Hi everyone.

This week, we’re looking at a structural part of our organisation: culture. 

When you think of hiring, do you think “culture”? Because, sure, there are parts of hiring that are pretty procedural. But ultimately, how we approach hiring, who we approach to hire, and how we then integrate these new hires is much more about the culture of our organisation than it is about the specific job and the specific skills we may be looking for in the news team member we’re trying to add. 

That said, if you’re the money person at your company reading this: We’d like more money for data please.

So this week, as we look at hiring and remote work, I hope you think about the greater picture that these questions touch on: how we organise work, how teams live, grow (and sometimes struggle) — rather than the mere practicalities of whether we do a daily Slack stand-up. 

And … this is our last newsletter for 2022: The Smart Data Initiative newsletter is taking the next installment off. I’ll leave you to fry up a latke, googling how much time what you’ve got in the oven really needs to roast, or to defend your holiday ornaments against the attacks of your cat. Whatever your particular brand of holidays, I hope they are filled with love and light. And perhaps a nerdy thought or two about data when the conversation around the holiday table takes a bad turn. Nobody needs to know where you escape in your mind.

All my best and see you in the new year,

Ariane

It’s tough to hire in data, but this forces us to question our hiring process to make it the best it can be

If you look around the career pages of tech or media companies, there are always open jobs in data. And this is despite the fact that data is not usually the large technical department (general engineering is usually several folds larger, unless you’re a pure data product company, and even then). 

But two different trends are fueling the faster growth of data teams: 

• The maturity of companies of all stripes in understanding the value they have in creating more data,and the value they would find in using it to further inform their business.

• The overall complexity and richness of the tools at our disposal also require larger teams. Whether you put data engineering in data, our data platforms are generally far more complex than what data engineering may have meant 10 years ago. Back in the day, data was really “just” business intelligence. And once you had ambitiously tagged your Web properties and apps and had signed a check to Omniture, you’d staff a few analyst jobs and be on your way to dashboard paradise. Such simple days.

So with demand being, to use a technical term, “hot-hot-hot,” well, money talks. And company culture, too: We talked about this in the previous newsletter, but if you’re an in-demand data professional, where do you want to work? At a company that actually does treat data like the great asset it can be — or at a company where the data presentation is just a conversation prop?

But there is another way to improve our prospects to detect and attract talent — and that’s improving the hiring game itself. 

Whose resume makes it past the filters of our recruiters? 

How do we approach understanding credentials and experience when so much of data is generalist enough that professionals from a wide array of disciplines possess a number of the fundamentals? Statistics majors feel like shoe-in, but the wider field of applied maths — of which you could consider statistics to be a part — really does prepare you for a lot of data careers, being focused on modelling). 

Meanwhile, a large part of data engineering — in particular data architecture — is really a flavour of general software engineering, for which training and experience can come from the most academic experience for folks who formally trained as engineers to folks who are self taught and learn the trade by doing. 

At The New York Times, widening the potential pool of candidates was done in a manner that tried to control for the thing that really mattered: the measurable fit-for-hire quality of the candidates. This was done in two ways: 

• Try and limit resume bias.

• Detect candidates with the right objective knowledge, even if their background is less conventional.

“We have SQL tests,”  Kendell Timmers, the SVP and head of data and insights at The New York Times, explained at the Monte Carlo Impact conference last month. “So one thing people often worry about, particularly coming from an unconventional data background, is that people will take a look at what you majored in or what school you went to and that will affect your evaluation.”

The tests, she explained, are a way to help both the interviewer and the candidate get an anchor into something more dispassionate than how well-versed a candidate is at the particular kind of interview they are taking. 

“This points towards having consistent evaluations because everybody takes the same test. The test is evaluated the same way for everybody.”

Now, the evaluation being the same for everybody does require some level-setting. How we may judge an answer could have natural variance. Complex SQL problems usually have several paths that could produce a correct result. Which one is the best answer does rest in the eye of the evaluator, to an extent. 

“We have this rigorous process where the people who are serving as a SQL evaluator sit down on each other’s evaluations occasionally to calibrate and make sure they’re reading the same way,” Kendell explained. 

On the other hand, job interviews will also consider the whole candidate and their cultural fit for the team they are joining. In data, where the diversity of profiles and experience can vary a lot, this would mean the hiring panel is going to be able to appreciate the diversity of these candidates in the first place — something The Times pays a lot of attention to, Kendell explained. 

A last component of a successful hiring process, Kendell noted, is how important hiring is made to be for the team. Is it part of the responsibility of the specific folks who work in recruiting or is it a part of many of your team members’ responsibility? 

“You really have to prioritise hiring to hire. We all have so much going on. That it’s very easy for this to kind of slip off to the side. And it’s not enough for you to say hiring isn’t your goal. Hiring has to be first of all in the goals of everybody on your team who’s doing the hiring. They have to have hiring as part of their goals or they’re not going to make time for it even though it will help them,” Kendell said. She noted that whichever part of hiring team members participated in — as test evaluators, as panel members — their participation needed to be part of their evaluation so their effort would be recognided properly.

This also means, in turn, that hiring is a stated goal: a measurable, stated goal that gets recorded as such — and where participation is a meaningful part of this person’s evolution in their own career ladder.

“All of this has to happen or hiring is always going to be something that you do on the side. I honestly think if you need to build a team, this has to take priority even over almost all the work you still have to keep the site running... . This really needs to be almost the first priority for a fair number of people that make it happen.”

Diversity, and a deeper talent pool, means supporting various working styles and lifestyles 

Not too long ago, I had an interesting chat with a data scientist at a publishing company in Asia. The data scientist has a few years experience, which means they are absolutely the hottest commodity: still very much “doing the thing,” using fresh-from-school training, and not yet incredibly costly relative to the work performed. 

This person’s boss and broader team were thrilled with them. They received accolades and bigger assignments.

This person works at a prestigious media company.  This person loved their job. Really cared about the mission. But also: This person is now leaving, they told me.

What happened? They were hired in pandemic times, when work-from-home was an option, and this meant a life where this person could work on a job they appreciated but also cultivate an athletic pursuit that was significant enough that they participated competitively at a very high level. 

Doing this meant they had to remain in a particular location — and couldn’t return to “the office” when the publisher decided that work-from-home was no longer a thing as the pandemic abated.

Regretfully, our young data scientist decided they cared about a work-life balance where they could continue to nurture their athletic pursuit. In short order, a job offer from a top tech company materialised — of course allowing them to work from anywhere.

Obviously, I’m keeping the details on the publisher and location very vague, because the goal isn’t to name and shame. But really, when you hear this story, it makes you a bit mad. “Boo hoo, ‘tis so hard to find great talent.”

How much are we making it harder on ourselves really?

This isn’t strictly because data is a field that has a supply-demand problem between staff and recruiters. This problem exists all over tech and, in fact, even in non-tech jobs. At this point, a work culture that feels arbitrary and process-based just doesn’t sit well with your teams.

In data, this is worse because much of the work is naturally goal-based and has measurable outcomes. For example: We create systems that didn’t exist yesterday but exist three months later. They connect with one, two, three different other tools, but they used to not. Such-and-such team used to not be able to extract this or that information, but now they can. 

What I described here is a road map, and road maps by their nature have measurable milestones. If your data team doesn’t have milestones to achieve, there are deep structural issues with your overall company — and not just your data team.

But, on the other hand, if your data team is measured through process-driven methods (“Did I see you at work today, in this building we call the office?”) rather than through outcomes, do you really think they will respect the organisation they work for? Do you really think they’ll choose you or that they will stay?

This isn’t to say that process isn’t useful and at times necessary. But many jobs are far more meaningfully measured on outcomes in the first place. And if your company culture only justifies process-based accountability on the basis of it having been the dominant culture for decades, expect that your hiring problems will not be improving in the future.

This also speaks to who calls the shot for creating company culture in the first place. 

The technical side of media companies has often been ahead of the newsroom and commercial organisations in terms of remote work. Offshoring part of the business has been common for a long time in tech. When my engineering team at Taboola was based in Los Angeles building analytics products, and my boss was in Israel (and I was based in New York), I had years of experience of working with quality assurance teams based in Bangalore, Florida and Russia — at a traditional media organisation. 

You could dismiss this as the anecdote of one person, except you’d find this is a very common scenario for Millennial tech workers, and it has been so for years. 

What our Asian media company essentially told our data scientist is that they didn’t believe he could do a good job working remotely. Except, they had. They had already made the demonstration that they did do great work remotely. 

So there’s that double cost right there: A smaller hiring pond to fish. And, culturally, one that states, upfront, that satisfying a process-based evaluation of your contribution is going to be the preferred form through which you’ll be measured. 

Hiring (and retaining) would be an uphill climb indeed.

Further afield on the wide, wide Web

For this last FAWWW, I’d be remiss if I didn’t bring you a fun reading selection about the newest AI on the block: GPT-3, the chat AI. This column from Kevin Roose in The New York Times (gift link for you!) has a great collection of good (funny) examples of how folks have tried GPT-3 and some (actually) good examples. It’s worth noting that GPT-3 seems less susceptible to becoming racist or otherwise inappropriate — certainly something that we know can happen to these large language model AIs.

We’ve come a long way. Indeed, compare this with what we worked with just a few years ago. Yes, I’m completely setting this up to bring up one my favoirite clips from the old Internet: the delightful experiment of a random Internet person who had two 2017 Google Homes chat with each other. An experience that turned into an absurdist performance art piece. This article from NY Magazine explains, but I particular recommend the YouTube clip where the two Google Homes are chatting. 

About this newsletter

Today’s newsletter is written by Ariane Bernard, a Paris- and New York-based consultant who focuses on publishing utilities and data products, and is the CEO of a young incubated company, Helio.cloud. 

This newsletter is part of the INMA Smart Data Initiative. You can e-mail me at Ariane.Bernard@inma.org with thoughts, suggestions, and questions. Also, sign up to our Slack channel.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .