Jerome Moeri On Navori Developing Its Own AI Analytics Add-On To CMS
February 24, 2021 by Dave Haynes
Switzerland’s Navori is among the most enduring and respected firms on the software side of the digital signage industry – widely used globally and known for being an early adopter of emerging technologies.
I did a podcast session with CEO Jerome Moeri about four years ago, and a new product release coming out of Navori presented a good reason to get back together recently.
The lab side of the business has been working, for several years now, on an AI-based computer vision platform designed to do audience measurement for retail and digital out of home. The product is called Aquaji, and it pairs with Navori’s well-established CMS software.
I asked Moeri about the thinking – given there are numerous commercial and open-source computer vision options already on the market. We get into why, what it does, and how it differs with what else is out there.
We also talk about the state of the business and industry on what we all hope is the tail end of COVID. We also hear his expectations that the coming year will see a lot of consolidation of the software ecosystem, through acquisition. Intriguingly, Moeri says Navori will be making a couple of acquisition announcements soon.
Welcome, Jerome. It’s been a while since we’ve talked, I looked up and saw that we first did a podcast almost four years ago now, which is amazing how time goes by. How has things been for Navori in the past year? I’ve spoken with many companies and generally speaking, they’ve done okay through all this mess.
Jerome Moeri: Yes, the pandemic was a moment of truth and the travel ban was very difficult for us because we are an international company and our business is based on traveling. So it’s been difficult. So we had to refocus on the research and development to end this pandemic with many innovations.
Yeah, I was curious about that. You’ve had to adapt to selling only online when so much of your work, with your with the guys I know over here in North America, is relationship-based and Jeffrey and Jordan are on planes a lot visiting clients, and now they’ve had to do everything online. Have you gotten good at that?
Jerome Moeri: It’s been difficult but surprisingly our revenue continued to grow last year in 2020 and North America was not affected at all by the pandemic. It’s quite surprising, but this is what happened and the Middle East and Asia also kept the same level of revenues.
In Europe, it’s a bit different. We had a slight drop because, in Europe, we were traditionally working on bigger projects, big deployments in retail, and most of the deployments were put on hold. It was a bit more difficult in Europe but North America and the Middle East and Asia are good. So we did not have any impact.
We’ve been able to do everything remotely using Teams and I guess it has not been too difficult for us because our company is 20 years old so we have a base of customers and all the recurring orders. But I had a thought of the young entrepreneurs, that puts a lot of effort into creating a company and for them, it’s been very difficult because they did not have a strong base of customers to face the pandemic.
Yeah. I would imagine a lot of your customers are kind of enterprise-level and as I’ve heard from some other companies, they just carried on knowing that this thing would end and they had the resources and they already had a plan in place.
Jerome Moeri: Yeah. We also had to open an online store and start selling online for the entry-level products and we have set up we had to set up full logistics, to take into account this pandemic
I’ve heard that from other companies where they’ve had to kind of branch into things they wouldn’t normally do or don’t really want to do, but you have to adjust.
Jerome Moeri: Yeah. We had to do it in such a way that we can still continue and not change our business models, and remain consistent working with partners. The development was a bit sophisticated, but we’ve been able to to to complete this development.
Has customer needs changed over the past year, are they asking for different kinds of things?
Jerome Moeri: Yeah when the pandemic last year came up, we were in the middle of research and development projects based on computer vision. So we had to stop everything and release a product that is a computer vision system that is integrated into our digital signage and it’s managing how many people can enter and how long they would wait if they have to wait to get in the store and we’ve been able to also to detect whether they are wearing a mask or not. And we did that to help our customers, especially retail in Europe because they needed a solution to open their stores while following the regulation from COVID and so we released an add-on called, “Access Control” which was dedicated to this type of use and it did help a lot our customers in Europe.
I’ve seen a lot of reports around access control systems and thermal readers and things that will meet the people coming in and out of a retail environment or another environment and I’ve been very curious about how much actual take-up there’s been of that. I think it’s quite interesting, but because I’m cocooned, so to speak where I live and I’m not traveling and seeing this stuff, I’ve not read a lot of indication that there’s been much take-up in retail, but are you seeing it happen?
Jerome Moeri: Yes. It’s very important in banks, in department stores where you have multiple entries. A human being cannot count and check how many people are in when you have multiple entries, for instance, and only the computer and the software can do that.
It has not been deployed so massively, to be frank, but for downtown department stores or banks, or flagships, it’s being used intensively and it was just a solution we tried to bring on the market and to help our customers.
You’ve just released a new product that you were referencing earlier with computer vision, it’s called Aquaji?
Jerome Moeri: Yeah, so access control was a digital signage product. So it was related to our digital signage product so a maximum of users may take profit of it and it’s because when the pandemic happened, I assigned 50% of our R&D team on computer vision starting in 2017 and we have made some prototypes and investigations and also market insights because we thought it was a market that was related to the digital signage or to the OOH and at the same time, it was different in the sense that it’s pure AI.
And we found this potential market interesting. This market would be worth, according to the insights we get, more than $1B within five years, just the software for artificial intelligence in retail.
Now the whole idea of audience measurement using computer vision and AI has been around for 15+ years, there’s a number of pretty well-established vendors out there doing it, and we’ve even seen some of the display manufacturers like NEC, in particular, coming up with their own version of it. And there are open source libraries that have computer vision, open-source code, all that sort of thing.
So I’m curious, why did you see the need to develop your own when there was a lot of it out there?
Jerome Moeri: So first because such companies do not have digital CMS software in digital signage and the connection between both systems is very interesting because the content is on the digital signage end, we thought we had to make these developments and to release a new range of products.
The second point is that this is true, that you have a lot of open source code, viable from the web, with some models enabling you to do some computer vision. When we did research and development, we found out that most of these companies have a level of accuracy at about 40% and this technology is consisting, mostly of counting bodies, not detecting people. So if you have someone passing by multiple times when you have employees, it’s just the body and the censors are doing a great job in counting bodies but the computer vision is not needed to count bodies or shapes. What we have developed is we created our own engine, just like we did in digital signage and what makes our system special is that we can combine and create multiple models. So we create models and we combine models to reach a degree of accuracy beyond 90%. This is the first differentiator.
The second differentiator is that because we can identify people when someone is passing by multiple times, we catch only one person, and because we identify people, we can say how long they waited in line and how long they stayed in the store.
And for the OOH industry, we have also developed a technology which is detecting the field of vision of the people passing by and we can determine, whether they had an object within their field of vision. It can be a product for the retail application, or it can be an advertising panel for OOH, for instance. And we can say if they had the object within the field of vision, and if they looked at the object, or if they interacted with the object and for how long. And so these are the main differentiators. And the reason why we’ve been able to achieve this is that the engine was created by us.
We used to collaborate with university researchers, and we also made our own models and we made an assembly of multiple models. So this is why we can reach a degree of accuracy of 90%.
Does the platform only work with Navori’s CMS?
Jerome Moeri: Yes, absolutely.
Okay, and how does it run? Is it running off of the same device that’s being used for the media playout or do you need a separate device?
Jerome Moeri: We need a separate device, like a PC for the moment, but in June we’ll be releasing a small device that would deliver digital signage, a media player plus computer vision, including the camera.
Okay, so an all-in-one thing.
Jerome Moeri: Yes and it will be far cheaper than the PC solution and it will be all in one.
The reason why digital signage and computer vision are interesting is that within the digital signage system, for each impression of an ad, we have the ideal audience demographics, how long they stay, what is the opportunity to see, conversions and stuff like this?
So it’s a plug-and-play solution that doesn’t need to play with API and to create complex and sophisticated systems. The second reason is that digital signage can play some content and choose content according to what the camera can see. So we can reverse the model and adjust the content according to the audience. And again, this is plug-and-play.
I have always been curious about the idea of audience measurement-triggered content, so a male 40 to 60 walks in front of the screen, serve content that’s contextual to that person.
It’s always been interesting, but I’ve wondered how often it’s used and how much of a demand is there from brands and from retailers to do that because it could get complicated in terms of the scheduling and planning for that, right?
Jerome Moeri: No, you just set conditions and within a few clicks, anyone can do it from the UI and it’s always good to adjust. With the content triggering, you have two ways. You may adjust the content on the fly, and you may trigger it. I agree that for the triggering, it’s a bit special or figuring is more for emergencies, but I just think the content on the fly is something fully automated and it’s very easy to do.
And do your customers have their heads around that? They understand the possibilities ‘cause I can see them going, “That’s interesting, but that sounds awfully complicated, maybe we’ll do that later.”
Jerome Moeri: I think it might take several years to make people use this type of solution, but the product is available now, so it’s still a product for pioneers. And you should also consider programmatic systems.
I’d like to connect Aquaji with a programmatic system so we can deliver some very detailed and accurate statistics on the audience so the cost per impression may rise because of the qualification of the audience. And at that level, we can also measure the level of interest of a given content, because we can compare one content to another, to find out which one is more efficient than the other.
Yeah, that to me is the kind of the secret sauce of these computer vision platforms that I don’t think gets enough attention is the idea that you can take a look at dwell times and attention levels, piece by piece, and adjust the content accordingly instead of just shoveling it out there and hoping people notice.
Jerome Moeri: Yes. Precisely.
Do you offer some sort of a dashboard that your customers can then use to see what’s going on and understand it? Because if it’s just log files and it’s just a bunch of numbers.
Jerome Moeri: Yes, we have beautiful dashboards within the Aquaji user interface and that’s not made much for scientists, but it’s more for marketing people and advertising people, so it’s for everyone.
We tried to simplify as much as possible. But at the same time, we also have an API for data scientists that may retrieve information of cross-analysis with other business intelligence systems.
Do you see this product working more in the digital out-of-home sector or do you see retail being the big take up?
Jerome Moeri: We’ve addressed both markets, but I think OOH might have maybe 30% this year and 70% for retail. This is originally a marketing product, enabling people to move better about their customers, the traffic, their activities, and the customer experience and most of the features are marketing oriented.
Have you found your company being drawn more and more into the digital OOH side of things, just because of contracts that have come up?
Jerome Moeri: Yeah, contracts/opportunities. I think digital out of home is a very interesting market, especially from the backend, because it’s quite complex, you have to create rules and you have a lot of algorithms.
From our standpoint, the requirements are quite busy because it’s full-screen content, you usually don’t have dynamic contextual content on the screen, no automation, and stuff like that. It’s a market that is very interesting from the backend. We are clearly a contender on OOH. There is an incumbent company, which we appreciate a lot by the way and we try to make a difference with this integrated computer vision solution and we think it would be successful.
We will start the test of Aquaji next week at Istanbul Airport. It’s a bigger deployment. They have, I think a thousand displays, it’s a combination of LED displays and system on chip displays and they will make some tests with the content automation. So according to the audience, we might adjust the content on the fly.
In the past, when companies have looked at using computer vision hardware and software, they have often tended to just do a sample of locations and extrapolate data based on that sample, just because the hardware and the software costs to do it across all of the display is just cost-prohibitive.
I’m assuming that’s changing and when it comes to things like Istanbul airport, maybe you’re not at every screen with a computer vision node, but you can deploy them more broadly.
Jerome Moeri: Yeah. We will release our own hardware and we have simplified the process in such a way that this technology becomes scalable and deployable. Because all the analysis is done on and the numbers are important, but they are less important than the comparison over time, especially in marketing, but also in OOH, because you have to find out the trend.
If you are a restaurant, you need to make sure that your customer has not waited more than 10 minutes in a waiting line, for instance, you need to limit how long people stay there in the store based on demographic, age, gender, and stuff like this because it reflects the attractiveness of your store, its assortment, layout and things like this.
You have to measure how many people are in store and it’s also very important for retail and we created a product that is doing these types of measurements and can adjust the signage at the same time and I think the cost of Aquaji won’t be so different from digital signage after two years. Today it’s 30% more expensive than digital signage, but within two years, I can tell you, it will be exactly the same price. So twice the price of digital signage to be clear.
So with scale, that’ll come down.
Jerome Moeri: Yeah. We’ll develop a small device, plug-and-play, and what is also interesting with Aquaji is that we can plug the system into an IP camera. So any camera pre-installed, we can use the video feed to make the analysis. So we don’t need a physical camera next to it or something, to make the analysis. We can plug our system into the security cameras because you already have security cameras to feed them data for inbound people, outbound people, queuing and so we can use these cameras, so it would be a facilitator, the deployments.
How much pushback do you get from venues when you start talking about using their security cameras?
The whole idea of computer vision, particularly in North America, gets people all excited about an invasion of privacy, which usually is completely wrongheaded, but nonetheless, they’re excited about it. So how do you work around the privacy issue?
Jerome Moeri: The degree of intrusion of Aquaji is far much lower than a traditional CCTV that retail companies have been using for the last 30 years because we don’t store biometrics. We don’t store data that are related to individuals. We aggregate on the fly information and so it’s very close to the sensor.
I remember you had a case in Canada, you had the case with Fairview, I think because this company was storing the biometrics on the backend, on the server for analysis. We don’t do that. We don’t store biometrics at all and we are compliant with GDPR. It depends on the regulation, whether you film inside or outside, but we are fully compliant with GDPR.
So privacy is really a concern for us. This is also why we don’t process the kids under 18 years old. We don’t track the races and we have a fully encrypted process and we don’t store anything that is personal, whether biometrics or images or stuff like this. So I don’t think this system is so intrusive. This is for Europe and Canada, with the GDRP. We developed the software with about 50 features. The user can adjust the features of the software to be compliant with local regulations because GDPR might change from one Euro country to another.
Then you have the United States, except one of two States, there is no regulation, so it means that the customer can store with Aquaji, the biometrics on the central database and share this information with business intelligence and other marketing material. So, It really depends on the country. We can do everything, but in some countries like Canada and Europe, the user has to restrict the software in such a way that it is compliant.
So you run a company, between yourself and your R& D people who are usually pretty early on emerging technology trends. You guys were early adopters of system-on-chip, you were early adopters of Android.
What are the trends you’re seeing out there that you think are going to get attention and traction within the digital signage ecosystem?
Jerome Moeri: I think the digital signage industry is pretty much stabilized now, the software, the display, and software targeting the low-hanging fruits so they deliver a commoditized software and they try to approach the market whether directly or indirectly. And then you have professional software like Navori and at Navori, I would say in the United States, for instance, about 40% of the top digital signage operators are using the Navori OEM and the scale is greater than it used to be a five-years back. And the way I see how the industry would evolve is that these digital signage operators would become stronger, they are doing a lot of acquisitions, including internationally, and these big operators would continue to grow and for the proficient digital signage network, they require sophistication, they require the support of multiple display brands and operating systems and they want to do everything. So, for the top part of the market, we would continue to get stronger.
We will also do some acquisitions, some acquisitions would be announced very soon and so there will be some kind of consolidation for the bigger digital signage operators and for the rest of the market it will be taken care of by display vendors and probably the software for all the basic use cases.
Yeah, you’ve had Samsung and LG out there for a while now with their own CMS software. and Samsung in particular has really started to aggressively market MagicInfo in a way that they didn’t do for a very long time. You see the big display guys doing that more and more?
Jerome Moeri: Oh yeah, you can tell how good the software is in its ability to incorporate artificial intelligence in its coming technologies.
It’s interesting when you’re talking about Navori doing acquisitions. I was curious about that because I get a lot of phone calls and emails from venture capital firms and independent investors who are saying, “Hey, we’re interested in acquiring companies, who are out there. Can you help us with that?”
And it seems like there’s a lot of activity around that right now. I don’t know whether they’re looking for distressed companies or they just see an opportunity to grow.
Jerome Moeri: Yeah, from the software standpoint, you have a lot of national companies, a company that is leading or a number two for a given territory like Germany or Italy, Spain, and these companies, they have a problem because their market is not large enough and they have some market share, but it’s not enough to finance the research and development. And these types of companies are typically the best company to acquire and these are our target companies.
So who are you buying?
Jerome Moeri: I can’t tell you today that if you are patient enough, I’ll tell you when it will happen.
I’ll find out when everybody else does, right?
Jerome Moeri: Nope. You would find out earlier, two days before.
Alright, Jerome. It was great to catch up with you.
Jerome Moeri: Thank you very much, Dave. I wish you a great day.
Thank you. Take care.