Marina Gorbis's Blog, page 787

Survey: Tell Us About Your Workplace Relationships

Richard Morrell/Getty Images

Research has found that employees are more likely to flourish at work when they have high quality workplace relationships with people from diverse backgrounds. These relationships can contribute to greater trust, facilitate information sharing, and engender positive emotional connections. Taken together, this can lead to increased productivity and reduced turnover.

There’s still a lot we don’t know about these connections, however. How often do they occur? Between whom? And what can better facilitate them? We’re trying to dig deeper, and we need your help. This survey, which takes approximately 10 minutes to complete, is the first step. We’d love to hear from you, and we’ll share what we learn both on this site and via an episode of the new season of HBR’s Women at Work podcast. You can also stay up to date by including your email address in the survey itself. All responses, of course, are anonymous.

We also wanted to let you know that this survey may be a bit different than ones you’re used to coming from HBR. It asks a series of questions about your personal relationships and activities both inside and outside of work, as well as some questions about you. We want to make sure we receive enough feedback from you in order to contribute to the literature on workplace relationships — and to help managers and employees better understand how to cultivate them. Thank you so much for being a part of what we hope is a project that will change work for the better.

Take the survey. 

How to Make Sure You’re Not Using Data Just to Justify Decisions You’ve Already Made

Alicia Llop/Getty Images

How can an organization tell whether it’s actually letting data inform its decision making — or if it’s merely using superficial analyses to retroactively justify decisions it has already made?

Traditionally, organizations have used data analytics as a tool of retrospection, as a means of answering questions like, “Did this marketing campaign reach our desired audience?” or “Who were our highest-value customers over the last year?” or “Did engagement peak at regular intervals throughout the day or week?” These answers are typically built around metrics — or key performance indicators (KPIs) — like click-through rates, cost per impression, and gross rating points, which companies all-too-often decide on too late in the process.

These descriptive analytics — that is, analytics that measure what has already happened — are undeniably important. But they’re just a bit player in the far more sprawling drama that is data-driven decision making. Within organizations that are truly data-driven, KPIs aren’t arbitrarily plucked out of thin air, but are generated at the start of a decision-making process. More precisely, it’s not an organization’s KPIs, but the key business questions (KBQs) — of which KPIs are an extension — that serve as the cornerstone of its success.

In their HBR article Big Data: The Management Revolution, Andrew McAfee and Erik Brynjolfsson arrived at a similar conclusion, writing, “Companies succeed in the big data era not simply because they have more or better data, but because they have leadership teams that set clear goals, define what success looks like, and ask the right questions.”

Insight Center



Scaling Your Team’s Data Skills
Sponsored by Splunk

Help your employees be more data-savvy.

However, arriving at “the right questions” is easier said than done, as any investigation must extend beyond, “What do the data say?” At my agency, our KBQs emerge from a rigorous four-step process that forces us to leverage data throughout the planning phases of our marketing campaigns. Though its specific applicability may vary slightly from industry to industry, our process provides a highly actionable model for deploying data analytics in a proactive, transformational manner; one that guides your decision making instead of justifying it.

Step One: Define your purpose. At the start of every planning cycle, an organization should make a concerted effort to engage stakeholders from every corner of its business in a wide-ranging discussion aimed at defining the campaign’s purpose. This begins with methodically zeroing in on the challenge(s) you’re trying to solve. Are you trying to improve a customer satisfaction rating? Cultivate long-term loyalty among a specific subset of customers? Increase the number of products that ship from a certain warehouse?

Don’t hesitate to interrogate the status quo — and, when appropriate, dismantle it. A history of maximizing pageviews is not itself a compelling reason to set a renewed goal of maximizing pageviews. Take a step back, survey the landscape (both internal and external), and carefully consider whether you’ve defined your purpose in accordance with anything other than the force of habit.

Step Two: Immerse yourself in the data. Once an organization has identified its purpose, it should conduct a comprehensive survey of what it already knows to be true. This is the stage where an organization should answer, “What do the data say?” That said, it should do so with a distinctly forward-looking mindset. At this stage of the process, an organization should take little interest in evaluating — and even less in justifying — past decisions. The totality of its interest should rest with how its data can inform its understanding of what is likely to happen in the future.

Like the previous stage, stage two is highly collaborative. In pursuit of broad-based collaboration, an organization should democratize its data to the greatest extent possible, funneling it into the hands of experts and non-experts alike. Not everyone at your organization is going to have a PhD in mathematics or a professional background in data science, but this doesn’t preclude anyone from getting their hands dirty in your data — after all, one doesn’t need to understand how a tool works to appreciate and take advantage of its utility. Ensuring that stakeholders across your organization come to a mutual understanding not only of the facts, but of their importance, is critical to the success of the rest of the process.

Step Three: Generate key business questions. While the previous stage pushes an organization to the edge of its organizational knowledge, this stage sends it tumbling into the unknown. With a goal and a set of agreed upon assumptions in hand, the organization has everything it needs to start posing KBQs, or lines of inquiry that propel it from “What do we want to achieve?” to “What do we need to know in order to achieve it?”

Using the precise purpose-defining language it established during the initial stage, an organization should now challenge stakeholders to ask as many questions as they can think of, first individually, then as teams. Good questions, bad questions, self-evident questions, unrealistic questions — it matters not. The objective is quantity, not quality.

While no topic or line of inquiry should be off-limits, an organization could start with these:

Can we predict which customers are at the highest risk of switching to a competitor, and design programs to decrease that risk?
Can we predict which customers have the highest probability of trying and subsequently adopting our brand, and design cross-channel promotional strategies to reach them most effectively?
Can we identify the optimal price point for our brand in order to maximize growth at a certain level of profitability?
Can we rethink the way we communicate with our target customers across our portfolio of products by understanding the combinations of products that are most often purchased by the same customers?

In many cases, such unfettered inquisitiveness requires feigning a degree of ignorance; that is, pretending that you don’t know what you know or pretending that your data doesn’t exist. This can be something of a high-wire act, especially for organizations new to data analytics, but it pays immense dividends if executed properly. Creativity and innovation are central to this phase of KBQ generation, and hewing too closely to your existing data is a recipe for the opposite.

To a similar end, it can be valuable to take the KBQs you generate and “invert” them. Just as sketching an object upside down can help an artist more accurately reproduce its likeness, rewriting your KBQs in the negative can produce more “Aha!” moments than would otherwise arise. Consider the following hypothetical progression that a pharmaceutical company might go through:

Purpose: Increase medication adherence among patients who have been prescribed Drug X.

KBQ: Which outreach methods do non-adherent patients respond to most reliably?

Inverted KBQ: Which outreach methods do non-adherent patients not respond to?

This slight shift in perspective can be a game-changer. Like any activity dealing with human behavior, marketing is an inexact science, and the value of strategically constraining your efforts cannot be overstated. Uncertainty is far more palatable — and far less problematic — when you know precisely where it exists than when it pervades your entire operation. In business, known unknowns are preferable to unknown unknowns.

Step Four: Prioritize your key business questions. Only after an organization has compiled an exhaustive list of KBQs should it begin evaluating, critiquing, and prioritizing them. In practice, some KBQs are highly actionable but lack the clear potential for making a business impact, while others have the potential to revolutionize your business but are highly inactionable. Pipe dreams, curiosities, and incremental improvements are all situationally valuable, but focusing on the pursuit of high-value KBQs will ultimately drive meaningful results.

Transforming a defense mechanism into a change agent. It’s tempting to place data analytics at a discrete juncture in your operational processes, but the reality is that data is not something to be used periodically, nor within strict project-based silos.

To drive real results, an organization must use data analytics throughout its business cycle. Today’s descriptive analytics are the foundation of tomorrow’s KBQ-oriented planning processes, which in turn are the foundation for a forward-looking analytics brief that details how an organization is going to answer its high-value KBQs. It’s this cyclical, mutually-informing decision-making architecture that both accelerates organizational transformation and disrupts your fixation on the rear-view mirror.

As Nobel Prize-winning physicist Niels Bohr once quipped, “An expert is a man who has made all the mistakes which can be made in a very narrow field.” Nowhere is this truer than in business. A well-conceived data analytics program empowers organizations to redirect their focus from justifying past decisions to learning from past mistakes. The sooner organizations make this pivot, the sooner they will enjoy the benefits of truly data-driven decision making.

The Motivating (and Demotivating) Effects of Learning Others’ Salaries

Dimitri Otis/Getty Images

Pay inequality is common in most workplaces. You get paid significantly more than your subordinates, your boss gets paid more than you, and your boss’s boss gets even more. In many large organizations, some employees can take home paychecks tens or hundreds of times more than others.

Whether you like it or not, your employees have wondered at some point about your salary — and their peers’. Should you be worried about that? Our recent research sheds light on this question, and our findings may surprise you.

We conducted an experiment with a sample of 2,060 employees from all rungs of a large commercial bank in Asia. The firm is quite representative of most companies around the world across some key dimensions, including its degree of pay inequality and non-disclosure policy around salary.

The first thing we looked at was manager salary. Through an online survey, employees had to guess the salaries of their managers. To make sure they had incentives to be truthful, we offered rewards for accurate guesses. The vast majority of respondents missed the mark by a significant margin (on average, employees tend to underestimate their manager’s salary by 14%). And this is where the action happens: by the flip of a virtual coin, we decided whether to “correct” a respondent’s estimate, by providing accurate information from the firm’s official salary records. So half of the respondents learned how much their boss truly earned — a salary higher than what they initially thought — while the other half did not.

Think about it this way: Let’s say there are two employees (similar in terms of level and experience) who think that their bosses get paid three times as much as them; but in reality, their boss gets paid five times as much. The flip of our coin randomizes which employee will learn that her boss actually gets paid five times more than she does, and which employee will not be corrected. Then we can compare the subsequent behavior of these two similar employees, to see how learning that your boss makes much more than you might affect your productivity.

To measure the behavior of these two groups of employees, we gathered daily timestamp, email, and sales data for the year following our survey. To our surprise, finding out that their managers got paid more seemed to make employees work harder than those who did not find out the true salary. Our estimates suggest that discovering that the boss’s salary is 10% higher than originally thought causes employees to spend 1.5% more hours in the office, send 1.3% more emails, and sell 1.1% more. (The higher the surprise, the larger the effect — finding out the boss earned 50% more led to effects five times larger.)

The evidence suggests that these effects were driven by aspirations. The effect of knowing manager salary was more substantial for employees who learned about the pay of managers who were only a few promotions away, whose shoes they could realistically aspire to fill. We find that, when the boss is fewer than five promotions away, for each 10% increase in the perceived salary of the boss, employees spend 4.3% more hours in the office, send 1.85% more emails, and sell 4.4% more. We also found that, after realizing that these managers get paid more, employees became more optimistic about the salaries they will earn themselves five years in the future. On the other hand, we found no effects on effort, output, or salary expectations when the employees learned about managers several promotions away (e.g., an analyst learning about C-suite salaries).

There is a caveat, though. While employees seemed perfectly capable of handling this vertical inequality, they did not handle horizontal inequality nearly as well.

In our experiment, we also asked employees to guess the average salary among their peers — that is, the other employees with the same position and title, from the same unit. Even though employees did better at guessing the salaries of their peers than that of their managers, most employees still guessed incorrectly. We flipped a second virtual coin to decide whether to “correct” their misperception about the peer salary.

We saw that finding out peers get paid more does have a negative effect on the employee’s effort and performance. Finding out that peers earn on average 10% more than initially thought caused employees to spend 9.4% fewer hours in the office, send 4.3% fewer emails, and sell 7.3% less.

This evidence suggests that it might not be wise to motivate individual employees through raises alone. If you increase the pay of one employee, that employee may work harder but the rest of the peer group could work less hard. You can avoid this by motivating employees through the prospect of a higher salary attached to a promotion. In other words, keep salaries compressed among employees in the same position, but offer them large raises when they get promoted to a higher position.

Our research raises the question: should you increase pay transparency at your company? Though surveys reveal most employees wish their employers were more transparent about salaries, the majority of firms maintain pay secrecy policies. But there is little evidence on how transparency affects the outcomes that managers care about. It is possible that managers choose pay secrecy because they think it is in their best interest when in fact it is not.

You may not need to worry too much if one of your employees catches wind of your salary. Employees in our study tended to underestimate the pay of their managers, and learning the actual amount led them to work harder. This degree of pay transparency seems to have given employees a sense of their earnings potential, driving up motivation. But we need further evidence to better understand how to best leverage transparency to promote productivity and employee satisfaction.

Of course, we must remember that salary information is sensitive, and thus there can be such a thing as too much transparency. For example, the majority of employees participating in our study were in favor of increasing transparency in an anonymous fashion, by reporting average salaries by position. However, when the same employees were asked about increasing transparency in a non-anonymous fashion, meaning their names and salaries would be shared, most of them opposed. And in a follow-up study, we found that most employees were willing to pay significant amounts in order to conceal their own salary from coworkers.

Many U.S. policies promoting pay transparency are mandating complete, non-anonymous salary transparency. For example, some states like California and New York publish online lists with the full names and salaries of every state employee. We think a wiser approach is what our study participants called for: transparency about average pay for a position, without disclosing individual salaries.

We encourage you to start experimenting with transparency at your company.  The first step is to figure out what your employees want. You can find out through anonymous surveys. Just mention some alternatives that you consider viable, and let them voice their preferences. For instance, do your employees feel informed about their salaries five years down the road? Would they want to find out the average pay two or three promotions ahead? Once you look at the survey results, you can decide what information to disclose and how. According to our findings, signals about the enticing paychecks waiting five years in the future is the push they need to be at their best.

Is Retail Dying? Plus, How Are Companies Spending their Tax Cuts?

Youngme Moon, Mihir Desai, and Felix Oberholzer-Gee discuss whether the “retailpocalypse” is real, try to figure out how companies are spending their Trump tax cuts, debate whether share buybacks are a good thing or a bad thing, and offer their picks for the week.

Download this podcast

HBR Presents is a network of podcasts curated by HBR editors, bringing you the best business ideas from the leading minds in management. The views and opinions expressed are solely those of the authors and do not necessarily reflect the official policy or position of Harvard Business Review or its affiliates.

How to Gracefully Exclude Coworkers from Meetings, Emails, and Projects

Here it is/Getty Images

You and about 20 of your coworkers are sitting around a crowded conference room table, discussing the details of some project. Some people are fighting for attention, trying to get a word in. Others won’t stop talking. Others have tuned the meeting out, retreating to their laptops or phones. At the end of the meeting, the only real outcome is the decision to schedule a follow-up meeting with a smaller group — a group that can actually make some decisions and execute on them.

Why does this happen? People hate to be excluded, so meeting organizers often invite anyone who might need to be involved to avoid hurt feelings. But the result is that most of the people in the meeting are just wasting time; some may literally not know why they’re there.

Whether it’s a meeting, an email thread, or a project team, people need to be excluded from time to time. Being selective frees people up to join more urgent engagements, get creative work done, and stay focused on their most important tasks. How, then, can leaders do this gracefully?

We recommend three steps.

Focus on key employees to protect them from overload. Most leaders try to pare down a meeting list or an email thread by looking for employees who clearly don’t need to be on it. But we suggest the opposite approach. Who is the valuable, collaborative employee you are most tempted to include? Now ask yourself: is she really necessary?

We pose this question because one of the foundational concepts to thoughtful exclusion is known as collaborative overload. The term was coined in a 2016 HBR cover story from leadership and psychology professors Rob Cross, Reb Rebele, and Adam Grant. Drawing on original research, they claimed that up to a third of collaborative efforts at work tend to come from just 3% to 5% of employees. These employees are often massively over-burdened and, in turn, at risk for burning out.

If the same small group of people get invited to every task force, every special project, every brainstorming meeting, there’s no way they can keep up with more valuable tasks. That’s why the first step to thoughtfully excluding people is to spot those employees at the greatest risk for collaborative overload, and then be incredibly selective about when to include them in meetings or other projects.

Address people’s natural social needs. The acts of excluding and being excluded are intensely emotional, even when people know they’re invited to too many meetings and resent getting too much email.

That’s because humans are social creatures; we naturally want to help those whom we consider close to us. The employees who suffer from collaborative overload take on such heavy burdens in part because they are compelled by these ancient impulses. It’s the same reason leaders over-include: They want others to feel like they belong.

The kind of exclusion that doesn’t trigger backlash or stymie productivity must address people’s varying social needs. If we look at who suffers from collaborative overload the most, we end up with two groups: employees who are too busy to be included in everything and employees who believe being over-included is a sign of prestige and status.

It’s up to leaders, therefore, to identify both groups and show them their time is better spent on projects with the highest return. Sample language might be variations on:

“I know you’ve got a lot of important work on your agenda, and I’d like to keep you off of this upcoming project so that you can focus on what you’ve already got. What do you think?”
“I’d like to take you off of this project, because someone else has a similar point of view. At the same time, you’d be able to add a ton of value to this other project because you bring a unique perspective. Would you be open to that?”
“I noticed that a couple of deadlines have slipped recently and that’s pretty unusual for you. Are there meetings, projects, or other things on your calendar that are consuming time or energy, that we might be able to reallocate? We all have times where we need some breathing room. How can I help?”

When leaders approach exclusion with employees’ social brains in mind, they can be more thoughtful in how they frame their directive.

Set clear expectations. Exclusion only hurts when people expect to be included.

The neuroscience of expectations shows there’s a great cost to mismatched expectations. When the anterior cingulate cortex, a brain region heavily involved in expectation matching and processing social exclusion, detects an error, it kickstarts a process that drains huge amounts of cognitive energy. This happens every time we encounter something unexpected, like seeing a favorite restaurant closed or getting disinvited to a meeting we’d normally join. That’s because the brain wants to make sense of the situation; it expected one thing and got another. Leaders eager to get the most out of their team members, by redirecting their efforts to more valuable activities, must understand and appreciate this aspect of the brain’s behavior.

If you only need a small subset of people attending a meeting, communicate with the rest of the group to ensure each person understands why they are not needed. Laying this groundwork also helps mitigate what psychologists call “social threat.” Just as loud noises and scary images can feel physically threatening, humans are wired to avoid threats in social situations, whether it’s anxiety, uncertainty, or isolation.

Managing people’s expectations ahead of time can act as a buffer against people feeling these kinds of social threats. For instance, the brain craves certainty, and being explicit about meeting participants’ roles offers it. Most of us also crave fairness, which you can provide by being transparent about the reasons for someone’s exclusion. That way, people can be excluded without the sting of feeling excluded.

Thoughtful exclusion in action

Leaders are responsible for appreciating these fundamental, albeit fragile, nuances of perception. When the time comes to launch a new project or host a big meeting, they should make it perfectly clear who needs to be involved, who doesn’t, and the reasons why. This way, employees will better understand how their role fits into the team’s larger mission, and with knowledge of other people’s roles, they’ll know who is working on what.

Think back to that chaotic meeting with 20 people. Thoughtful exclusion pares down that meeting to a core team of six or seven. Since the project manager now thinks hard about whose skills and time are most valuable — and whose would be better served elsewhere — she graciously decides you (and a dozen other people) have more important things to work on. As a result, the project reaches the finish line earlier and those employees who were excluded make greater progress on their own work.

Scale that behavior throughout an organization, and you have more people making better use of their time, tackling projects where their contributions are known, not assumed, to add value.

Exclusion may earn a bad rap in a climate where leaders are admirably sensitive about others’ sense of belonging. And it’s important to remember that thoughtful exclusion is only possible with an appreciation of the benefits of diverse perspectives and inclusive decision-making. But in order to avoid the dreaded logjam of over-inclusion, the brain science makes it clear that, with the right approach, thoughtfully leaving people out could become one of the greatest managerial moves a leader makes.

Auditing Algorithms for Bias

Laurence Dutton/Getty Images

In 1971, philosopher John Rawls proposed a thought experiment to understand the idea of fairness: the veil of ignorance.  What if, he asked, we could erase our brains so we had no memory of who we were — our race, our income level, our profession, anything that may influence our opinion?  Who would we protect, and who would we serve with our policies?

The veil of ignorance is a philosophical exercise for thinking about justice and society. But it can be applied to the burgeoning field of artificial intelligence (AI) as well. We laud AI outcomes as mathematical, programmatic, and perhaps, inherently better than emotion-laden human decisions. Can AI provide the veil of ignorance that would lead us to objective and ideal outcomes?

The answer so far has been disappointing. However objective we may intend our technology to be, it is ultimately influenced by the people who build it and the data that feeds it. Technologists do not define the objective functions behind AI independent of social context. Data is not objective, is it reflective of pre-existing social and cultural biases. In practice, AI can be a method of perpetuating bias, leading to unintended negative consequences and inequitable outcomes.

Today’s conversation about unintended consequences and fair outcomes is not new. Also in 1971, the U.S. Supreme Court established the notion of “disparate impact“ — the predominant legal theory used to review unintended discrimination. Specifically, the Griggs vs. Duke Power Company ruling stated that independent of intent, disparate and discriminatory outcomes for protected classes (in this case, with regard to hiring), were in violation of Title VII of the Civil Rights Act of 1964. Today, this ruling is widely used to evaluate hiring and housing decisions, and it is the legal basis for inquiry into the potential for AI discrimination. Specifically, it defines how to understand “unintended consequences“ and whether a decision process’s outcomes are fair. While regulation of AI is in early stages, fairness will be a key pillar of discerning adverse impact.

The field of AI ethics draws an interdisciplinary group of lawyers, philosophers, social scientists, programmers, and others. Influenced by this community, Accenture Applied Intelligence* has developed a fairness tool to understand and address bias in both the data and the algorithmic models that are at the core of AI systems.

How does the tool work?

Our tool measures disparate impact and corrects for predictive parity to achieve equal opportunity. The tool exposes potential disparate impact by investigating the data and model. The process integrates with the existing data science processes. Step 1 in the tool is used in the data investigation process. Step 2 and 3 occur after a model has been developed. In its current form, the fairness evaluation tool works for classification models, which are used, for example, to determine whether or not to grant a loan to an applicant. Classification models group people or items by similar characteristics. The tool helps a user determine whether this grouping occurs in an unfair manner, and provides methods of correction.

There are three steps to the tool:

The first part examines the data for the hidden influence of user-defined sensitive variables on other variables. The tool identifies and quantifies what impact each predictor variable has on the model’s output in order to identify which variables should be the focus of step 2 and 3. For example, a popular use of AI is in hiring and evaluating employees, but studies show that gender and race are related to salary and who is promoted. HR organizations could use the tool to ensure that variables like job roles and income are independent of peoples’ race and gender.
The second part of the tool investigates the distribution of model errors for the different classes of a sensitive variable. If there is a discernibly different pattern (visualized in the tool) of the error terms for men and women, this is an indication that the outcomes may be driven by gender. Our tool applies statistical distortion to fix the error term — that is, the error term becomes more homogeneous across the different groups. The degree of repair is determined by the user.
Finally, the tool examines the false positive rate across different groups and enforces a user-determined equal rate of false positives across all groups. False positives are one particular form of model error: instances where the model outcome said “yes” when the answer should have been “no.” For example, if a person was deemed a low credit risk, granted a loan, and then defaulted on that loan that would be a false positive. The model falsely predicted that the person had low credit risk.

In correcting for fairness, there may be a decline in the model’s accuracy, and the tool illustrates any change in accuracy that may result. Since the balance between accuracy and fairness is context-dependent, we rely on the user to determine the tradeoff. Depending on the context of the tool, it may be a higher priority to ensure equitable outcomes than to optimize accuracy.

One priority in developing this tool was to align with the agile innovation process competitive organizations use today. Therefore, our tool needed to be able to handle large amounts of data so it wouldn’t keep organizations from scaling proof-of-concept AI projects. It also needed to be easily understandable by the average user. And it needed to operate alongside existing data science workflows so the innovation process is not hindered.

Our tool does not simply dictate what is fair. Rather, it assesses and corrects bias within the parameters set by its users who ultimately need to define sensitive variables, error terms and false positive rates. Their decisions should be governed by an organization’s understanding of what we call Responsible AI — the basic principles that an organization will follow when implementing AI to build trust with its stakeholders, avert risks to their business, and contribute value to society.

The tool’s success depended not just on offering solutions to improve algorithms, but also on its ability to explain and understand the outcomes. It is meant to facilitate a larger conversation among data scientists and non-data scientists. By creating a tool that prioritizes human engagement over automation in human-machine collaboration, we aim to inspire the continuation of the fairness debate into actionable ethical practices in AI development. 

* An early prototype of the fairness tool was developed at a data study group at the Alan Turing Institute. Accenture thanks the institute and the participating academics for their role.

Managing a Data Science Team

Ragnar Schmuck/Getty Images

Many managers of data science teams become managers because they were great individual contributors and not necessarily because they have the skills or training to lead a team. (I include myself in that group.) But management is a skill unto itself, and relying on your experience as a successful individual contributor is not enough to ensure that you are able to retain and develop great talent while delivering valuable learnings, products, and outcomes back to the organization. Great data scientists have career options and won’t abide bad managers for very long. If you want to retain great data scientists you’d better commit to being a great manager.

What does it take to become a great manager? Volumes have been written on that subject, of course, including from HBR. But in my experience, a few areas are particularly important for those who lead data science teams. Great management means caring about your team members, connecting their work to the business, and designing diverse, resilient, high-performing teams.

Build trust and be candid

Trust, authenticity, and loyalty are essential to good management. That’s particularly true in data science where confusion around the discipline and its role in the organization means the team manager is responsible for insulating team members from unreasonable requests and for explaining the team’s role to the rest of the organization. Your team needs to trust that you will have their back.

Having your employees’ back doesn’t mean blindly defending them at all costs. It means making sure they know that you value their contributions. The best way to do that is to make sure your team members have interesting projects to work on and that they’re not overburdened by projects with vague requirements or unrealistic timelines (which is all too common given the high demand for data scientists.)

To build trust over time, you should invest in candor. Data scientists are smart people who are trained in how to interrogate and handle information. Therefore, my heuristic is to be about 20% more direct and candid than you think you should be. Be transparent with the good and the bad during the entire process, from recruiting, to onboarding, to the day-to-day, to performance reviews, and when discussing the team’s, department’s and organization’s strategy. It’s painful but critical for success. The moment you start “being nice” to avoid a tough conversation, you and your team have begun to lose.

Finally, feedback should be consistent and bi-directional, and great data scientists will smell bullshit a mile away. If you say you’re a believer in candor but become defensive or (worse!) don’t actually act on feedback, then your best reports will want to leave.

Connect the work to the business

To get the most from a data scientist’s time, they need to have a clear understanding of what the business goal behind the project is. Anchoring your team’s work in the context of the broader organizational strategy is among the most important jobs a manager of data science has. Unfortunately, it’s not always easy to do.

Data science projects often start with a question from someone outside the team. But often the question that the person asks isn’t exactly what they actually want to know. A lot of managing data science involves discussing and fine-tuning questions from stakeholders to better understand the information they actually want and how it will be used. Don’t let questions or requests become projects for your team until you know exactly what the stakeholder wants to understand and how they’ll use it. Having very clear objectives for the data-related questions that come your way is one of the most important things you can provide for your team.

Of course, stakeholders can’t always answer these questions on their own. They might not have a clear idea of what a finished data science product would look like (or how they would apply it). To fill this gap, make sure members of the data science team are regularly invited to product and strategy meetings. This way they can be inputs into the creative process rather than merely responding to requests.

Design great teams

There are many professionals trying to break into the “sexiest profession of the 21st century” and so, as a data science manager, you’ll get lots of applications and will have to be picky. Take advantage of that to be picky in the right ways. Care about your hiring process.

One of the biggest areas where people fail as managers is in the tradeoff between the short- and the long-term. For instance, it’s easy to start thinking that you don’t have time to recruit. This is a huge mistake. If you don’t have the time to find great team members and to scrutinize your interview and onboarding processes to ensure that you have good ones in place, then you don’t have time to manage a new direct report. Creating a great hiring process will pay off in the long term.

What does a great hiring process look like? For one thing, it doesn’t just focus on technical skills. Social skills like empathy and communication are undervalued in data science and the disciplines from which data scientists usually emerge, but they’re critical for a team. Make this a part of your hiring (but not in a way that amounts to hiring just for ‘culture fit’ and reinforces your affinity and confirmation biases). Instead of focusing on whether you can get along with a candidate, ask yourself if there is a lens though which this person sees the world that expands the boundaries of the team’s knowledge sphere—and value that dimension as highly as you value other attributes such as technical ability and domain expertise. This is why it is important to prioritize diversity. That includes diversity of academic discipline and professional experience but also of lived experience and perspective.

A few areas in particular stand out as important for data science. First, don’t just hire senior people. Not only are they in high demand and expensive, but less experienced employees have the “luxury of ignorance” and can ask “dumb” questions. These questions are not actually dumb, of course, but are unencumbered by the usual assumptions that more experienced professionals stop being aware they are making. It’s not hard to become infatuated with a particular way of doing things and to forget to question whether a favored approach is still the best solution to a new task.

Second, data scientists come from a variety of academic backgrounds: computer science, physics, statistics, and many others. What matters most is having a creative mind coupled with first rate critical thinking skills. I have a team member who studied marine biology and this diversity of expertise has proven extremely valuable. (The ability to translate domain knowledge about how pods of dolphin behave in the wild can be surprisingly useful when modeling a fleet of robots.)

Third, it’s important to hire individuals whose strengths complement one another, rather than building a team that all excels in the same area. A “big picture” person, someone who can articulate stories with data, and a visualization wizard working together can collaborate to produce things none could independently. To take the most advantage of these complementary skills, it’s important to make sure that the team actually works as a team and collaborates. You want your team working with each other and not just alongside. Regularly requiring members to read each other’s code and reports and fostering team activities centered around technical discussions ensure that you get the most out of this sort of diversity.

Finally, it’s also important to build a team that reflects the people whose data you’re analyzing. This is the only way to ensure that you have a resilient team that will ask better questions and a have wider aperture of perspectives from which to ask these questions. This way, each individual’s blind spots are covered by another’s past experiences and skill set.

When to specialize

One final piece of advice: When a data science team is just starting out, everyone on it will “wear many hats” and do lots of different kinds of data science. That’s ok—it’s like when someone joins a startup. But as your team matures and proves its value, recognize that roles will become more defined and some activity will move to other teams (infrastructure, ops, etc.).

Having said this, I would caution against specializing too soon. Specialization only works when well-defined and clear requirements are available to offset the coordination delays and costs associated with multiple teams working together. “Full stack” data scientists are very hard to find, but it is possible to find smart and driven “partial stack” data scientists who can learn, with a little dedicated coaching, how to appropriately frame a problem, manage a small project, develop and train a model, integrate with APIs, and push to production.

If you’ve done your job right as manager, this evolution will proceed relatively smoothly. You’ll have been picky in your hiring and created a great team with a balanced skillset. Your employees will trust you, and they’ll understand how changes support the organization and its goals.

End the Corporate Health Care Tax

Bjarte Rettedal/Getty Images

Imagine if a single piece of legislation could effectively eliminate all U.S. corporate taxes, subsidize hundreds of millions of dollars in new corporate investment, increase the take-home pay of most U.S. employees, ease state and local budgets, and reduce the U.S. trade deficit — all without increasing the federal budget.

It sounds completely impossible, but it is not: All we have to do is put aside the moral and political debates about Obamacare and recognize our health care system for what it is: a burdensome and unnecessary tax on corporate America.

U.S. companies pay $327 billion in income taxes, but they pay $1.1 trillion — more than three times as much — in health insurance costs. No other OECD country imposes anything close to such a heavy “health care tax” on its businesses. Eliminating this tax by shifting all responsibility to the federal government under a single-payer system would create a massive economic stimulus, providing Democrats with the universal coverage they seek while offering corporate America a far greater stimulus than any proposed Republican tax cut.

Insight Center



The Future of Health Care
Sponsored by Medtronic

Creating better outcomes at reduced cost.

After all, when the 2017 Tax Cuts and Jobs Act (TCJA) lowered corporate tax rates by 40%, saving corporations an estimated $950 billion over a decade, it created an immediate economic stimulus that bolstered corporate earnings and pushed the stock market to record heights. Eliminating the corporate health care tax would free up more than a trillion dollars of corporate earnings every single year, a stimulus 10 times more powerful than the TCJA.

Transferring all responsibility for health care to the federal government would not only offset 100% of what companies now pay in income taxes, it would provide an additional $773 billion a year in immediate bottom-line corporate profits that would be available for new investment. (Depending on how companies use these funds, they may end up paying tax on their increased profits, but even so, the net increase in after-tax income would substantially exceed their total taxes.)

The magnitude of this stimulus is hard to comprehend. It is comparable in scale to the $835 billion emergency Troubled Asset Relief Program (TARP) bill signed in 2008 that helped the United States recover from the worst economic crisis in our lifetime — but it would put that much money into the economy every single year. In fact, such a health care stimulus bill would dwarf any previous economic stimulus effort in modern times.

Wouldn’t shifting responsibility for health care to the government simply add a trillion dollars to the government budget? Not according to economic studies and the experience of other countries. Studies have shown that a single payer plan would save over $900 billion a year, eliminating more than 80% of the costs now borne by employers.

Much of this saving would come from reducing redundancy and inefficiency in seven areas: unnecessary services ($210 billion), inefficient delivery of care ($130 billion), excess administrative costs ($190 billion), inflated health care prices ($105 billion), costs from the failure to pay for preventive care ($55 billion), and fraud ($75 billion), and running Medicare and Medicaid as separate programs ($136 billion). Other studies have suggested that billing and insurance-related administrative costs would fall by 70% or more. After all, private insurance overhead averages 12.1%, compared to 2.1% for single-payer Medicare. Countries with single payer systems spend an average of 8.5% of GDP on health care vs. 18% by the United States. If administrative overhead were to drop to the level in Canada’s single-payer system, that alone would save $400 billion. Add to this the potential increase in tax revenue that would come from the growth in corporate earnings and wages, and the entire cost of eliminating the corporate health care tax could be fully offset.

Unlike the TCJA, which the Office of Management and Budget (OMB) estimates will add the full amount of corporate savings to the federal deficit, a health care stimulus bill would offer far greater corporate savings with no net impact on the federal budget. Such a stimulus plan would not only increase corporate profits but would also strengthen our economy in multiple ways.

Wage growth. The economic recovery has not translated into wage gains for the average American worker. Worker productivity grew 72% in the 1973-2014 period while median pay rose only 9%. In real dollars, the median income of middle-class households declined from 2000 to 2014 by 4%. Although there has been modest improvement in recent years, one of the largest drags on wage growth has been the increase in health care costs, which have taken an ever-larger bite out of workers’ take-home pay. Ninety percent of CFOs polled agreed that reducing health care costs would enable them to increase wages. Shifting the entire cost of health care to the federal government would eliminate the employee contribution to employer health care plans and would immediately raise the take-home pay of 156 million U.S. employees by an average of $1,443 per year.

Consumer demand. The stimulus effect of a broadly distributed increase in take-home pay would be far greater than the effect of the TCJA. It is estimated that 60% of the benefit of the TCJA went to stockholders rather than to employees or new capital investment. This increase in wealth went overwhelmingly to the top quintile of households that own 92% of all stocks. Studies show that that wealthier households tend to save or invest the extra money that comes their way, dampening its impact on the overall economy.

A broad-based increase in wages for the average worker, however, immediately translates into increased consumption that has a much greater stimulating effect on the economy. The Congressional Budget Office estimates that a onetime increase of a dollar in income would result in 84 cents of increased consumption by those in the bottom third of income distribution and 57 cents by the middle third compared to only 30 cents of increased consumption by the upper third.

Balance of trade. The United States has recently imposed massive tariffs on foreign goods in an attempt to reduce the trade deficit. The health care tax, however, puts U.S. companies at an even greater global competitive disadvantage. For example, U.S. automobile manufacturers General Motors, Chrysler, and Ford estimate that health care costs add between $1,100 and $1,500 to the sticker price of every car sold. By contrast, Toyota’s financial statements indicate that health care is not a material cost that is even worth reporting.

Corporate leaders know this well: 93% of CFOs, in a recent survey, agreed that the high cost of health care in America gives foreign companies a competitive advantage. Harold McGraw III, CEO of the McGraw-Hill Companies and chairman of Business Roundtable, declared that “health care costs are one of the top cost pressures… hurting America’s ability to compete in global markets.” Add to that the impact of poor health on productivity for employees that do not have coverage. And, unlike the tariffs that hurt U.S. farmers and many domestic industries and also lead to retaliatory tariffs from our trading partners, eliminating the health care tax would simply level the global playing field without any negative consequences.

Easing state and local government budgets. State and local governments have been increasingly squeezed by growing health care costs. A Pew study found that state and local governments were spending 31% of their revenues on health care by 2012. And a “baseline” projection by the Brookings Institution found that, by 2034, the increased health care burden on state and local governments “is more than the entire amount that states and localities spend on police and prisons annually. And it is almost as large as spending by states and localities on highways and the judicial system combined.”

Without the federal government’s luxury of deficit spending, state and local governments have had to compensate for increasing health care costs by cutting spending in other critical areas. In many states, teacher salaries, school budgets, hiring and wages for police and fire departments, and numerous other essential services have already suffered, and the ability to continue, let alone expand, these services is in jeopardy. Relieving state and local governments of their health care burden would immediately free up billions of dollars that could be used for better schools, safer streets, and emergency services.

Universal single-payer healthcare would also save lives and reduce suffering for millions of people, a massive benefit not to be overlooked. And the idea of federal government providing single-payer universal coverage is already gaining popular support with a majority of Americans. But leaving aside issues of humanitarian concern or political popularity, the economic case alone justifies Congressional action. Repealing the corporate health care tax would be a massive economic stimulus that singlehandedly addresses many of the nation’s toughest economic challenges. It might well be the only economic stimulus that could satisfy both parties, boosting the stock market and corporate earnings while providing meaningful economic and health benefits at every income level across America.

If Congress moves to act on this idea, we can expect health care insurers and providers to lobby hard to protect their profits, since much of that $900 billion in savings will come out of their revenues. But there is no reason the health care sector, representing 18% of our economy, should be entitled to impose a tax on the other 82%, especially when that tax undermines our global competitiveness, undercuts wages, inflates our deficit, and compromises essential public services.

As the midterm elections draw closer and the economy tries to sustain the longest bull market in our history, politicians know well that an economic decline is the surest omen of a change in political power. Republican leaders have proposed a second round of tax cuts in an effort to further stimulate the economy, while progressive Democratic candidates are promoting the radical idea of “Medicare for all.” Each side is deeply entrenched in its own political ideology and utterly rejects the views of the opposing party. But there is a single solution that fulfills both parties’ most deeply held goals: repeal the corporate health care tax.

The 6 Fundamental Skills Every Leader Should Practice

Thomas J Peterson/Getty Images

There’s an old story about a tourist who asks a New Yorker how to get to the storied concert venue Carnegie Hall and is told, “Practice, practice, practice.”  Obviously, this is good advice if you want to become a world-class performer — but it’s also good advice if you want to become a top-notch leader.

Over the past year we have been writing the HBR Leader’s Handbook — a primer for aspiring leaders who want to take their careers to the next level. As part of our research for the book, we interviewed over 40 successful leaders of large corporations, startups, and non-profits to get their views about what it takes to become a leader. We also explored several decades of research on that subject published in HBR; and we reflected on our own experience in the area of leadership development.

Our research and experience have shown us that the best way to develop proficiency in leadership is not just through reading books and going to training courses, but even more through real experience and continual practice.

Take the case of Dominic Barton, who served as the Global Managing Director of McKinsey & Company from 2009-2018. In an interview with us, reflecting back on his own development as a leader, he didn’t cite education programs or books he had read, but rather described several “learn-by-doing” experiences that would shape his successful career.

As the office leader of McKinsey Korea, for example, he realized he had “a small playground to… try new stuff” — and against all advice of local colleagues to be cautious and follow cultural norms, started writing a provocative newspaper column that challenged traditional ways of working among local businesses as their markets continued to globalize. “I took a risk, and it helped put us on the map, as never before.” His tenure in Korea also taught him that he was better at some things than others: “My performance evaluator used to beat me up regularly during those days, because I was better at opening up new initiatives than bringing them to completion. When I later became head of McKinsey Asia, he helped me see that I had to hire a solid COO to work with me—which substantially increased my leadership effectiveness in that bigger role.”

Our research also pointed to six leadership skills where practice was particularly important. These are not mysterious and certainly aren’t new. However, the leaders we talked with emphasized that these fundamental skills really matter. Aspiring leaders should focus on practicing these essential basics:

Shape a vision that is exciting and challenging for your team (or division/unit/organization).
Translate that vision into a clear strategy about what actions to take, and what not to do.
Recruit, develop, and reward a team of great people to carry out the strategy.
Focus on measurable results.
Foster innovation and learning to sustain your team (or organization) and grow new leaders.
Lead yourself — know yourself, improve yourself, and manage the appropriate balance in your own life.

No matter where you are in your career, you can find opportunities to practice these six skills. You’ll have varying degrees of success, which is normal. But by reflecting on your successes and failures at every step, and getting feedback from colleagues and mentors, you’ll keep making positive adjustments and find more opportunities to learn. Research by Francesca Gino and Bradley Staats published in HBR shows how important this reflection can be to your improvement: they found that workers were able to improve their own performance by 20% after spending 15 minutes at the end of each day writing reflections on what they did well, what they did wrong, and their lessons learned. Leaders often have a bias for action that keeps them from stepping back in this way — but it is the reflection on your practice that will help you improve.

Don’t wait for learning opportunities to be handed to you. Seek them out and volunteer to take them on.  And if you don’t see the opportunities in your own organization, find them outside your professional work in a community group, a non-profit, or a religious organization, which are often hungry for leaders to step in and step up. For example, Wharton’s Stew Friedman has described how one young manager who aspired to become a CEO joined a city-based community board, which allowed him to hone his leadership skills; three years later, he was on a formal succession track for CEO.

Eventually, as you progress, you’ll reach a level of capability in these areas such that you’ll start seeing results: you’ll successfully make things happen through the people who work for you on your team or in your division. As you succeed, these results will begin to build upon one another—you’ll oversee a new product that becomes a runaway hit or take charge of a transformational initiative that redefines a major market. More and more people will want to sign up and work with you. Clients or customers will ask for you by name. You’ll be invited to represent the company at major industry conferences. Whether you use this momentum to guide a new initiative or to start your own company, you’ll have begun to truly deliver major impact. You’ll have become a leader, capable of rallying an organization of people around a meaningful collective goal and delivering the results to reach it.

When Men Mentor Women

David Smith, associate professor of sociology at the U.S. Naval War College, and Brad Johnson, professor of psychology at the United States Naval Academy, argue that it is vital for more men to mentor women in the workplace. In the post-#MeToo world, some men have shied away from cross-gender relationships at work. But Smith and Johnson say these relationships offer big gains to mentees, mentors, and organizations. They offer their advice on how men can be thoughtful allies to the women they work with. They are the authors of Athena Rising: How and Why Men Should Mentor Women.

Download this podcast