Watchful x Proper Case Study: Part 4 - Advice for New Users of Watchful

Key Takeaways

‍

• For Proper, the biggest advice they can give to a new user of Watchful is to treat your data as your most valuable asset and seek to understand it before you try to constantly throw more people and money at just getting reliable data.
  ‍
• It’s expensive to hire an increasing amount of people to be continually labeling and verifying in addition to the overhead of full-on machine learning. You have to approach data for machine learning in a different, programmatic, way.

‍

Shayan Mohanty (Watchful):  ‍

I got to say it's been such a pleasure working with you and your team like guys like became power users so fast. It was like, really humbling to see to that end. I'm curious like, we've got a whole bunch of folks who are kind of new to using Watchful. You guys have been using it for a while, you've sort of gone through several iterations. I'm curious what tips and tricks you might have like to share with some of our newer users. we've got a whole bunch of folks who are kind of new to using Watchful. You guys have been using it for a while, you've sort of gone through several iterations. I'm curious what tips and tricks you might have like to to share with some of our newer users. 

‍

Michael France (Proper):

‍Yeah. Well the thing that I think Watchful was built for this, right? But I've always kind of felt this way and particularly going through this with Watchful and with this business and the problems that we're trying to solve. I feel more strongly than ever that really understanding your data set is crucial. And it's ama ing. I think how that how little that's done.. Like why don't I just throw a whole bunch of stuff into a black box and try to get something acceptable. But understanding the data set I think is absolutely crucial. And that's the way that Watchful works to kind of meets you at that level, right? So Watchful was helpful, as I said, in showing us aspects or making it very easy for us to find places of like, hey, that association is interesting.  Let me actually dive in and actually look at that right now. Watchful makes that process first of all, I mean, it literally hints and helps you, right? It is you're both learning together as you're using it. That ability. And that mode, I think is really key though, you know. So I get to know your data set really well. Particularly the machine learning application or environment. Understanding the system I think is critical to being able to develop products and services that use it well, use the output as well. And there's always the struggle between a logical approach and then approach where maybe you can understand the concepts and you can think about them, but you don't really know how to compress that into an algorithm, right, which is exactly what we do with a lot of machine learning. Ideally, at least you kind of like structure it well. I think with Watchful, you structure a really good foundation for this and you learn about your data set as you're going. Um, I think that's a good level to engage. 

The other thing is that I think we very deliberately did this  And we looked at a competitor who had 20 data scientists in counting and not to mention whatever was outsourced down in South America or whatever else. Just to fundamentally solve labeling problems. I think when you look into how would I actually build a product or service on this, but how do I also build a business on this for us? It was critical that we took just one of the fundamental problems right on the nose and it was part of our solution. We cannot afford to hire a whole bunch of people in an increasing amount of people to be continually labeling and verifying and doing these things. And also accepting in any one given shot. The overhead of full-on machine learning. Everything at once was not something we could do. It was not really an option for us, and certainly not desirable. I think Watchful was really key for us in helping us jump-start this whole problem get us to actually. Like I said, like we expected to develop a model we didn't need to yet. We still don't need to based upon our recall performance and precision and run time. So the thought that someone who knows their business and knows their dataset can get to better know their data set and very quickly kind of translate that into actionable output that can either replace machine learning models or then feed them. And that is really, really efficient in terms of what ends up in most businesses and the real world being one of the biggest bottlenecks, just like labeling. That was crucial to us to make sure that we could actually hit gross margin targets with our solution while scaling. 

‍

Shayan Mohanty (Watchful):

‍Yeah, that resonates. So well that's the entire idea behind like, data centric AI. It's really just focusing on your data. And making sure your data is solid before you start throwing things at algorithms. I love the articulation here because that's exactly that was our intention with the product. It lets you focus on the data rather than having to think too much about all the machinations behind the algorithms and so on just focus on your data and everything else fall out of that. Yeah. And that is the most valuable component. I mean, there and you know, the data is the most valuable the thought that you wouldn't understand that or you need to constantly throw more people and money at just getting reliable data. That doesn't make sense to me, right? 

‍

Mike France (Proper):‍

Yeah. Absolutely. 

‍

Shayan Mohanty (Watchful):
Mike, listen, thank you again for spending all this time with us. Once again, it's been such a pleasure working with you and your team. I'm excited to continuing to hopefully impact how you guys are growing your business and we look forward to shipping features. So thanks. 

‍

Mike France (Proper):

Yeah, thank you. And really appreciate it. And, I obviously have to say how supportive you guys have been as well. Thank you in helping us just understand evaluate overall options. Not just how to use Watchful, of course. But you know, how do we just structure a system that actually does what we need it to and solve the problem well, so really, really appreciate it. Thank you. 

‍

John Singleton (Watchful):

Absolutely, man. It's been a blast yeah. And come on. We got to make sure that you're still around. So we can charge you next year, right? Come on?  I got to help you along in a cost-efficient manner. It's the front effort here. We're selfish. It's a healthy relationship. 

‍

Mike France (Proper):

Incentives. 

‍

John Singleton (Watchful):

Well, great Mike. Really enjoyed it. Thank you so much for taking the time. Maybe with one last thing I want to give a plug to proper. 

‍

Mike France (Proper):

Yeah, sure. Well, you can find us on the website of many leading brands, particularly in California. Probably the biggest brand in the world size. I we power their store finder and we power their product e-commerce. If you want to check that out. We also have our own marketplace at www.aproperhigh.com where we are starting to deploy more of what we've already done for brands here. But ultimately, like I said, to help consumers shop for branded cannabis. 

‍

Shayan Mohanty (Watchful):

So awesome, congrats again on all the success. And thank you again for joining us. It's been awesome. 

‍

John Singleton (Watchful): 

Cheers. 

‍

Mike France (Proper):

Have a great day guys.  Thank you. Bye. Talk soon.

‍