How to run a pilot involving smart sensors risk free

You, my dear reader, manage an elderly care facility or hospital. One pressing problem you and all your colleagues from other care homes have is staffing all shifts. Because of the aging population, the demand for care increases while the availability of caregivers declines, making staffing all shifts increasingly difficult.

You read the newspapers. You know about the enormous breakthroughs in artificial intelligence. You have heard from your peers that today, artificial intelligence helps to look after the well-being of your residents. In particular, smart sensors offering fall detection and restlessness recognition relieve caregivers at night. That way, artificial intelligence helps you provide better care. At the same time, the new technology makes your staff more productive. Smart sensors may be part of the puzzle to help you with your staffing problem.

But which smart sensor or which artificial intelligence application should you pick? An expert from Vilans, a renowned organization advising Dutch care homes, told me last week that they counted twelve fall detection vendors in the Netherlands alone. These vendors all claim their sensors are reliable and reduce false alarms significantly. But that is what they all say. How do you make up your mind? And can you make evaluations risk-free? Can you get your money back?

This LinkedIn post is about smart sensors, conducting successful pilots, and on how to get a refund if things don’t work out. I have been involved with smart sensors for years. Let me also be honest in that I am biased: I am the CEO of Kepler Vision Technologies. We build the artificial intelligence that makes smart sensors smart.

How to pick the best smart sensor?

If you are in the market for a smart sensor, I recommend trying to see this technology for yourself. Go to a tradeshow where they are on display. Walk around and talk to the smart sensor vendors. This should already be very revealing. Note which vendors do not showcase their products live on the tradeshow floor. Ask yourself, or better: Ask the vendor why they only show a video or leaflet. Does the technology really work?

At Kepler, we always showcase our product life—“Demo or die,” as we like to say. Our technology does not need to be fine-tuned. It works everywhere, even on a very crowded trade floor. If the vendor shows a live demo, then the fun starts. If you lay on the floor, does that get detected? Is the technology based on computer vision? If so, does it recognize each and every person visible? Because it should. By watching closely for a few seconds, you get a very clear indication of the solution’s reliability. If the smart sensor misses a person now and then, you can bet it will also miss a fallen client of yours now and then. In that case, find yourself a better smart sensor.

Another option to get a live demonstration is to request a live online demo. At Kepler, we give these almost every day through Teams meetings. We also have an Experience Center near Eindhoven where we welcome visitors. Lastly, all system integrators selling our technology have demonstration rooms and will be eager to show how our software works.

Having seen one or two demonstrations already gives you many insights into the possibilities smart sensors offer. But the question I posed earlier in the post, which one to pick, has not yet been answered.

Here, I recommend you be pragmatic and pick the smart sensor everyone will use in five years. You can already tell the winners, and here is how:

1. 1. Look at the number of reference cases they have. Is it easy to get to talk to a reference care home? In the Netherlands, we have plenty of care locations you can visit.
   2. Is the smart sensor supported by many (international) system integration partners? At Kepler, we have 23 international system integrators working with us.
   3. Can the solution connect to many different nurse call systems so your care staff can receive alarms on their phones? At Kepler, our website lists that we connect to 12 different nurse call systems. We connect to four more systems, but these are not GDPR compliant, so the total is 16 nurse call systems, depending on how you count.
   4. Will you get updates on the software, and how often? This indicates whether the application is still being improved or was a “build and forget” application. At Kepler, we release once every two weeks and have been doing so for the past seven years.
   5. Is the vendor focused on a single market vertical? Or is she a “Jack of all trades, master of none?” This is important because one of the promises of software is that you get updates and that, over the years, you get more and more functionality. That is not likely to happen if your vendor also develops software to count visitors at supermarkets or detect empty spots in parking lots. This is not the type of functionality that you need as a care home manager. For Kepler, our most recent extensions were “getting up in bed,” “getting up from a chair,” and “lying position” detection.
   6. Does the vendor take care of its certifications (ISO 27001, NEN 7510 are required) and offer patent infringement protection?
      Okay, let’s say you have seen demonstrations and spoken to a reference care home that recommends the new technology. You are ready to make a purchase. But someone in your team may still not be convinced. You are aware that you are not buying toothpaste. You are about to purchase a solution that will run at your care home for years. Picking such a solution is like choosing a new school for your children: you do so with care.

What is important to run a pilot properly?

In that case, you probably want to run a pilot. If you do so with the Kepler technology, you have two choices:

One is that you do not have any cameras yet. In that case, try the Mobotix C71 camera. Our software is pre-installed. Call your system integrator and ask him to switch our software on. On the C71, our software is called “Nurse Assist.” You can try it out for free for one month.

The other choice is for when you already have cameras installed and you have not deprecated them. In that case, try our on-premise server solution.

If you run a pilot, I see two things that are often not properly taken care of. These are:

1. 1. Not having defined – before the pilot starts – when the pilot is successful.
   2. Not having cleared what happens once the pilot is successful. And if it is not successful, do you get your money back?

Let me explain.

Before you start a pilot, you need to know how to decide whether it is successful. On a high level, a pilot with smart sensors is successful if the caregivers trust the system and, therefore, can change their way of working. With that, I mean that caregivers no longer need to make physical checks at night in every client’s room to see if the client is okay. This allows them to get the same amount of work done but with fewer people. This is part of the puzzle to solve your staff shortage problems

However, “trust” in a software application is challenging to quantify. Instead, I recommend three metrics:

1. 1. 1. Is it possible to send the alarms generated by the system to your existing nurse call system yes or no?
      2. Is the number of false alarms generated by the new smart sensor significantly less than before, say at least 50%? Yes or no?
      3. Are all the true false detected, yes or no?

A pilot is successful if all questions are answered by yes.

More tips:

• • I also like it if a care home adds an extra constraint such as “Is it true that the pilot did not in any way disturb our normal caregiving process, yes or no?”
  • Do not limit the pilot to one or two cameras. Statistically, one or two cameras give you little chance of a reliable pilot outcome. Right now, we have five pilots running. The smallest number of cameras involved is ten, and the largest is 24.
  • The duration of a pilot where we help is typically one to six months. By the way, I think six months is way too long. After a few weeks, you should be able to tell whether a system works.

The next thing that often is not taken care of is what to do when the pilot is unsuccessful or when it is successful.

When unsuccessful, you should get your money back. At Kepler, we never deal with care homes directly. Instead, our software is always sold through system integrators. Do you think our Chief Financial Officer will likely approve a credit invoice request if the system integrator claims a failed pilot? No, he will not. Therefore, we want to be involved if you run a pilot process. This means that you organize a monthly catch-up meeting to discuss the progress of the pilot. We want to call in to ensure that the metrics above are met.

If the pilot is successful, you should have prevented what happens if inflation suddenly soars, camera prices go sky high, and new and unexpected software price hikes occur. Negotiate your five-year license – before the pilot starts. Make sure that a failed pilot automatically terminates the five-year license agreement. With such a contract, you significantly increase your odds of a successful pilot.

In the end, when a pilot fails, this is because of events that are difficult to anticipate. I have seen a system integrator mess up when trying out a new, untested nurse call system during a pilot. Or care staff that threaten to go on strike if our new technology is introduced because they are afraid they are going to lose their job. But so far, we at Kepler have never had to refund a pilot. We always make absolutely sure our technology works as expected.

You can always send me a private message if you want to learn more about smart sensors and pilots and how to run them successfully. We will get you going.