HRExaminer Radio Executive Conversations Badge Podcast Logo

HRx Radio – Executive Conversations: On Friday mornings, John Sumser interviews key executives from around the industry. The conversation covers what makes the executive tick and what makes their company great.

HRx Radio – Executive Conversations

Guest: Darrin Lipscomb, Founder and CEO, Ferretly
Episode: 336
Air Date: August 23, 2019


Join John Sumser at this year’s HRTech conference



Important: Our transcripts at HRExaminer are AI-powered (and fairly accurate) but there are still instances where the robots get confused (or extremely confused) and make errors. Please expect some inaccuracies as you read through the text of this conversation and let us know if you find something wrong and we’ll get it fixed right away. Thank you for your understanding.

Full Transcript with timecode
Today’s Show is brought to you by the Human Resource Executive Magazine’s HR Technology Conference and Exposition, held October 1st through 4th at the Venetian in Las Vegas.

Join me and thousands of your colleagues at the world’s largest exhibition of HR Technology. Act now using the code HREX and you can receive a $300 dollar discount on your ticket.

Thanks. We’ll see you there. And by the way, don’t miss the Women In Technology segment.

Good morning, and welcome to HR Examiner’s Executive Conversations. I’m your host John Sumser and today we’re going to be talking with Darrin Lipscomb, who is the CEO and founder of a company called Ferretly. Darrin, [00:00:59] How are you doing?

Doing great, John. Thanks for having me.

Yeah, take a moment and introduce yourself, would you?

Sure, well at the risk of giving away my age, I spent my entire career pretty much in software industry. It’s about 30 years now. I was 24 when I first created and sold my first software product to the federal government and then late nineties.

I co-founded a CRM software company I sold a few years later to Remedy in Mountain View and I spent some time in Silicon Valley running their engineering and product management team. Second Venture was a smart City Big Data play that I founded. We developed a web platform for modeling IOT and spatial temporal data for visualization.

[00:01:41] I know it’s a mouthful but basically we could consume just about any data set including social media and use that for predictive modeling such as crime prediction and then I sold that company to Hitachi in 2014. So Ferretly is my third startup software and probably the one I’m most excited about I think has the most potential to be [00:02:00] disruptive.

[00:02:00] So tell me a little bit about Ferretly.

Well we spent I think it’s going on now to two full years building a powerfully simple web platform that leverages AI to perform social media background screening. So distill it down to a simple value prop. I just say organizations use our application to mitigate their employment risk.

[00:02:19] And we achieve that by flagging public social media posts based on a dozen or so categories such as bullying, toxic, obscene language, threats of violence, political extremism, and, and some other ones. The best way to describe it is sort of to do a little bit of revisionist History you remember Reality Winter, that name ring a bell to you?

[00:02:40] Well you have to follow the news but in 2017 she was a contractor for NSA and she ended up she was an Air Force veteran 25 year spoke four Middle Eastern languages and Farsi Pashtu and others and she was hired by a contractor [00:03:00] called Pluribus International her actual name was Sarah Winters and she went by the nickname Reality Winter and they Pluribus hired her in 2017.

[00:03:08] And you know, they passed all the background checks. She had a security clearance and the necessary skill set obviously with their translation skills and they hired her to basically translate NSA intercepts or intelligence that they gathered around Iran’s Aerospace and nuclear program from Farsi into English.

[00:03:29] And so she was an ideal candidate on paper but prior to your employment and this is early 2017. She was very outspoken against. Which is an abnormal in and of itself, but you know, she she also expressed empathy towards Iran and and just the sheer volume of clothes, you know, put her in a in a situation there where someone someone should have raised a red flag, but that’s really where fairly comes into play had they had our.

[00:03:56] You know within about 30 minutes, they would have been able [00:04:00] to do a social media background check and that would have, you know, raise those flags around political extremism and seeing language that she had posted. It was only four months after she was hired that she leaked NSA secret documents on the 2016 Russian election meddling to a news Outlet called I think it was the intercept and she was found guilty two months later.

[00:04:21] I think it was and violating the Espionage Act sentenced to five years in federal prison. So, you know have had an employer taking that 30 minutes and run the check, you know, they could have avoided that that whole Fiasco certainly pluribus to hit to their to their brand obviously as well as revenues and I think just this past year they ended up selling to a competitor.

[00:04:42] So yeah, that’s a big deal so that again it sort of speaks to the risk factor and that what we’re trying to mitigate that. That’s a pretty amazing story. What you’re saying is that NSA which is kind of has the largest data set of social media data. Can’t run it on the unemployed. [00:05:00] I wouldn’t say can’t they don’t have the tools and that’s the saying if you look at you know, they just passed a law all the Visa applicants.

[00:05:07] I think there’s about four million on average every year coming into our country statement department has required by law to check social media that came went into effect. I think late last year earlier this year and they have no mechanism today in the state department and that’s you know something we’re trying to raise awareness that hey that you know, we exist and you know, there’s a tool we can use Rai to.

[00:05:29] We perform that check more the sure a lot of talk about in that tell me about the AI components of the product before we go deeper into the. Yeah, just real quick. There’s really two two aspects or subset of AI the AI umbrella once our natural language processing that we use for sentiment analysis.

[00:05:47] And the other is our machine learning for post an image classification. And we combine those two things into what we call a social media score and what that does is it basically just normalizes the data. So a [00:06:00] user can make a quick determination at the candidate, you know should come under further scrutiny.

[00:06:04] So, you know, we sort of take these two different substances. That’s and we combined them and deliver a result pretty quickly. So how is this validated? Right? I get that you scavenge the data, but father what’s expressed in social media and what’s actually the case there’s some variance between that and so do you do some sort of actual validation of your Discovery or are you just like, so how do you validate?

[00:06:32] So there’s there’s two aspects here to you where human intelligence is required to augment the aii, right? We wouldn’t we’re not going to rely upon the a I want but what it allows you to do is take thousands of post and distill it down to ones that you probably should look at right. That’s the first thing but really before that is the identity resolution.

[00:06:53] So you really need a human to there are tools out there on the market the health and we have some built-in mechanisms into our [00:07:00] application to allow you to sort of discover. Uncover the social media profiles given individual and their location Etc. But so that identity resolutions really important first step.
[00:07:10] And then once the AI gets done and you know, we make it really easy to review what they I produced and then you can adjudicate that information both on the sentiment side if you got the sentiment wrong or and if we got the flagging of the poster, Tell me how this gets used sort of in practice. Is this at the border checking that you’re talking about doing?

[00:07:33] So I wish we’re two new and we’re just sort of getting our name out there, but that’s certainly a goal of ours to to use it not but not just you know federal government. I think organizations of all sizes and all Industries can take advantage of this. We have customers as small as you know, folks that own a pizza chain or running it, you know and making sure they’re hiring the right folks, you know [00:08:00] 21, I mentioned before the call and for college athletic recruiting and it’s really just a really any industry can benefit from from this type of background screening.
[00:08:10] That’s an interesting proposition. So what are the big what would you describe as the big questions that you’re trying to answer? I think there was a Sherm study. I’m sure you’re aware that turnover costs can get into the six figures on average depending on the employee that you’re you’re trying to hire.

[00:08:28] And and then if you look at that combined with what’s the impact of a toxic higher to the company’s brand or even you know, possibly these days workplace safety. So I thought isn’t an original thought is why that we came to the conclusion that this the market needs something like this is that we feel like traditional screening such as criminal background checks and employment verification drug screening.

[00:08:51] They really don’t answer some of the most important and basic questions about a candidate, you know, so we think by leveraging social media and AI we can get a better handle [00:09:00] on these risks. So if let’s talk about is the the existing ways of doing background checking get you what percentage of the decisions right and then what’s the added value of apparently by the preview making is that you have additional risk prevention so quantify that little.

[00:09:22] Well, I think just the example I gave you on Sarah winners was Air Force veteran security clearance. No prior criminal record. No drug can no conventions nothing. But then when do you you know, what cat someone you know, and on their social media platform and what they’re saying on a public forum, you can glean a lot about an individual and that’s really what this is about when it gets down to it.

[00:09:46] So it’s not just about the negatives, right? Your traditional screenings are going to point out negative things that have happened to you and your life whereas this stop location and social media screening can can actually point out positive things about [00:10:00] the individual that you may want to bring to the Forefront.

[00:10:03] That’s interesting. So you said there are categories of alert that either tell me about the positive categories of work that you use to do that. Well, we can identify things like someone does a donation, you know, Facebook for instance volunteering those sorts of things, you know, spending time with family and co-workers.

[00:10:23] All those things can be, you know leaned as in positive sentiment as well as you know positive flag. So I’m not just negative things like bullying and violence and those sorts of thing. So you get bullying violence political extremism. And then what are the categories of positive thing to Kasich language, which is sort of super sad of obscene.

[00:10:45] We’ve toxics a little more thank passive-aggressive and then you have obscene you have threats of violence. Don’t harm hate speech political extremism and Center then, you know image image classification such as drugs and alcohol related [00:11:00] damages explicit racy and even violent images. Okay, I’ll leave that alone.

[00:11:04] I was looking for a reason for a list of the positive things that you discover. Why don’t you hurt you?

[00:11:15] That’s why your stuff to find out positive things about this place. But if you’ve got categories that you deliver that sort of balance out the political extremism and violence. So you get a more balanced. Look at somebody. I’m interested in understanding. Well in sentiment really covers that as well.

[00:11:34] So you’re having a person that’s very positive can sort of offset some of the negative flagging so and that’s the importance of combining those two factors and which makes us unique in the industry as well. So who makes the determinations about things like that balance between negatives and positives who decides that the ins and outs of that algorithm.

[00:11:56] And is that publicly available? Yeah, we [00:12:00] do. We leverage open source. In fact our sem analysis is one of the compound it’s an algorithm. It’s an open source, and it’s literally had millions of data points trained by a community and it was specifically built for analyzing so. Media post and post like on things boards like Reddit and such so it’s been invented and it’s used a lot of act.

[00:12:22] In fact, it’s used by some of these other application for sort of brand recognition. You know, what is the public think about this new release our new car new model car for instance and. So it’s been valid and been using industry for over a decade and it just keeps improving because it you know, it is a so, so hang on so that open source tool is designed to categorize and evaluate in the categories that you’ve described bullying and I’m referring to sentiment.

[00:12:55] So that’s the setup. That’s the natural language processing. It’s open source piece that we [00:13:00] use. Okay. So the question that I asked you the question that I asked was who makes the decision about what you said when you do an analysis the positive stuff can balance out the political extremism and to the question who was who makes the decision about how that balance works.
[00:13:21] Yeah, so there’s the fire you to tell me. Yeah, I’m going to here’s how we get at the political extremism thing here is that we get a positive thing and here’s how we make the tear and here’s how the decision is made to blend the two. Well, these things are dependent upon data sets, right so a training dataset and the case of Open Source and why that’s valuable is that you get not just community that doing providing the training sets.

[00:13:47] But also you have incentive by the car. May’s the products and our space not just back on screen but in the social media analysis space as a generality, you know, there are incented to get it right so that [00:14:00] feedback into that those training sets is really important and it’s been going on for over 10 years and it constantly is being improved upon so.

[00:14:09] We found is actually fairly accurate. But again, you know, we always said that human aspect on the end of this to make sure that what we’re you know, we’re producing is accurate. So we have that that ability to make Corrections where needed to make corrections as the community the users make the corrections as those and then we take those Corrections and we actually feed that back into our training set.

[00:14:34] So it only improves over time so it’s not a saint so. Don’t don’t perceive this as you know, this big bad guy sitting here and manipulating the training data sets. It’s actually a community involved thing that and that’s the way we’ve architected the solution to allow for feedback from our end users to make sure it is sort of a Federated approach not a not a centralized decision-making on [00:15:00] that.

[00:15:00] Well, I would love to meet the person who gives you feedback on the quality. We have all flavors, right?

[00:15:14] Well, depending on your bat, you know, if it’s political extremism, obviously, if you’re in the left or right of that equation, then you know, hopefully the idea is a balance itself out kind of like our politics is supposed to work, right? Yeah. Okay, so. Sure. So in eastern you you’re in Easton Maryland, which is which is not a giant Tech Hub.

[00:15:39] Although there’s all of that all of that Fort Meade related stuff in the Maryland Countryside. How do you compete for development? Well, I think the market and the industry has really changed a lot and I’ve see it just the last 5-10 years, you know, you can really start a company anywhere and and there’s we’re pulling from folks in DC [00:16:00] Philadelphia Baltimore.

[00:16:01] So we’re close to all the major hubs here, you know, and and Marilyn actually has a pretty healthy set of Education, you know institutions that really feed those startups Maryland’s pretty active in that folks like Ted. On such but you know, it’s really up to the organization. Do you have the right ideas that people want to be a part of that team building something that is disruptive.

[00:16:25] And so I you know, we use a combination house and Outsource talent and but it really comes down to the world will reward system obviously competitive pay an options, but really more importantly it’s a, you know offering environment that’s challenging to everyone. And there’s a certain group of people that seek out startups and they thrive on that energy and that environment and those are the ones that we try to identify and we like to bring them on full-time this creates and how big is the company we just got going this year.

[00:16:53] We’re less than 10 employees strong and growing and you know, we plan to grow [00:17:00] pretty quickly here throughout the rest of 2019, especially in the 2012. You got it. Okay. So the big question is in all of this is what are the what are the key ethical issues in your work? So there are several, you know, one of the things one of the things that actually makes us feel good about it is that if you look at the current state.

[00:17:20] How organizations are using social media screen candidates, it’s fraught with issues. There are over there was a study done. I think was Careerbuilder. I think in 2018 did a study and they showed 70% of hiring managers or scanning their candidates social media already and then just doing it manually me up on Facebook or your Twitter feed and it’s highly subjective and it can create EEOC and fcra violations doing it this way.

[00:17:50] But if you’re Outsourcing this function, and if you couple it with a more objective way, I to evaluate the post and and evaluate the post thoroughly, you know, we’re [00:18:00] really addressing a big shortcoming in the market we feel and you know, another thing that we’re proud of is the fact that unlike traditional screening Mac method, like I mentioned earlier, you know, we’re really going out not just the negative aspects but also positive and.

[00:18:16] Lastly from an ethical standpoint. I think we’re only it’s important to note that we only evaluate public posts for the candidates. We don’t look at private posts. In fact, I think it’s 12 to 26 States. Now that have social media privacy laws in place that prevent employers from looking at private posts.

[00:18:35] So that that employer cannot ask the candidate to say Hey, you know accept my friend request or. Even worse give me your password to your social media account so I can check it out before I hire you that’s been made illegal in about 26 states, which we totally agree with someone post publicly and you know, they’re making a statement and I think that’s anybody could see it and evaluate that.

[00:18:59] so [00:19:00] I get the what you’re saying is legal to do. This is the question is is it ethical to do this? What’s the that’s should have the real meat of that question is. What’s the trade-off between invasion of privacy and we can talk about whether or not you believe that that’s the case and the employers need to spot risk, right and then if there’s some balance there and so I’m interested in anything.

[00:19:28] Well, I think from an ethical standpoint, you know, we think it’s absolutely ethical and we focused only on public posts, right? So if someone doesn’t want to say something to the world, they only want to say it to their friends, that’s all for me. We don’t get that we can’t get that and so we’re only looking at public and public accessible information.

[00:19:47] Yeah, it’s the same information. It’s you know, criminal background check. This is court records and this is public information to write. So so we view it in that vein. You know long as you’re you’re just looking at public posts. I [00:20:00] think you’re perfectly fine. And in fact the FTC agrees with us on that because they roll in that that way so that’s that’s that’s our position.

[00:20:08] So you’re scoring people? Yeah, we do only two we score the data is really not the person we score what we see from a cinnamon and a flagging standpoint. And again, this is the normalize it so in other words if you had a you got two candidates for a job one post a lot, right? They have thousands of posts.

[00:20:31] And they may have fun. Let’s assume they had high flag post and let’s assume for this case that they have the exact same sentiment over time. Let’s say somebody else only posted a hundred post, but they had four five flag post.

How do you represent that data to the user so they can make an informed decision so that Goering reflects that proportion of flag post to the total domain of public posts that were analyzing and that just allows you to quickly clean the.

[00:20:59] You’re [00:21:00] clear. We don’t we don’t are making a statement of risk. The organization that is up to the user of the application to make that final determination. And we’re just giving them the tools to enable that or make that a little easier for. For so many places to go with this. I guess the question that I have for you is do the people whom you investigate this way get notified that that’s going on and do they get to see a copy of what you pass on about?

[00:21:28] So fcra guidelines is you have to have a permissible purpose. So employment is one. You have to let the person know they have to give you permission to run that particular type of background check. Okay. So first off you have to have a permissible purpose you have to get permission from the candidate and then you have to have a mechanism for that person to easily dispute that report.

[00:21:54] And so if you follow those guidelines and there’s a slew of other regulations around that which we [00:22:00] follow to the letter then you’re perfectly fine on a little long I’d say most background screening company certainly follow fcra guidelines, just like we do. Well what an interesting conversation. Is there anything you want a listener to take away?

[00:22:15] Again, I appreciate the time and there’s a little bit of misunderstanding about social media. I think the Cambridge analytic thing sort of got people a little upset on the amount of day of the sheer volume of data that they were analyzing but we’re simply providing a tool, you know, another arrow in the quiver for employees or employers to manage better manage risk.

[00:22:41] And you know what makes us unique is we allow organizations to you know, quickly sign up under a minute on our website that can run a report have a report in their inbox in about 30 minutes. So we think it’s a game changer. And we we certainly want to follow all the legal and ethical aspects of this [00:23:00] Market, but we think it’s important and thus are all the surveys are showing that most organizations do find this important and we just give them a legal way to do this in a more efficient way to do it and there than they are today interesting take a moment reintroduce yourself and tell people how to get in touch with you.

[00:23:17] Thanks, John. Yeah, I certainly appreciate opportunity again during the podcast and the share, you know my thoughts and on this up-and-coming space and my name is Darrin Lipscomb, and I’m the CEO of Ferretly, the best way to reach me is my email and it’s Darrin at D-A-R-R-I-N at f-e-r-r-e-t-l-y dot com.

[00:23:40] Thanks very much.

It’s been a treat having you on and thanks everybody for tuning in today. You’ve been listening to HR Examiners Executive Conversations. We’ve been talking with Darrin Lipscomb who is founder and CEO of Ferretly a [00:24:00] 21st century background checking tool.

Thanks for everything, bye-bye.

Join John Sumser at this year’s HRTech conference

Read previous post:
Building a Case for Evidence-Based Interviewing

The job interview is a poor predictor of success and while current interviewing practices are ineffective, they are also familiar...