Data Analytics in the World of Bullshit Jobs

I don't think companies use data very well

What are you doing at work? Let’s start with the basics: your company sells something. Maybe they sell physical goods like food or tires. Maybe it’s intangible like software. Maybe they sell some complex financial or IT services that you don’t understand. Maybe you work in a nonprofit or government entity. In that case, you probably object to the word “sell,” but the term “offer” should apply. Keep that thing at the front of your mind.

Now let’s get back to the original question, what are you doing?  You probably have a few primary tasks plus some short-term requests that come up from time to time. You probably create some recurring reports or decks, or you might have already automated them. Keep that in your mind as well.

Cool, so now we have two things in our heads: the company output and your day-to-day workload. Theoretically, these two items should align. Companies hire people to make a profit, right? McDonalds doesn’t hire fry cooks to support the local economy. One can see the value of a role through a counterfactual: how would the company work without them? It’s hard to imagine a plumbing business without the plumbers or a software company without the developers. Beyond that, a company needs accounts, HR reps, and salespeople. Sure, the accountants don’t sell the computers at Apple, but the operation would fail without them. 

For data professionals, it’s harder to think of the counterfactual. A minority work on the product itself (think the recommendation algorithms at Netflix), but most analysts work for finance or operations. For you, what happens if you put down your current project and play World of Warcraft instead? Would anyone notice? Would anyone care?

I’ve asked this in euphemistic terms to various coworkers over the years and received responses like these:

• An important person requested this report

• We use these numbers for a presentation

• We need these numbers for a recurring meeting

• We always make this report

I find these responses uncompelling for one minor reason and one major reason. I’ll start with the minor one: companies can automate most recurring reporting. A good analyst should be able to automate reporting through some basic coding in Python, VBA, or another high-level language. If they can’t, a software developer could figure it out. If it involves a lot of manual work, a coordinator or executive assistant can put it together.

The major reason stems from the first couple of paragraphs: none of this impacts the thing your company sells. Does this report impact your customers? Will it cause them to buy more of the thing? Will it cause the thing to cost less money to produce? In my experience, the answer is usually no. White collar work tends to bloat and metastasize. Oftentimes, we find ourselves doing work that supports other work, and it’s hard to draw the line to any of it and the final product. 

Consider these two types of requests that one might receive as a data analyst:

1. We need this data for a recurring meeting.

2. At the recurring meeting, we use this data to decide which customer segments show the lowest forecasted demand in order to target our marketing activities.

The former treats a recurring meeting as an end to itself while the latter shows how the data helps sell the product. Data analysts provide value through the latter. However, maybe I’m violating the principle of charity towards the former request. Business leaders might say the former, but they always mean the latter. The manager at Starbucks tells the barista how to make the drink without informing them of the Pumpkin Spice Latte’s role in the broader business strategy. Can’t you just do your job? Do you really think leaders ask for useless crap?

Well, yeah, I think they do ask for useless crap. I know that some will refuse to believe this. Detractors will argue that firms maximize profit, so they wouldn’t pay for tasks that provide any value. I’ll address this argument in a different piece. For now, I’ll present an example from my own time in the workplace.

My team worked with a high-ranking executive that I’ll name Adrian. Adrian obsessed over certain parts of the business that didn’t seem to impact the customer. She also created numerous last-minute ad hoc requests that prevented us from focusing on longer term tasks. Adrian also required mountains of recurring reporting for her meetings with other executives, and we often made reports solely to justify her decisions. One day, Adrian was “no longer with the company.” After Adrian got no-longer-with-the-companyed, we stopped all these reports. The new leaders never replaced these reports. We never focused as much effort on her preferred tasks. We stopped updating the recurring reports, and our ad hoc volume plummeted. Yet, the customers continued to buy the product. In fact, this leaderless group performed better than they had under Adrian’s leadership. In short, Adrian wasted our time.

Yes, this is just one anecdote, but I have several more. All analysts I’ve spoken to could provide their own list of similar cases. One of my favorite Reddit posts discusses this. Data science is a new field, so some companies might struggle to figure out how to use it. However, I don’t think that’s the full story. I think our economy has a bullshit jobs problem.

In David Graeber’s book Bullshit Jobs, he finds that a lot of people work for Adrians. The book showcases people in a predicament: they know their jobs produce nothing of value but they’re afraid to tell anyone at the office (or Zoom call). If these pseudo-workers informed their bosses of the role’s uselessness, they’d get fired. While losing their jobs would provide an existential boon, people need money. As a result, people labor away at useless crap for years. In my favorite example from Graeber, one manager’s direct reports performed their roles without supervision, so the manager wrote and published novels during work hours. 

In the book, Graeber classifies five types of bullshit jobs:

1. Flunkies: they exist only to make their bosses look or feel important

2. Duct tapers: they fix problems that didn’t need to exist in the first place

3. Box tickers: they provide the illusion of the firm doing something

4. Task masters: they manage workers who don’t need to be managed (like our aforementioned novelist)

5. Goons: they exist to counteract the measure of other firms’ goons (e.g. telemarketers)

For my purposes, I think Goons differ from the other four types. A talented goon helps their company. Company’s need search engine optimizers to counteract their competitor’s search engine optimizers. At the macro level, they cancel each other out, so it might be socially beneficial to not have any of them in the first place. That’s a political question, however, and this post isn’t about politics

Removing goons from the equation, I’ll list ways that analysts can fit each type of bullshit job:

1. Flunkies: A non-technical manager might hire flunkies to increase his or her technical cred. A technical manager might do so to increase the size of the team and their own responsibility. In Bullshit Jobs, Graeber notes that white-collar workers often gain prestige and salary from obtaining more direct reports. Thus, analytics leaders might want to hire to increase their own brand. I’ve also heard some admit to hiring more entry-level positions so they can justify promotions for the more senior analysts. 

2. Duct tapers: I don’t think any analysts spend all forty hours duct taping, but 1) I wouldn’t rule it out and 2) I think almost every analyst spends too much time on it. Analysts may be asked to finish something on short notice when it’s not actually time sensitive. Pile these requests over time and the firm accumulates technical debt. This can take the form of data being poorly documented, pipelines breaking consistently, different teams defining key terms differently, or inaccurate reporting. Analysts can also serve non-analytical duct taping roles, like providing post-hoc justifications for errant budgeting decisions.

3. Box tickers: Every company wants to be a tech company right now. CEOs and investors see the success of Facebook, Google, etc and know that much of their success derives from data. Thus, they hire analysts to show that they’re a modern company.

4. Task maskers: Analytics managers who supervise those who don’t need supervision.

I think there’s also two additional types of bullshit jobs that don’t fit Graeber taxonomy

1. The right-hand man: This one might fit into the flunkie category, but I think the role differs a little bit. Flunkies don’t help their boss perform their jobs. Graeber’s examples include doormen in fancy buildings or receptionists who never recieve any calls. In both cases, the bosses hire these people for optics. Firms think they look fancier when they have doormen and receptionists. On the contrary, the right hand man actually helps their boss. In data, they can make their boss look data savvy while providing no value (or even negative value) to the organization. For instance, a right-hand man could make biased reporting that makes their boss’s crappy decisions appear profitable. The right-hand man could also manipulate the data to sandbag company goals in a way that increases their boss’ end-of-year bonus. In short, a right-hand man helps an individual manager at the expense of an organization.

2. The tradition keeper: These people produce reports because the company has always produced these reports. They maintain a dashboard that’s always been there. They analyze a certain metric because a company has always chosen that metric. Usually, people retroactively figure out they served this role. A manager suddenly leaves, a pipeline breaks, or someone forgets to update a report. Then, nothing happens, and the analyst realizes all the time spent on it was pointless. 

In the book, Greaber also discusses the “bullshitization” of jobs. Doctors and university professors spend a larger amount of time on administrative tasks than they used to. These jobs aren’t wholly bullshit, but they contain a lot more bullshit than they used to. I think most analysts fit into this bucket. In my experience, no analyst spends all forty hours on nonsense.

Given this, what should an analyst do? First, I think it’s important to separate individual actions from political tasks. Some readers will see this problem and argue that we need socialism, deregulation, or some other economic solution. In a sense, I agree. I don’t think individual actions solve social problems.  In the meantime, however, you have to work within the system that exists. From here, there’s two options: 1) apathy and 2) work towards useful tasks. I don’t mention the first option sarcastically. Many view work as the way to earn their material living and nothing more. If that work involves digging up holes and burying them again, so be it. This seems reasonable, and I won’t argue against it. Instead, I’ll delve into the options for the second path:

1. Suggest an alternative. Many business leaders lack ideas (see the Reddit post). If you ask them what could change, they might offer little more than aesthetic revisions to current reporting. Did you receive a request for a dashboard that you know will result in no business decisions? Counteroffer with something you find more useful. Even if they reject large portions of your suggestion, some piece of it may make it into the current workflow.

2. Half-ass the useless stuff. If you know the business will not make decisions based on reporting, don’t spend too much time on it. If the numbers look close enough, don’t double check for accuracy. Don’t make it aesthetically appealing. Get it off your plate to focus on the important projects.

3. Baffle them with bureaucracy. A good analytics team will handle incoming requests through some project management process. For the stakeholders that provide meaningful requests, feel free to bypass this a bit. Don’t document it, or make the ticket for them. For the annoying ones, make sure they fill out every box in the form. Demand a business case. Suggest an alternative location of the data, even if it’s some crappy CSV download from the CRM. Your power will be limited there; you probably can’t bureactize the CEO to death. Still, find the stakeholders you can push back on, and push them back. If they complain about the request process, offer them time to discuss potential changes to it with other business leaders. They probably won’t attend these meetings, thereby maintaining the process.

4. Find the right people to work with. Managers differ in their philosophies. Some do not care what people below them think. Some want to execute on the ideas they saw in a conference or their MBA classes. Others admit to a lack of certainty and analysts to help them. Find people in the third group and try to work with them. Your organization might also have leaders that came from an analytics background, so they’ll be more receptive to your ideas. 

5. Accept that some work sucks. As mentioned above, even doctors and professors spend time on bullshit. Even in Star Trek, our heroes waste time on bureaucratic bullshit. Focus on the things you can control.

6. Pass bullshit onto others. Some people will not care that work is bullshit, while others might not consider it bullshit. Finally, some might enjoy the exposure to data analytics projects, even if they don’t understand the business context. If these people like it, hand it to them.

7. Consider that you might be wrong. Maybe some of those tasks aren’t bullshit.

8. Find a different job. 

Do not assume that leaders know how to use data. Google released BigQuery in 2010 and Snowflake opened shop in 2012. Data scientists introduced the XGBoost algorithm in 2014. In other words, modern analytics is young. No one has mastered it yet, and some leaders don’t know what data can do for their team. In general, don’t assume that the higher-ups know much more than you do. However, remember to pick your battles. If people don’t like you, they won’t execute on your ideas. With the right approach, your job can contain a bit less bullshit.