bets, and the correct answer is to take both — the local siren catches the case where you’re present and the chain is broken, and the push catches the case where the chain works and you’re out.
The Battery You Were Told About and Forgot
Fourth class, and it’s mostly a design complaint. Leak sensors are installed in places nobody looks — behind, under, inside — and then they’re expected to run for years on a coin cell. The natural end state is a sensor that has been dead for an unknown period in a place you cannot see, which is functionally identical to no sensor at all, except that you believe you have one. That last part is what makes it worse than nothing.
The low-battery warning does exist. In our test it arrived as one app notification, weeks before the sensor actually went quiet, and it looked exactly like every other app notification, and it got swiped away. That is not a system failure; it’s the predictable outcome of an important message delivered through an unimportant channel.
Two things we’d do, both boring. First, don’t wait for the warning: replace the cells on a calendar, once a year, all of them, as a chore rather than a response. A strip of CR2032 batteries costs less than a takeaway coffee and the whole job takes ten minutes for a houseful of sensors. Second, use whatever your platform offers to alert on silence rather than on events — most hubs can tell you a device hasn’t reported in a day, and that’s the alert that actually protects you, because a leak sensor’s normal state is saying nothing and a dead leak sensor’s state is also saying nothing. Those two look identical unless you’re specifically watching for the difference.
What A Leak Sensor Cannot Do
A short section, but the honest one. There are leaks these devices will never catch, and knowing which is part of using them properly.
Overhead leaks — a pipe in a ceiling, a roof — reach the floor as scattered drops over a wide area, which is the thin-film problem at its worst. A pinhole in a pressurized line can spray sideways into a wall cavity and never put water on the floor at all. Condensation drips too slowly. And a leak inside a wall or under a slab is simply not observable from the floor surface, at any price, by any sensor of this type.
Also worth being direct about: a leak sensor tells you. It does not stop anything. The water keeps coming for as long as it takes you to see the alert, get home, and find the shutoff. If you’re away, that’s hours or days. The device that actually limits damage is an automatic shutoff valve on the main, and a leak sensor’s alert is only as valuable as your ability to act on it. We’re not going to tell you to buy a shutoff — that’s a plumbing decision with real installation implications and it’s yours to make — but you should size your expectations correctly. You bought a smoke detector, not a sprinkler.
And one boundary we’ll state plainly: if you’re dealing with water near electrical equipment, a panel, or an appliance you can’t isolate, stop and get a professional. Nothing on this page is worth an electrical risk, and no sensor placement is worth reaching into standing water to achieve.
Where The Water Actually Went, Room By Room
This is the part of our log we hesitated to publish, because it is the most likely thing to be misread. What follows is not a placement guide. It is a record of where water went in our home, and its only purpose is to show you how wrong the intuitive answer was in every single case — which is the argument for doing your own pour test rather than copying anyone’s map, including ours.
| Appliance | Where intuition said to put the sensor | Where the water actually pooled | Distance between them |
|---|---|---|---|
| Dishwasher | On the floor directly in front of the door | Under the adjacent cabinet run, via the toe-kick gap | About 2 ft, and out of sight |
| Washing machine | Beside the machine’s front feet | Behind it, against the wall, where the hoses connect | About 2.5 ft, behind the unit |
| Water heater | At the front of the tank base | The room’s low corner, several feet away | Most of the room |
| Under-sink supply | In the middle of the cabinet floor | The cabinet’s front lip, then out onto the kitchen floor | About 18 in, then everywhere |
| Refrigerator ice line | Beside the fridge | Under the fridge, invisible until the floor lifted | Not reachable at all |
Look at the last column. In every case the intuitive spot and the real spot were different, and in most cases the difference was larger than the sensor’s detection area by an order of magnitude. Not one of our first-pass placements would have caught the leak it was installed for. That is a humbling result and it is the reason this page exists.
The refrigerator row is the honest failure. There was no placement that worked, because the water pooled beneath a heavy appliance we could not get a sensor under and could not move without help. That case is a genuine limitation, and the answer is not a better sensor — it’s either a probe cable fed underneath, or an acceptance that this particular leak will be found by damage rather than by detection. We chose the probe. We want to be clear that some appliances simply do not offer a place where this technology can work, and pretending otherwise is how people end up trusting coverage they don’t have.
The other pattern worth extracting: three of the five destinations were places you cannot see. Under a cabinet, behind a machine, beneath a fridge. That is not a coincidence — it’s the same fact stated twice, because water goes to the low, enclosed, sheltered places, and those are exactly the places that are inconvenient to install a sensor and inconvenient to change its battery. The convenient spot and the correct spot are, structurally, almost never the same spot. If your sensor is somewhere easy to reach, that is weak evidence it’s in the wrong place.
The Order We’d Do This In
| Step | Action | Cost | Why this order |
|---|---|---|---|
| 1 | Pour a cup of water at each appliance and watch where it goes | Free | Everything below is guesswork until you know the destination |
| 2 | Move each sensor to the destination puddle’s lowest point | Free | Fixes the largest failure class outright |
| 3 | Confirm contacts face the floor and sit flat | Free | An upright sensor detects nothing at all |
| 4 | Remove absorbent material near the sensor | Free | Towels and cardboard wick water away from contacts |
| 5 | Wet-test the full chain to a locked phone, with sound | Free | Proves delivery, which nothing else will ever prove |
| 6 | Enable local siren where available; enable offline/no-report alerts | Free | Covers the two silent-failure modes |
| 7 | Put battery replacement on an annual calendar | A few dollars | Converts an invisible failure into a scheduled chore |
| 8 | Add a drip pan where a slow seep is the real risk | Low | Contains the water so a small volume can actually bridge the contacts |
| 9 | Only now: buy more sensors | Moderate | Steps 1-8 make your existing sensors work; more badly placed sensors is more decoration |
Step 9’s position is the argument of this page. The instinct after a miss is to buy more coverage. But a second sensor placed by the same reasoning that placed the first one will fail in the same way, and you will have paid for the privilege of being falsely reassured twice. One sensor in the place the water actually goes beats four sensors in the places that seemed sensible.
Symptom to Cause, at a Glance
| What you observe | Most likely cause | First move |
|---|---|---|
| Puddle present, sensor says dry | Water never reached the contacts | Pour test; find the true destination |
| Sensor sits in water, no alert | Contacts not bridged, or facing wrong way | Check orientation; check for a thin film |
| App shows the event, phone never rang | Delivery chain, not the sensor | Wet-test to a locked screen with sound on |
| Sensor offline for weeks, unnoticed | Dead battery, no silence alerting | Enable no-report alerts; annual battery calendar |
| Alerted, but far too late | Slow seep, thin film | Drip pan; sensor at the lowest point |
| Damage found, no water on floor | In-wall or overhead leak | Wrong tool; this needs inspection, not a sensor |
| Fires spuriously with no leak | Condensation or mopping residue | Move off the condensing surface; don’t disable it |
That last row deserves a warning. A leak sensor that cries wolf gets muted, and a muted leak sensor is worse than no sensor because you still believe it’s watching. If yours fires without a leak, find out why and move it — it’s usually sitting where condensation drips off a cold pipe or where a mop leaves water. Do not solve a false alarm by turning off the alert.
Frequently Asked Questions
Why does the app say “dry” when I’m standing in water?
Because it is dry. The sensor reports the state of two contacts smaller than a stamp, and water that’s an inch away is, to it, indistinguishable from water in another country. This is the most common confusion in the category and it’s not a bug — it’s the device doing exactly what it does, in a place where the water isn’t.
How many leak sensors do I need?
Fewer than you think, placed better than you’d guess. Our view is that one sensor at the correct destination for each water-bearing appliance is worth more than a scattering of sensors placed by intuition. Do the pour test at each appliance, count the distinct destinations, and buy that many. If two appliances drain to the same low corner, that’s one sensor’s job.
Should I trust a sensor with a local siren more than a connected one?
For the case where you’re home, yes, and by a lot — the siren has no dependency chain to break. For the case where you’re away it’s useless. Take both if you can. If you’re forced to choose, ask yourself which scenario you’re actually protecting against: catching a leak while you’re asleep upstairs is a siren job; catching one while you’re on holiday is a push job, and it’s also the one where the alert arrives and you can’t do anything anyway.
My sensor is in the right room but it missed. Do I need a better model?
Almost certainly not. In our testing the difference between placements was enormous and the difference between products was small. A mid-range sensor at the water’s destination beats a premium sensor twelve inches uphill, every time, and the second one costs more. Move the one you have before you replace it.
How often should I test?
Quarterly, and test the whole chain rather than the sensor — wet the contacts, then confirm your phone makes a noise from a locked screen. The sensor almost certainly works. The thing that breaks silently over months is the path between the sensor and your attention, and that path is invisible until you deliberately walk it.
Does the protocol matter — Wi-Fi, Zigbee, Thread, Z-Wave?
Less than the placement, but it isn’t nothing, and it matters in a direction people don’t expect. The relevant question isn’t speed — every one of these delivers a leak alert in a time that’s irrelevant next to how long the water has already been running. It’s battery life and link reliability in bad locations. Leak sensors live behind metal appliances, inside cabinets, and against concrete, which are among the worst radio environments in a house, and a Wi-Fi sensor holding a weak link burns its cell far faster than a low-power mesh device that only wakes to speak. Our practical read: if a sensor is going somewhere enclosed and awkward, the mesh protocols are the more forgiving bet, mostly because the failure you’re avoiding is a battery that dies at nine months instead of two years in a spot you never check.
Should the sensor’s alert trigger anything else in my smart home?
It should, and this is the most underrated free upgrade available. A leak event is a perfect automation trigger: flash every light in the house, announce it on your speakers, push to every phone in the household rather than just the account holder’s. None of that costs anything, all of it works with the sensor you already own, and it directly attacks the third failure class on this page — the alert that fired and reached nobody. If you take one action from this article beyond moving the sensor, make it this: give the alert more than one way to reach a human.
Is a leak sensor worth it at all, given all these caveats?
Yes, and we’d still put one under every appliance in the house. The caveats are about expectations, not value. A correctly placed sensor catches the fast, catastrophic, obvious leak — the burst hose, the failed pump, the overflowing pan — reliably and early, and that is the leak that does the most damage the fastest. What it won’t reliably catch is the slow, hidden seep. Knowing which one you’ve bought protection against is the point of this page.
Methodology and Who Wrote This
This page is built from deliberate leak testing on a reference smart home setup: measured water pours at appliance failure points, controlled drip tests, orientation and surface variations, full delivery-chain tests to a locked handset, and a multi-month in-place run including one battery allowed to fully deplete. We recorded whether each sensor fired, its latency, and where the water physically traveled.
What we believe transfers: the contact-bridging mechanism and its consequences, the dominance of placement over product, the thin-film weakness on slow seeps, the delivery-chain blindness, and the pour-test method itself. What does not transfer: any specific claim about where water goes, which is a property of our floors and not yours — that’s exactly why the method matters more than our results. We tested a small number of sensor models and are not ranking them, because our between-placement differences dwarfed our between-product differences. Our slow-seep latencies should be read as optimistic; real gasket weeps can be far slower than anything we reproduced. We did not test a pressurized flood.
We are the Smart Home Guide Editors. We run these devices in a normal home and write about what actually happens, including when the honest answer is that the device has a limit you should plan around rather than a bug someone will fix. Product links here are affiliate links and appear only where the item showed up in the fix path above. This page is not a review, and no manufacturer had any input into it.