Interview with Paul Mendoza
I bring 12 years of data science, machine learning and algorithm development to the task of helping sales teams eliminate manual data entry, enable higher rep productivity and deliver better client outcomes. I founded Dragnet Technologies to create tools for sales teams and managers like SigParser
. With our tools we eliminate manual data entry.
When did you first realize there was a need for products such as Salesmino and SigParser?
So I realized there was a need for Salesmino when I was working with sales teams and I realized that 80 per cent of contacts would never make it into a CRM system. So they were just throwing away all of the sales intelligence people they had interacted with these relationships they had developed but they were never in the system. So if the company ever wanted to utilize that or even if the sales rep wanted to find someone's phone number they would have to go digging through their e-mail. So I wanted to solve that problem for sales teams. So what I did was I built Salesmino which is this database of every contact a sales team is interacted with. And then out of that, we learned that there was this value in this email signature Parser we had developed to capture the phone number, the title, the address, LinkedIn URL, Twitter URL, of all of these contacts. We realized there was value in that algorithm. So we spun that off into SigParser and then that's turned into its own thing which has been incredibly successful for us.
From idea to prototype, how long did it take to develop/launch Salesmino?
So from idea to prototype, I would say that it probably took around three months to go from the initial idea and concept to having a semi-functional prototype. It was super rough in the beginning. I would say that almost no one could use the product in the beginning. It was just sort of a demo product that we could show to potential investors. We have our friends and family kind of sign up but even then we weren't really seeing retention. People didn't really understand what it was supposed to do. So then it took a long time of really iterating to really bring the concept down to a single sort of focus that we were trying to have now. I think one thing that we struggled with in the beginning was trying to make the product may be a little bit too diverse early on. We really should have focused on one particular vertical. So, for now, Salesmino you know right now is really focused on trying to build an automated contact database for a sales team to make sure they don't lose any contacts that they wouldn't have manually entered into a CRM system. So we want to grab everything possible for our sales team so they have every contact, every phone number, every title that's ever arrived in their e-mail system even if they didn't manually capture. So that's one of those things where we thought we wanted to be bigger sort of play that did a lot more than that but we realized that to really market we had to really focus it down to one sort of use case. So that's the use case we're focused on right now and then for SigParser, SigParser we took that from idea to prototype in about maybe a week and a half. That was just hey let's take some code we have already get it going, get it out there and see if anyone bites and turns out people who are loving SigParser, thanks.
What would your best advice be for anyone thinking of launching a startup of their own, especially in the data science or machine learning industry?
I would advise anyone starting their own data science or machine learning startup to not undervalue what a human can do running an algorithm vs. a machine an auto-generated machine learning output. A lot of times the machine learning outputs are kind of a black box which means that later on years down the road it's somewhat difficult to or very difficult to figure out “OK well how do I just tweak this a little bit”, it's especially hard for someone to come along and reproduce the data set that originally generated the algorithm. Unless you really document it really well. So sometimes it is better to try to hand-code what you can and then bring in machine learning to augment what you can't. For example, in SigParser we do use machine learning for parts of the process to identify certain spots that are really good for machine learning but then we have lots of the signature Parser code that is just handwritten because you just it's it was much easier to maintain that code. Long-term as something this is written by a human than it is to have it all as a machine learning algorithm.
What's next for Dragnet Tech? Any new service offerings or projects coming up in the last half of 2018?
So what's coming next for Dragnet will probably be a heavy focus on our new product which is SigParser, it parses the e-mail signatures as just an API and they're trying to make that easier to use for our current customers. And then also a lot of our new customers, people are just finding out we have this offer. So we're really trying to make it integrate with just everything and make it so that anyone can parse e-mails signatures whether they have Salesforce, GreenRope, any CRM system they have to make it really easy to capture all of those contacts into their system. Then we also have Salesmino which is our automated database that uses SigParser. So we're really trying to make that so that for those teams that just need a contact database that we're really improving that process of just making it super easy to sign up and capture all of your contacts without having a lot of integration set up. With SigParser you have to do a little bit more. With Salesmino we’re really are trying to work on our onboarding process to make it really really fast so that someone doesn't get confused. So it's totally obvious what they should do next. We are trying to work on that flow just for new users to make it really simple.