Self-serve analytics and other corporate accountability sinks

One of my scorching-hot takes about corporate life is that executives and managers should strive to see reality clearly. While this seems uncontroversial, it isn’t, which you can tell by witnessing the uncomfortable fidgeting that occurs whenever a company’s most intelligent and grounded employees watch their CEO talk gibberish at an All-Hands meeting. (One day, I’ll launch an IoT startup that measures this ambient discomfort and sells it as an “Engagement” score to gullible executives; like all such scores, it’ll be a busy box for them to obsess over so their employees can get on with actual work.)

Given my pro-reality stance, it’s time I came clean: I’m as susceptible to reality avoidance as even the most credentialed consultants and executives. This harrowing realization came to me recently when I found myself earnestly discussing that Loch Ness Monster of data work, “self-serve analytics.”

Self-serve analytics is the idea that anyone should be able to do their own data analysis, regardless of their technical ability. While the premise is somewhat misguided (I don’t pretend to “self-serve” marketing or software engineering), its goals are noble. Let’s “democratize data!” Let’s “empower domain experts to leverage the data they need when they need it!” Let’s “free up data resources by taking ad-hoc requests off their plate so they can focus on high-leverage strategic work!” (Alas, nobody asked for this high-leverage strategic work, which calls into question whether it is, in fact, high-leverage or strategic, but that’s a topic for another day.)

Why do initiatives to deliver self-serve analytics fail? Perplexity has a compelling list of reasons here. I would especially co-sign the difficulties associated with documentation and user-friendly tooling. But if I might take a dour perspective for a moment, the fundamental reason is simpler: Stakeholders don’t actually want self-serve analytics, or at least not in the ways Data Teams want them to want it.

Data practitioners are from Mars, non-technical stakeholders are from Venus

After interviewing scores of non-technical stakeholders about self-serve analytics and observing their use of data, I have the perfect solution to address their needs. Are you ready to see it? I must warn you that this product is genius. If I didn’t have an upcoming fantasy football season to focus on, I would totally start a company to productize this idea on a flimsy LLM foundation.

Are you really ready?

Okay, here it is:

“Wait!” you say. “That’s just a white square!”

That’s where you’re wrong.

The key thing about the square is that it outputs the correct number or chart 100% of the time, tailoring the results to the specific wants, needs, and — critically — domain context of the stakeholder. There is no nuance the white square isn’t aware of. There is no strategic implication it fails to consider. If a stakeholder’s team defines “new user” slightly differently than another stakeholder’s team, it will know. Its main feature is that it always knows.

Incredibly, it even knows without manual prompting or painful configuration. When a stakeholder lands on the white square, the perfect visualization or metric is presented. This is true even when the person doesn’t have a specific question in mind! The white square realizes this and outputs the perfect statistic in the precise format that will earn said stakeholder the admiration of peers and supervisors. (That is, if their peers and supervisors actually pay attention on Zoom.)

Meanwhile, back on Mars…

The only problem with this vision is that the overworked Data Team often doesn’t realize this is what people want when they ask for self-serve analytics. The typical data practitioner hears “self-serve” and translates it into something like:

Phenomenal! There is so much straightforward ad-hoc work for them to take on. And it’ll be so easy! They only need to learn this tool and read the relevant documentation. Oh, and I guess learn about the nuances of how Finance and Product think about retention differently. But that’s fine. They must already understand that to do their job, right?

Sorry, data practitioners, the white square has spoken.

It may seem like I’m making fun of our stakeholder and data practitioner pals, and that’s because I am. But I also empathize with both parties.

In many non-technical roles, the use of data is sporadic and inessential, a nice-to-have rather than a must-have. (This is especially true when the CEO professes to be “data-driven” while only requiring data for decisions he disagrees with.) In such cases, learning how to self-serve data has an uncertain payoff at best.

Meanwhile, on the data side, there genuinely is so much exciting proactive work to take on and so little bandwidth with which to do it. For data practitioners, self-serve is the white whale that finally enables execution on the Data Team’s mythical roadmap. Can’t their stakeholders meet them halfway?

---

Here is where my take gets controversial:

Although the impasse I’ve sketched out is at least partially solvable (I promise to explain how in a future blog post), in many dysfunctional organizations, it’s never meant to be. While the promise of better decision-making and superior utilization of expensive data science resources are captivating dreams, dreams are all they are. In such an environment, self-serve analytics becomes a convenient lie — an elaborate cover-up for poor prioritization or inadequate resourcing. A punching bag that absorbs energy and attention from the company’s more profound problems.

It’s an especially advantageous lie because self-serve analytics is so naturally tricky to get right. There is something between the white box and the data practitioner’s SQL tutorial that will solve 80% of the use cases self-serve is supposed to solve. But what if we did something worse instead? What if we implemented a solution that is too shoddy to be genuinely helpful, and what if the organization put no muscle behind training employees and setting the cultural expectation that data should be incorporated into the modal decision-making process?

Um… This would be bad, right?

It would be if you aimed to create “Actual Business Value” (bless your heart). But again, you must remember that your organization — like 94.44% of all organizations (thank you for the statistic, white square!) — is dysfunctional.

In a dysfunctional environment, shoddy self-serve yields several critical benefits.

First, consider the stakeholder. When challenged on subpar results, they can blame their lack of timely access to data for their failings. Again, this is a wonderful excuse because it sounds important, even if timely access to data would have provided little in the way of practical short-run benefit. And because the self-serve problem is such a boondoggle, it buys them and company leadership many quarters of diversion and delay. It becomes a bike shed to paint, and painting bike sheds is always more fun than addressing the foundational issues — personnel, prioritization, expectation/standard-setting — that actually drive results.

Next, consider the data manager, who, for reasons I’ll one day break down in a blog post/doctoral thesis/therapy, has a uniquely stupid job in many organizations. (Stupid is a statistical concept provided to me by the white square, so it is once again indisputable. Sorry.) The data manager is fenced in by extremely intelligent reports who understand intimately how much better things could be, and leadership that is at best vaguely hostile to the manager’s pleas for better prioritization and partnership between data and stakeholder teams. For this poor intermediary, self-serve analytics becomes a life-preserver. Rather than solve the genuinely challenging organizational problems that hold data teams back, data managers take out their paint brushes and go to town alongside their counterparts in company leadership.

“I know we haven’t gotten to our roadmap,” the data manager tells their team, paint dripping onto the ground, “but if we spend a couple of quarters building out self-serve, we’ll free up so much time for high-impact initiatives.”

At the risk of layering on another metaphor, self-serve analytics becomes a carousel for management to ride. It looks exciting. It feels exciting, and with the wind flowing through one’s hair, it can even feel like progress is being made. That is, until managers look up one day to discover that all their best employees have left, which seems like it would be a concern until we recall the key word: Dysfunction!

Sadly, many organizations — precisely 94.44% of them — would prefer to create intractable issues to obsess over rather than focus on the challenging and potentially fraught organizational concerns that prevent true excellence.

Don’t get pulled into corporate accountability sinks

This post contains a meta-lesson, one that I almost feel guilty sharing but that is critical if you’re hoping to climb your local corporate ladder (bless your heart): Some problems are not meant to be solved. This is true even — and perhaps especially — when an organization complains about them loudly and persistently. (Embarrassingly, it took me years to understand this.)

These low-status problems are a special case of the broader concept of “accountability sinks,” the unforgettable name for entities designed to absorb blame for an outcome but who purposely don’t have the authority to address it. (Think of the poor gate agent who must deal with customers angry about a delayed flight but who can’t provide redress.) Accountability sinks shield the overall system from blame, much like self-serve analytics shields incompetent management from blame.

Alas, once you learn about the accountability sink phenomenon, you see it everywhere. People processes are particularly prone to becoming accountability sinks: “I wish I could explain our compensation system, but I think HR is working on a refresh, so hold tight!” Or “I know the leveling system at our company is broken, but that’s why we formed a cross-functional task force to address it!” For every Netflix that is ruthlessly specific about its approach to talent, there are (roughly) nineteen companies that allow HR to become one big accountability sink so that employee dissatisfaction can reliably be channeled into a topsy-turvy world of bureaucratic obfuscation.

Learning to distinguish between low-status accountability sinks and high-status problems to be solved is an essential skill. (You could also try working for one of the 5.56% of organizations that isn’t dysfunctional, but that’s a more challenging career path to recommend.)

Sadly, I’ve noticed that the people who see or intuit these dynamics often leave management. But they are precisely the people we need in management. So if you’d indulge me in one last meta-meta lesson: Accountability sinks aren’t inevitable. Even when they’re convenient, they’re often unintentional — the result of poor process or unclear roles and responsibilities rather than deliberate malice. If they’re recognized for what they are, they can be overcome (again, I promise it’s possible to implement a version of self-serve analytics!). However, recognition is often the biggest hurdle: It requires someone — you — to have the courage to tell the truth about the polite corporate fiction.

And if the organization insists on continuing to live a lie, you know it’s time to polish the resume.