Why Every Startup Needs People Operations.

Yes, Even You.

I am not ashamed to admit: I love numbers. Math was my favorite subject in grade school, I graduated university with a finance degree, and currently I work at a big data company. Numbers are logical, objective, and pragmatic — all qualities I embrace.

Knowing this, some people who know me well were surprised that I chose to focus my career in People Operations. It’s admittedly an area littered with stereotypes ranging from party planning to a glorified complaints department. And for some companies, that might be okay.

I disagree.

To me, People Operations is the marriage of the irrational, emotional complexities of people and the strategic, pragmatic structure of operations.

People aren’t a mathematical formula where you can input parameters and solve for x. If only it were that straightforward.

People are complex. One person may love to be publicly recognized while the same recognition could leave someone else mortified. The same email could be simultaneously perceived as either concise or rude. An interviewee could click perfectly with their hiring manager, but grate on the rest of the team.

Irrationality isn’t always a bad thing.

But that complexity is what makes my job so interesting. And as complicated and irrational as we can be, I like people. I genuinely care about the happiness of the people I work with, and I know I can make their work lives better through logic, structure, and yes, numbers.

This isn’t only true for me and my company though. People Operations is critical in all startups, whether or not there is a team or individual dedicated to it.

If you’re a founder, a leader, or just someone who wants to improve your team, you can leverage People Operations.


People focus is needed because it’s mandatory that employees care more. To work in almost any modern tech startup you’re expected to care deeply about the work you’re doing and the success of your company.

And this personal investment from employees means they expect more from you as well. A better work environment. Smarter coworkers. Cooler perks. Deeper learning. More accountability.

Operations is necessary because startups have limited resources so you need to prove your ROI. I’m not just talking about money (although that’s a real thing), but also people’s very limited time. You also need the buy-in of some analytical, naturally skeptical folks. Engineers and data scientists are good at their jobs because they question assumptions and test hypothesis. They need data before they change their ways.

But when so many of your key people metrics are qualitative, how do you prove your operational value?

If you’re Google, and have over 70,000 employees at your fingertips, you run the numbers. You run them on everything from the optimal number of interviews (4), to where to place the salad bar in the cafeteria (at the front), to an algorithm that predicts employee retention.

Prasad Setty, VP of People Analytics at Google told Slate Magazine:

“We make thousands of people decisions every day — who we should hire, how much we should pay them, who we should promote, who we should let go of.
What we try to do is bring the same level of rigor to people decisions that we do to engineering decisions. Our mission is to have all people decisions be informed by data.”

Which is fantastic, and one of the things that makes Google, well… Google.

But if you’re reading this, you’re probably not Google.

You (like me) probably don’t have thousands of employees and literally millions of resumes coming in to dissect, analyze, and experiment with.

So what’s a data-loving startup to do?

1. Learn from others

This is the fastest and easiest way to leverage data into your People processes — use someone else’s!

Whether it’s individual companies such as Google open-sourcing their internal analytics or rigorous peer-reviewed studies, there is research abound related to every facet of People Operations.

Looking to revamp your performance management system?

Harvard Business Review published an in-depth breakdown of how — based on internal and external research — Deloitte threw out their cascading objectives, once-a-year reviews, and 360-degree-feedback for a new system that provides more meaningful and accurate results while reducing the bureaucratic time drain.

Wondering which interview techniques actually predict future job performance?

Psychologists Frank L. Schmidt and John E. Hunter reviewed 85 years of research into what methods are actually predictively valid when it comes to hiring. The most important was general mental ability (smarts) combined with either integrity assessment (values), work samples, or a structured interview. And despite being common hiring practices, neither years of experience nor “someone you’d like to have a beer with” came anywhere close in validity.

Refreshing, but not a particularly valid interview technique.

Still in awe of Google and want to follow in their footsteps?

Luckily for you, they launched re:Work, as “a curated platform of tools and lessons from Google and our partners, designed to help you use data and science to make things better, no matter where you call work.” Those experiments I was talking about earlier? The lessons they’ve learned are now at your disposal.

I highly recommend checking out all that re:Work has to offer, but some of my favorites include:

2. Customize it to your reality

I’ve said it once already, but to reiterate: You are not Google.

Are there wonderful lessons to be learned? Is the scientific rigor of the above examples stronger than what you would be able to do yourself? Definitely.

But just like there is no “perfect employee” that will fit every job in every company, there’s no one-size-fits-all plan for employee engagement either.

Maybe your company is small enough that formal employee surveys would seem silly. Maybe you don’t have the budget to spend on professional general mental ability assessments.

That’s okay.

At Granify, our interview process is a constantly evolving mechanism. While built on the research I listed above (among others), it is still far from perfect.

For example, over the past year we’ve moving towards more structure in our interviews. Ideally we want candidates for the same role to have the same process and answer the same questions, but there is definitely variance at times. Sometimes we need to test new work sample tasks, ask new questions, or reorder steps to fit hectic schedules. Overall though, we’re making progress. We’ve even gone so far as to layout our entire hiring process on our careers page for everyone to see.

We’re also trying out better ways to test if someone is “smart”. Research has shown time and again that your general mental ability has a huge impact on future job success. But how do you test if someone is smart? Grades, gut feeling, brainteasers?

How many golf balls would fit inside a 747? A real, although now retired, Google Brainteaser.

We’ve used all three in the past, but now we’re rolling out a new online mental ability assessment to help us evaluate it. This third-party test has a few advantages over anything we’d be able to do on our own. Not only was it created by test-development experts, it has been tested hundreds of thousands of times — another custom way we’re able to leverage others’ research!

3) Understand statistical significance

Whether you’re using other people’s tools or running your own experiments (we like to do both), I have to emphasize the importance of everyone understanding basic statistics, especially statistical significance.

One of the first books I read when I started working at Granify was Statistics Done Wrong, lent to me by our Director of Data Science, Marcin. I’d taken some intro stats classes in university, so I knew the theoretical basis of p-values and confidence intervals. But understanding how easily stats can be misinterpreted — or worse, manipulated — helps put your numbers in perspective.

Marcin, who is actively involved in hiring our Data Scientists, recently sent me an article about how a company had used data to hire a data scientist. This company implemented a new structured, data-driven process and it was a huge success! They hired someone, and she’s great! There were hundreds of applications, so their process was tested hundreds of times… right?

According to Marcin, not really (and I trust him on this stuff — he does this for a living, not to mention his PhD).

Hiring one person doesn’t prove your process works. Only one candidate went through the whole process, so there is only one data point in their experiment. And at the time of publication, that employee had barely started, so it’s not even guaranteed that one was a success. N=1 is nothing more than a statistical anecdote.

Sample size matters.

Even when you have a hundred examples to pick from, like a hiring manager with decades of experience, there are still pitfalls to be careful of. One that has a concerning prevalence when it comes to hiring decisions is survivorship bias. As in, you only remember all of the amazing people you hired, while conveniently forgetting the false positives along the way.

Sometimes a decent sample size is hard to get, or takes longer than you have.

Then what?

4) Set future you up to succeed

Okay, so maybe you don’t have enough data now to come to any meaningful conclusions.

You’re a startup.

You’re growing.

You will want this data eventually, right?

The best time to plant a tree was 20 years ago. The second best time is now.

Figure out what stats are important to you, and start tracking them now. That way you’ll not only have more data, but you’ll be able to identify trends and see how things have changed after months or years.

Not sure what to track? Think about the challenges you’re facing now, and what information you wish you had to answer them.

  • What were the best sources for employee recruitment?
  • Why have people left your company?
  • How diverse is your organization?
  • Who are your best managers? What makes them the best?

You don’t need fancy tools to do this; a simple Google sheet can track most things when you’re starting out. Or if you are already using an HR software, such as BambooHR or Rise, that’s a great central place to keep all your data over time.

No matter how big you are, or whether or not your official title is People Operations, using data to make your team’s lives better is a real thing you can be doing right now.

And it’s simple. It doesn’t require an HR certification or PhD in Statistics. All you need to do to get started is:

  1. Learn from others
  2. Customize it to your reality
  3. Understand statistical significance
  4. Set future you up to succeed

Every great company is built by great people.

Why not do everything you can to support your great people?