Todd Carlisle is Vice President of People at Ipsy, a beauty company based in San Mateo, California.
A former HR director at Twitter and Google, Carlisle says “Ipsy” is a play on the Latin word “Ipse,” which means “self,” and the company is all about empowering people to express their unique beauty. Workforce Editorial Associate Bethany Tomasian recently spoke with Carlisle.
Workforce: How does HR within a startup such as Ipsy differ from more established companies you worked for?
Todd Carlisle: There are definitely a lot of differences. The first is that you are forced to become a generalist really fast. At more established companies, you have an immigrations team, benefits team, an analytics team or an inclusion and diversity team. However, at smaller company there might be five of you total and all five of you have to learn how to do all of those things. I quickly learned that some of the stuff that I’ve done well in the past, like recruitment and analytics, were fine but there are all these other parts of HR that needed attention. At a startup, there’s no one to turn around and pass the ball to. You have to do it.
We’d never had a formal diversity and inclusion plan and even though I had been involved that in the past, I was never the one to drive what that plan should be. That also brings a lot of excitement for people in all levels of their career in HR. They can take on a bunch of different things that they can learn about that they might not have had a chance to otherwise because you have your own lane at a big company.
For example, we had our recruiting coordinator plan our university programs approach. Usually that is its own team at a medium or large size company, but we didn’t have that. We couldn’t have designated a headcount for that, so we asked who was interested and passionate about it. Our recruiting coordinator raised her hand, and then we put together a plan. A couple months later: off we go to campus.
I also see that decision-making is faster at a startup, but the results are even more important because whatever you are deciding becomes a part of the company’s DNA. At a larger company there might be more checks and balances, You might have to run ideas through a lot of committees and get people’s views. While those checks and balances are useful, you don’t have time for that at a smaller startup.
One example of that situation was when we’d not had traditional peer-reviews in our performance appraisal process. I thought that was odd because everywhere else I’d gone to had them. I introduced the concept of having peer-reviews as a part of Ipsy. I think if I just said, “We’re having peer reviews,” that we would have done it. Instead, I opened it up for conversation. It made me pause and think that we should go slower, or else we could decide big elements of performance review in this single conversation.
Another difference is that whatever you’re offering to employees, even at a small start-up, you’re still going to be compared to the big companies. Since Ipsy is located in Silicon Valley, we’re compared to big tech companies. People don’t always understand the differences in funding and what you want to put that company’s emphasis on. You have to think about the trade-offs in what you want fund. Know that you can’t offer all everything all the time.
Workforce: What is the promise of AI for recruitment?
Carlisle: I think there is a lot of promise in the early stages that could (and should) be automated around finding people within LinkedIn or other publicly safe platforms for work-related information. We should automate scraping through those and compare the best ones to the job description requirements. After that process, it should be given to the human recruiters.
I remember talking to some executives on how many software engineers there were in our area that we could reach out to. That’s the kind of thing that should be automated over time where the humans can then apply filters to narrow the results to a select group. However, before that stage where humans take over, there is a lot of stuff that can be automated to make things easier.
For example, I am looking for an executive assistant to join Ipsy. I have my specific criteria for what they should have for the job regarding their experience and their employment history. For all that criteria, I should be able to talk into my Amazon Echo and have all that information be presented to me. Automation definitely helps recruiting in those initial stages.
I think automation is also good for is ensuring that there is no bias in job descriptions. As an example, there is an app called Textio, which makes sure that you have no gender-specific language in your job descriptions. I think that should be everywhere because it’s hard sometimes as a human to even know what that is.
This also applies to the initial technical screening engineering candidates. During these screenings you’re solving hard math problems that don’t require you to talk to another human, as long as it can be proven that you haven’t looked up the answers. I think that there is a ton of promise for AI during that initial screening process. After that though, I think the humans need to take over because there are skills that recruiters have that requires them to actually talk to a candidate. There are a lot of complicated interactions that make amazing recruiters worth their weight in gold.
Workforce: Is the data for AI recruitment made available through social media and other forms of public online profiles?
Carlisle: I’ve talked to various social scientists about it. You want to allow people to be one person on social media and another person at work. Who they are on social media is not necessarily who they are in their work-related life. I’m always cautious about trying to passively grab whatever you can find about a person and throw it into an algorithm because that social media identity is not necessarily who you need them to be at work.
My preference would be to use platforms where people tell you things about themselves in an appropriate way that is work-related. LinkedIn is a good example of that. You know whatever you put on LinkedIn is going to be scrutinized by recruiters and employers, so people are more careful with that information. HackerRank is another good example. If you’re an engineer and you’re taking a quiz on HackerRank, most of the time it is because you think it will help you to get a job. So, you’ll put you best skills out for that. I would prefer for people to train their tools on those kinds of platforms rather than what they can scrape off the internet.
Workforce: How might employee privacy be affected?
Carlisle: I hope that training people to use other tools would help protect employee privacy. You should tell your recruiters to look at these sources of truth like LinkedIn and HackerRank, that people share information on. Don’t use anything that is from a source that people don’t have a reasonable expectation that the information they put out might be used in a pre-employment decision.
Most places, when they offer pre-employment tests, have a disclaimer that the information provided might be used to make a pre-employment decision. There are explicit protections of the data that you share. I would much rather use that kind of information than tell recruiters to see what else they can find on a person. This also goes for background checks. There is a reason that companies outsource background checks to professionals in that area. These are people who can comply with the law and perform those checks in an appropriate way.
Workforce: Along with neutralizing gender-specific language, how can an automated HR help solve the diversity barriers in tech?
Carlisle: Aside from the recruitment stage, there is a lot that has to do with compensation. There is the still a gender pay gap. I’ve never talked to a single person who has been OK with the pay gap. However, the big problem is how to tackle it. There are ways to do it on the policy side. I think California has passed some great laws recently that have helped.
When it comes time to make those compensation decisions, I would love to have some help. For example, we just got done planning a round of salaries and raises for the whole year. I asked managers what they think certain employees should get and then I will give the managers market data. At the end of planning, I do the analysis to see make sure that there is no gender discrepancy for raises and promotion. We have total pay equity and total promotional equity, but I wish there was a tool that would suggest that I take a closer look at employees where the only difference in their pay is gender. Same thing for recruiting. I wish that there was a little bell that went off if the pipeline for certain roles had too many dudes in it. I eventually find out because I go around and ask but we can’t be everywhere to check everyone at once. I wish there was something monitoring the pipeline of candidates for roles to see a trend in gender that would then suggest waiting to decide until there was a more diverse pool. That should be built in to the recruiting process. On the compensation side, clearly, we should get to a point where when everyone comes in to work that they are all offered the same pay. However, if equal pay wasn’t the case when someone started working, then you only get a couple times a year to adjust that. I would love AI to monitor that and flag certain groups that look off so the humans can figure out what’s going on.
Workforce: How can an automated HR change the global landscape of human capital in tech companies?
Carlisle: There’s a huge element of HR and dealing with people that I can’t imagine being automated. I can’t imagine ever putting person in a room with AI to talk about workplace harassment. I want to keep the human in human resources.
I think 20 percent of the work we do in HR is rote and repetitive that could be solved by using a chat-bot or virtual human. For example, when there is a pregnant employee who is wondering about which forms they need, while the pregnancy is very human, the steps that the employee needs to take are always the same. I would like something that is in between an FAQ page and a chatbot that can take on some of those rote processes.
Workforce: How do recruiters and HR professionals respond effectively to the information provided by the data analytics without becoming daunted by it?
Carlisle: One thing we are stringent on is A/B testing. We never decide on something without the data. When we do make a decision with data, we test it and get more data. That’s important when we do these people-related things.
We had some reservations when we first started using HackerRank because our engineers thought the automated screening was something that they could do already. I had to assure them that the only thing adding HackRank could do to our recruiting process would be giving us a few false-negatives of people we shouldn’t have rejected, and we did. Then, we had our head of engineering review 500 resumes then sort them into rejections and acceptances. We ran all 500 through HackerRank and discovered that there was a significant amount of people that we might have rejected that in the end we ended up hiring. We aren’t removing the humans, but we are allowing HackerRank to do things that are saving engineers a lot of time.
Have both the humans and the technology there. This is a mistake I learned from early in my career. If you try to have the data be the only decision-maker, the humans will revolt. So have the data there and educate your teams on how to respond to it. Then over time you can see how accurate the humans are and how accurate the data is.
Workforce: Since launching People Analytics, what have been the greatest challenges of an automated recruitment process?
Carlisle: I think the hardest thing is that our thinking is still behind when it comes to how much people want to trust the data instead of their gut. When it comes to hiring someone, there are people who only need five minutes with a candidate to know if they’re a good fit. It’s hard to extract that gut feeling. It’s like telling someone who you’re going to marry because an algorithm picked them out. No one would sign up for that. It would be like a “Black Mirror” episode. When it comes to HR, it is a field that people see themselves as an expert on and they want to be involved.
When I started doing People Analytics 15 years ago, nothing like it had ever existed before. People are not used to looking at this kind of data when they make these decisions. I think we have to be patient. I love seeing all these programmers coming out of college where looking at that data is natural to them. When that group becomes executives, they would have already been swimming in that water. It’s just taking us a little while to get there.
Workforce: What have been the successes in the workplace thanks to data analytics?
Carlisle: Compensation is one. I think a problem HR always faces when doing compensation is worrying if someone is going to leave. HR might give someone more money if they are worried about that or perhaps HR believes an employee isn’t being compensated enough. A lot of times we have to battle with executives over those issues. Over time we have to carefully track why people leave a company. In exit interviews, I started asking very specific questions around compensation and if that played a role in leaving the company. Even asking them when they knew it was time for them to go, and what the factors were in their decision. I can quantify that information and keep that handy when I talk to the executives. Once a month, we have a meeting and I show executives who has left, why they’ve gone and how much compensation played a role in their exit.
Although compensation seems like an easy lever to pull, it’s a very expensive lever. Having that data handy always helps to drive that conversation and we can look at all the other factors that contribute to attrition risks. I think that’s been a change over time how HR professionals have come up with their plans.
Workforce: What were your motivations to move on from Google and Twitter?
Carlisle: When I started at Google it was pre-IPO and much, much smaller. I stayed with them for 11 years until they grew to 70,000. Now, they’re even bigger. Then I was at Twitter, and I thought to myself, “What’s really fun and what gets me excited?” I found that it was the building and scaling parts of a company, especially at a place where there’s still room to make changes and the DNA is not yet complete. I combine that with wanting to go to a place where there is really nice and smart people that are passionate about a product that brings nonstop joy to people’s lives. There weren’t that many companies that combined all those things for me and that’s when I decided to make a move.