Home Assistant Sensors: What Should Your Smart Home Actually Measure?

One of the easiest ways to overcomplicate Home Assistant is to start with the hardware. You see a new sensor, integration or ESP32 project and immediately start thinking about where you could use it.

I think that question is backwards.

A better place to start is: What information about my home do I wish Home Assistant knew?

That small change in thinking makes a big difference. Instead of collecting sensors because they are interesting, you start collecting information because it helps you understand what is actually happening in the house. Once Home Assistant can reliably understand what is happening, useful automation becomes much easier.

That is really the foundation of a smart home.

A Smart Home Has to Sense Before It Can Automate

We spend a lot of time talking about automations: turning on lights when someone enters a room, closing the garage door if it has been left open, sending an alert when water is leaking, or adjusting heating and cooling based on occupancy.

Every one of those automations depends on Home Assistant knowing something first. It needs state, context and, most importantly, information you trust. An automation can only be as good as the information feeding it.

I have started thinking about this in three layers: Sense → Automate → Interpret. First, understand what is happening. Then build predictable automation around that information. Only after those pieces are working does it make sense to ask whether AI can add useful interpretation.

This article is about that first layer.

Start With the Problem, Not the Sensor

Instead of asking what sensor you should buy, walk around the house and think about what you repeatedly check manually, what problems worry you when you are away, what equipment you wish gave you better information, and what changes you would want to know about immediately.

Those questions lead to better Home Assistant projects. A temperature sensor is interesting. Knowing your freezer is getting warmer than it normally does is useful. A power sensor is interesting. Knowing equipment is consuming electricity when you thought it was shut down is useful. A contact sensor is interesting. Knowing the garage door is still open after everyone has gone to bed is useful.

The sensor is not really the product. The information is.

Water: Know Before the Damage

Water is an easy example because the value is obvious. You do not want to discover a leak because the basement carpet is wet. You want the house to notice first.

That might mean placing sensors near water heaters, washing machines, sinks, sump pumps, HVAC equipment or other places where water would be a problem. The first goal does not need to be complicated. You simply want Home Assistant to know whether water is somewhere it should not be.

Once that condition is observable, you can decide what should happen next. Home Assistant might send an alert, notify several people, sound an alarm or eventually close a compatible water valve. Notice the order: first make the problem visible, then decide how the house should respond.

HVAC: Measure the Condition, Not Just the Calendar

HVAC maintenance is another good example. A traditional filter reminder says it has been 90 days, so change the filter. But 90 days does not actually tell us whether the filter is dirty. One house may run its HVAC system constantly while another barely uses it.

What we really care about is the condition of the system. That could mean measuring runtime, airflow, differential pressure, supply and return temperatures, or some other characteristic that helps tell us whether the system is behaving normally.

Gavin Campbell and I talked about exactly this kind of problem in HGG690, including using differential pressure to get a better indication of HVAC filter condition.

That project stuck with me because it captures the real value of Home Assistant and ESPHome. You are not necessarily making the furnace smart. You are giving Home Assistant information the furnace never exposed before.

Pumps: Behavior Can Matter More Than On or Off

A sump pump does not need an expensive connected controller to become useful to Home Assistant. Sometimes the important questions are simply whether it is running, how long each cycle lasts, how often it runs and whether that behavior has changed.

That history is where things get interesting. A pump running every four hours might be completely normal. The same pump suddenly running every five minutes tells you something has changed. Nothing has necessarily failed yet, but the house is behaving differently.

That is exactly the kind of change I want Home Assistant to notice.

Temperature and Humidity Need Context

Temperature sensors are probably among the most common smart-home devices, but a single reading often does not tell you very much. If the garage is 91 degrees, is that unusual? Maybe it is 98 degrees outside. Maybe the garage door has been open. Maybe it was 75 degrees twenty minutes ago because equipment just started running inside.

Now compare that with garage temperature + outside temperature + garage-door state. That starts to tell a story.

The same idea applies elsewhere. An attic temperature becomes more meaningful when compared with outdoor temperature. A network-rack temperature becomes more useful when you know what equipment is running. A basement humidity reading becomes more useful when you can see whether it has been climbing for three days.

Context is what turns sensor readings into information.

Energy Monitoring: Measure Things That Can Change a Decision

We spent a fair amount of time on energy monitoring in HGG691 with Phil Hawthorne.

It is an area where you can go very deep very quickly. Whole-home monitoring, individual circuits, smart plugs, computers, HVAC, solar, batteries and EV charging can all become sources of data. Before long, you can measure almost everything.

But should you?

I think the same rule applies here: measure something because knowing the answer might change what you do.

How much power does the studio really use? What happens when the large computers are shut down? Is an appliance drawing power when nobody expects it to? Did a device suddenly begin using more electricity than normal?

I recently went through this myself. My studio could be pulling roughly 600 watts with everything running. By shutting down equipment I did not need all the time and doing more of my everyday work from the Mac Mini, I could get that closer to 60 watts.

That number changed my behavior. That is the point. A dashboard full of numbers nobody ever acts on is less interesting.

The Home Assistant Podcast recently featured a similar path. What started as an attempt to measure an old refrigerator’s energy use grew into a much more data-driven Home Assistant installation involving energy monitoring, solar forecasting, batteries, ESPHome and other measurements.

I like that progression because it starts with a real question: How much energy is this thing using? The answer creates the next question, and the system grows from there. Home Assistant’s Energy documentation follows the same general idea: collect useful consumption, production and storage data, then use it to understand where energy is going and make better decisions.

Presence: Motion Is Not the Same as Occupancy

Presence sensing is another place where defining the question matters. A motion sensor tells you something moved. It does not necessarily tell you someone is still in the room.

For a hallway light, motion may be enough. For office lighting, climate control or room-level automation, occupancy may matter much more. That could mean combining motion, mmWave presence, door state, device presence and time, but the goal is not to install the most sophisticated sensor possible. The goal is to use the simplest information that answers the question reliably.

Outdoor Sensors: Measure the Need, Not Just the Schedule

Outdoor automation is another place where sensing can replace assumptions. Watering every Tuesday morning is a schedule. It does not necessarily mean the lawn or garden needs water.

If you care about soil moisture, recent rainfall, temperature, reservoir level or whether irrigation actually ran, measuring those conditions gives Home Assistant something much more useful than a calendar. Gavin and I talked about some of these ideas during HGG690 as well.

A schedule tells Home Assistant what you planned. A sensor can tell it what actually happened.

Sometimes the Best Sensor Just Establishes “Normal”

Sensors do not always need to trigger immediate automations. Sometimes their first job is simply to collect history.

Let Home Assistant watch something long enough to learn what normal looks like. A freezer has a normal temperature pattern. A sump pump has a normal cycle frequency. A room has a normal humidity range. A server rack has a normal thermal pattern. A house has a normal overnight energy load.

Once you know the baseline, abnormal behavior becomes much easier to see. That may be one of the most underrated uses of Home Assistant. You do not always need to automate immediately. Sometimes you should measure first.

ESPHome Changes What You Can Measure

This is one reason ESPHome keeps coming up in my Home Assistant conversations. Commercial smart-home products are great when somebody already sells the device that solves your problem. ESPHome becomes much more interesting when they do not.

An ESP32 combined with the right sensor can potentially expose temperature, humidity, pressure, current, voltage, water presence, switch states, distance, light levels and other physical conditions directly to Home Assistant. I put together a separate guide on what ESPHome is and how it fits into Home Assistant if you want to go deeper on that side of the stack.

Phil Hawthorne and I also spent time talking about ESPHome Bluetooth proxies in HGG686. Instead of relying entirely on the Bluetooth radio near the Home Assistant server, strategically placed ESPHome devices can extend Bluetooth Low Energy reach into areas such as garages, basements, workshops and distant rooms.

That does not magically make unsupported Bluetooth products compatible with Home Assistant. It simply improves reach.

Again, the important question is not, “What ESPHome project should I build?” It is: What information do I wish Home Assistant could see?

Don’t Measure Everything

There is a trap here. Once Home Assistant makes data collection possible, collecting data can become the hobby. I enjoy that part too.

But more data does not automatically make a better smart home. Every sensor adds another battery, wireless device, ESPHome node, entity, dashboard item and potential maintenance point.

So I think every sensor should earn its place. The simplest test is this: What decision does this information improve? If you cannot answer that, maybe you do not need the sensor yet.

A Simple Way to Decide What to Measure

When I am thinking about adding another sensor, I start with the problem I am trying to understand and the state I wish Home Assistant knew: wet or dry, open or closed, occupied or empty, running or stopped, normal or abnormal.

Then I think about how accurate that information needs to be. A bad reading that turns on a decorative light is very different from a bad reading that controls water, HVAC or security.

From there, I would rather collect some history before getting too clever. Watch the signal, learn what normal looks like, and only then decide whether I trust it enough to automate from it.

That is the transition into the next layer.

Sense First. Automate Second.

We call Home Assistant a home-automation platform, but before it can automate a home well, it needs a useful model of the home. It needs to know whether someone is there, whether something is open or leaking, whether equipment is running, whether a system is behaving differently than usual, and how much energy is being used.

Those are sensing problems. Solve those first, and automation gets much easier.

That leads directly to the next question in this series: Once Home Assistant knows what is happening, how do you build automations that are useful, reliable and maintainable instead of creating an unmanageable pile of rules?

That is where we are going next.

Related Home Gadget Geeks Episodes

HGG691 — Home Assistant AI Automation, Tesla FSD and Energy Monitoring with Phil Hawthorne
Phil Hawthorne and I talk about Home Assistant, AI-assisted automation, energy monitoring, Docker updates and how to decide which jobs should remain deterministic.

HGG690 — Tesla, Home Assistant, ESPHome and Smarter Automation with Gavin Campbell
Gavin joins me for a practical discussion around ESPHome, custom sensors, leak detection, pump monitoring, HVAC filter monitoring, soil sensors and making a house more observable.

HGG686 — Phil Hawthorne on Smarter Home Assistant Automations, ESPHome, E-Ink and AI
Phil and I discuss ESPHome, Bluetooth proxies, automation architecture, E-Ink displays and ways of extending Home Assistant beyond off-the-shelf devices.

Additional Reading and Listening

Home Assistant Podcast — Roland “Accidentally” Installed Home Assistant, Then Automated Everything
A useful real-world example of measurement leading to a larger automation system, beginning with refrigerator energy monitoring and expanding into whole-home energy, ESPHome, batteries, solar forecasting and other data.

Home Assistant — Home Energy Management
Home Assistant’s documentation explains how consumption, production, storage and individual-device measurements can feed the Energy dashboard and help show where energy is actually going.