How Onepanel CEO Donald Scott grew Onepanel to $118.8K revenue and 300 customers in 2019.

hello everybody my guest today is don scott he founded taizu robotics and dslr pros forging partnerships with dgi hitting uh 10 million in annual sales and his background in technology development and orbital orbital atk missile systems and u.s navy and cyber security at experian sony and warner bros now general aviation pilot having owned and operated several types of aircraft he's also a triathlon enthusiast training for an ironman but in the meantime focused on his company one panel workflow automation for computer vision don you ready to take us to the top absolutely all right man so what do you like people calling you by the way donner donald dawn's fine don very good all right so i'm a little i'm a little confused here what did you do with dji and how does that relate to one panel uh so we actually built a robotics platform we leveraged a lot of the dji drones this is much early on and one of the one of the biggest challenges that we were faced with was how do you efficiently find uh data scientists computer vision experts and how do you actually scale a compute environment to actually build an object tracking model or platform and this was back in 2013 and 14 it was very difficult to do and ever since then uh we one of my uh friends actually his name is russ tehrani he actually started to look into how to leverage kubernetes very early on and how to scale a compute infrastructure environment and uh we started the company and now we're able to scale almost nearly infinitely across um kubernete using kubernetes native for any type of uh computer vision ai workflow right now very cool now when was founding date what year uh we we founded one around november december 2017. okay 2017. and then um obviously scaling today what's the revenue model is it a sas company or how do you price uh so we actually have two different models we have a cloud based offering more like a platform as a service where you can uh you know we have a monthly subscription fee but also uh we charge for compute right so if you're using if you want to use a gpu 8 gpu cluster v100 you know we'll charge you per second for that our other pricing model is we offer a on-site on-premise deployment option so uh you could then you know we would you know you would we would charge you uh per seat per node perhaps uh depending on the organization depending on the size scale that you need so in other words one panel can not only provide you with a cloud-based platform as a service but you can deploy in a hybrid configuration on-premise or um you know in any you know private cloud if you wanted to as well if you're like a government for example um you're doing facial recognition uh you don't want to put all your data in anybody's cloud you can we can deploy our platform like an operating system do you sell to china uh we actually are working with some folks uh potentially in china mainly aws and senate cloud we see that there's a huge opportunity for uh providing a way to scale infrastructure for uh product-based companies um you know we're looking to help them uh provide uh you know access uh this type of scalable compute they don't have that right now so uh we do have some client customer chinese customers right now we're looking to expand them to china very very considering the current macroeconomic environment how on earth are you convincing to be okay with an american company potentially owning part of their data so the partnership for us wouldn't be um you know our our instances keep in mind that we are deployable on any cloud so imagine tencent or alibaba cloud uh you know using our software as an on-premise type of platform right so they would offer that separately um none of our data would be held you know our us or european data would be held in in china whatsoever it would be an offshoot if you will of uh of of one panel and mainly it's going to be around the marketing and sales engine for a one panel type of instance in china sitting on top of alibaba cloud or aws out there interesting okay i know you have a lot of different customer cohorts but give me an average you're on average what's the customer going to pay you per year for full deployment of the platform so it depending on the type of customer so right now we see users spending close to you know several thousand dollars per month on our platform that's just on the compute uh but also uh you know we're partnering up with by cloud factory uh some of these annotation companies and they can they can uh you know consume about maybe four or five thousand dollars worth of uh annotation related work and since our platform tightly integrates with that annotation workflow for computer vision um we can actually do pre-annotation using gpus so we're expecting that uh that consumption of uh gpu infrastructure for annotation scale up quite right what is an annotation so annotation annotation is a you know taking an image or maybe it's a video feed say you're dropping two gigabytes worth of video from a self-driving car um and you want to identify hey that's the person that's a bike um that's a pothole don't drive over that so you'll have somebody in india the philippines or even africa drawing a box around it and labeling that's a pothole that's a truck that's a car example oh my gosh yeah you do it automatically we can actually help pre-annotate that so we once the initial data sets are are annotated classified we can then build a model or help them build a model or use a third party uh to help build a model that uh basically populates that uh that that bounding box if you will ahead of time so now it becomes a qc function uh we literally popped we literally annotated 22 000 objects in less than 45 minutes imagine how long that would take you and and days and not thousands how much are you doing how many how many annotations are you processing monthly right now across your entire base so i'll give you an idea one of our customers processes close to 250 terabytes per month on our platform right now what does that mean in terms of annotation uh annotation is uh thousands of dollars per customer uh we have it's it's hard to quantify right now but i can give you on average what our customers are doing in terms of uh volume of data right that's how we measure it yeah what i'm the reason i'm asking these very specific questions because i'm trying to get i want to make sure that i give some um example that my audience will understand that relates back to how much data annotations and how you just described it i think everyone understands what those are so i was wondering if you had a number i assume it's in the millions across your entire base uh so in terms of so we look at in terms of gigabytes and terabytes right so let's say you take a gopro camera you drop in uh maybe a gigabyte worth of data so out of that gigabyte we can pre-annotate 22 000 frames or objects and the number of hours associated with that would take you know a couple thousand hours per month if you multiply that times uh the number of workers doing it you're looking at maybe four to five thousand dollars per month for that one individual so we have close to two almost close to two thousand uh users on our platform now um we have maybe five to ten percent are actually using annotation at that particular volume right now interesting um when you say two dozen users are they all paying customers or is there a free plan we do have uh some free access to the platform that's our total uh user base around 2000 uh we are starting to onboard more larger enterprises right now where um you know you'll we'll be looking at uh you know you know per seat per node type licensing around that yeah so you launch 2017 and then fast forward to today about how many customers have you actually have paying are we talking like 20 it's an enterprise player is it 200 or more well over 200 okay sure yeah well i mean by the way i'd prefer you just give me the accurate number so it's less than 2 000 but more than but more than 200 is it what's the actual number paying customers i i don't have the data in front of me right now but it's it's uh because some each customer will pay at different levels right so some might be teams um you have individuals that are paying uh maybe intermittent uh intermittently right so they'll come on for a couple a couple weeks or a couple months to actually run compute to build some models then you'll have annotation so it's actually sporadic that it depends over the course of the year okay but i guess maybe the right question is over the course of the year right how many customers will pay you something to do some you know use your technology some way uh well over you know 200 to maybe 500 uh okay okay fair enough so a good portion of your of your free users are paying call it 25 30 percent yeah roughly yeah interesting okay and and kind of how did you get your talking back to the the founding days 2017 what was the struggle like on your first kind of 10 customers how'd you find them yeah that's a that's a good question um i think mostly it was word of mouth through our accelerator we went through a accelerator out of minneapolis called generator and we you know uh they actually exposed us to a lot of interesting companies from like cargill to a lot of startups and how we got our first customers was really to say look um we part we started to partner up with uh different uh ai consulting firms um who are who specialize in different types of ai model building and uh what we did was we connected them together...