The truth behind “99% accurate” occupancy sensors

When it comes to occupancy sensors, “accuracy” is a word you’ll see everywhere. Vendors tout it, prospects ask for it and procurement teams sometimes make it a requirement: “Must be 99% accurate.”

But here’s the truth: Accuracy isn’t a single number. And it’s never as simple as it sounds.

Why accuracy is more nuanced than “99%”

In practice, accuracy depends on context: The type of space, the way people move through it and the decisions you’re trying to make with the data. For example:

• A conference room might require near-perfect accuracy when it’s nearly empty (so you know if it’s truly available) but can tolerate small variances when it’s packed.
  
  
• An open desk neighborhood can handle a little more wiggle room about people count without impacting decisions about workplace policy or layout.

So in real-world environments, the concept of accuracy is layered. A space planner trying to decide if a room needs to be reconfigured might need “perfect” counts for certain situations. A workplace strategist looking at trends across thousands of square feet might be fine with small variances.

That’s why claims like “99% accuracy” should raise a flag. Does it apply when the space is nearly empty and when it’s at full capacity? Is it measured in a lab under ideal conditions? Or in a bustling office where people stand in doorways, cluster in groups and walk past one another? Without that context, the number alone doesn’t tell you if the data is decision-ready.

How Density thinks about accuracy

At Density, we’ve always approached accuracy from the perspective of the customer’s decision-making needs:

• Is the data accurate enough to make the right decision?
  
  
• Does the accuracy hold up across the conditions you actually operate in?
  
  

Take our newest device, Waffle. We introduced a “Capped Count” approach, reporting only 0, 1, 2, or 3+ occupants in a space. Why? Because the likelihood of undercounting rises in crowded rooms due to point cloud merging and occlusion. By capping at 3+, we can confidently report on what matters most—telling you if a space is empty, occupied by one person, or in use by a group—while keeping accuracy as high as possible. With Waffle, we can tell you with 99% accuracy whether a room is occupied or available. 

For our Open Area (OA) sensors used to measure larger spaces, we focus on Time Within Tolerance. That’s how often the system’s reported count is “close enough” to reality to make the right decision for the use case. For small groups (1–3 people), our tolerance is zero—meaning we aim for perfect accuracy. 

For medium groups (4–7 people), being off by one person is acceptable. For larger groups (8+ people), we allow a tolerance of plus or minus two. Our target to meet or exceed 90% Time Within Tolerance better reflects real-world usage than a flat “99%” claim. It’s also validated against thousands of hours of annotated ground-truth data.

Measuring entrances and exits has its challenges

Entry sensors introduce another layer of nuance: They measure occupancy by counting everyone who walks in and out over time. But if the sensor misses someone or accidentally counts an extra person, which is completely normal, that little mistake sticks around and adds up all day long. We call this drift. 

Imagine you’re keeping score in a game, but you forget to add one point early on—the score will be wrong for the rest of the game. In a busy place, that can turn into a big gap between the real number of people and what the sensor thinks. That’s why we work very hard to keep drift very small—our accuracy target for Entry is >90% of all spaces on all days having <5% drift rate.

We’ve built automated ways to fix it during the day, so the “score” always stays close to the truth.

Customers say we’re best in class

Many of our customers came to Density after trying other occupancy sensors that promised high accuracy but didn’t deliver in practice. From the largest nightclub operator in Denmark to a “Big Four” accounting firm, the feedback is consistent: Our accuracy claims hold up when tested in their actual spaces. 

We run continuous comparisons against ground-truth data collected in real environments—thousands of hours of annotated footage—and incorporate those results into every algorithm update. 

That’s why we’re comfortable saying we’re best in class. Not because we chase a near perfect-sounding percentage. But because we’ve built a system where the numbers mean something, the limitations are transparent and the data is trustworthy for decision-making. 

And don’t just take our word for it. We’re more than happy to introduce you to one of our customers for a reference call. 

The takeaway

If you’re evaluating occupancy sensors, look beyond the headline percentage. Ask how accuracy is defined, in what conditions it’s measured and whether the vendor can explain the trade-offs they’ve made.

Simplified marketing claims might sound appealing. But in practice, accuracy is contextual, dynamic and worth understanding deeply. 

The good news: When accuracy is done right, like at Density, it becomes a powerful foundation for every workplace decision you make.

Interested in learning more about how Density’s solutions can help your business?