Intuition in a Data-Driven World

N.B. No corporations were harmed in the writing of this essay.

(This is from a set of old essays I'm putting back online.)

Structural Tensions

An increasing proportion of resources must be expended on overhead--administration, coordination, "work about work", if you will, as a system expands. Overhead's natural optimization impulse is a mid-century cybernetic fever-dream. Standardization, fungible components, feedback cycles, and automation.

(Management consulting was McMaster-Carr for X long before Uber for X was a thing.)

I don't pass judgement. Without structure, massive enterprises simply don't work--public or private. Reified management science serves a serious purpose indeed beyond being a rich target for artists' disdain.

The other side, of course, is that many system components are people, not servers or robots. And people are not their job description--the shoddy strictures of their "role" never captures the full picture.

Think for a moment about what your job description says. Then contrast that with what you actually do. Finally, frame both against the rich fabric of your full lived experience: travel, relationships, victories, failures, disappointments, realized dreams, and so on. An awkward juxtaposition, no?

Your impact comes from those blurry margins, humanity applied illegibly, if not secretly, to the common goals of your team, org, company.

Administrative Legibility

Doing some gradient of your "job" is table stakes for a place in the game, but you won't long keep it without the something special you bring--nor will the enterprise do particularly well without it. While common knowledge among people, precious few policies or management playbooks will even acknowledge this.

To adapt an idea from Scott's Seeing Like A State, the enterprise needs you to be legible in order to administrate the processes in which you are constituent (such as your employment), ergo the job description, millions of policies, many layered hierarchies both formal and matrixed, management information systems, and on and on and on. Again, this isn't inherently bad, it just isn't the whole story.

The more administratively legible you are, unfortunately, the less you bring the real stuff of your humanity to bear on your work. Usually then reducing your performance (among other things, like satisfaction..) and making the company's outcomes worse. Your immediate manager knows this, or ought to, but your distant skip probably doesn't (despite knowing the same for their directs).

(Tangentially, I think this is a particularly egregious contributor to the information filter that inevitably arises moving up through any hierarchy.)

We Are Data-Driven!

Now let's zoom back in a little. I suspect nearly everyone reading this has heard something akin to "we make data driven decisions," if not today, in the past week. I'll save my critique of how that usually works in practice. Instead, let's assume that we're talking about a more or less stereotypical org culture that embraces data, rationality, and metrics-governed management.

For some companies, the reality is... notional. Others truly live and breathe metrics--and others still fall everywhere in between. This matters to us for our purposes here because that "data driven decisions" interface is the primary administrative surface most employees interact with most of the time--think WBR, MBR, QBR, annual planning justifications, impact justifications during perf reviews, in writing docs for designs or proposed work or promotion justifications.

(See a theme? There's much to explore regarding justification supplanting judgment.)

Everywhere numbers, and in the most extreme cases nothing happens period without a metric to support it--which leads to all sorts of hi-jinks. Goodhart's Law, anyone?

I am a huge exponent of a rigorous quantitative picture underpinning all ops and most material decisions generally. Without the numbers, you're lost. However, paraphrasing Vannevar Bush, "Data Is Not Enough"--high function requires those illegible elements, too.

Intuition Revisited

Data holds precious little worth unless it's interpreted, contextualized, and put to work somehow. High quality decisions come from information. The brute existence of data is not equivalent to even a tiny piece of information.

Reducing the interpretation of the data that drives decisions to "more number more better" anchored to some KPI goes all-in on legibility at the great expense of effectiveness. The people who build, maintain, move, sell, support, repair, and use a thing will deeply understand it in ways that are hard to describe. Expertise both critical to your mission and most especially not legible.

When data and anecdote don't align, find out why.

The data might be wrong or the orthodox interpretation skewed. Whatever it is, figuring it out can only help you. For this approach to work, and it does, folks need to be empowered to share those anecdotes, to give voice to their intuitions without fear.

When they do, leadership needs to listen.

That sliver of humanity that sneaks into the great machine in the form of being a little illegible is the soul of operations. Instead of you--or your people--working in the dark, integrate a little intuition into your general ops activities. It's not an easy thing, but some illegible activities can become legible with sustained leadership support.

Embracing intuition enhances quantitative decision making by taking seriously inputs from outside of the system structures that capture and process your data. Computers do only what we tell them to do, and despite the cleverness with which we manage to do so, can't recover from incorrect instructions. Taking human intuition seriously is a mechanism by which you can suss out such issues over time.

An ounce of humanity will take you far keeping the math in check.

I almost always champion a middle path. Here, I no more support intuition-only than I do data-only decisions. Reductionist thinking is a luxury one can ill afford when playing a game without knowing the rules. The competence, command, and comfort that you're used to is gone.

Adopt a beginner's mind--it cannot be simultaneously, generally true that global polycrisis, the dawn of AI, deglobalization, etc. are harbingers of paradigm shift AND more of the same thinking, talking, doing that you've always done (maybe with a fresh coat of paint) to be a path forward. Honor what has gotten you this far without getting lost in it.

Succeeding in predicting specific outcomes using general theory is the pinnacle of applied knowledge--and a flow that we've all become accustomed to in the triumph of science over the past 100-150 years. The world is changing, in big ways that we don't yet understand, we have to shed this hubris and rebuild back toward answering thorny questions--potentially over and over again. Counting on having the cake and eating it, too, is a perennial seduction with a universally unfortunate end.

Practical Ways to Adopt Intuition

Intentionally integrating intuition into operations is practically effective for many purposes. I treat models for knowledge in another post, arguing for the merits of a rather peculiar one that requires the application of intuition to work well.

Like any cultural or mindset shift, there are no shortcuts--I don't have a secret Latin phrase for you to say three times over your JIRA board toward instant intuition integration. That said, here are some things that have worked for me.

For Managers

I normalize intuition by asking people for theirs and openly offering mine. Use stand ups, retros (dedicate a column and contribute to it yourself), regular syncs, one-on-ones, whatever your rituals are--asking questions like:

• Is there anything on your mind that just doesn't seem right? I'm not looking for a metric to justify your feeling, I'm sincerely interested in the way things seem to you.
• If you could do any work to improve things for our customers or make the team's life easier that you don't know how to justify, what would it be? Why that?
• Is there anything you're aware of that feels under/over-invested that you don't know how to justify pursuing? What do you think a world in which that work is done looks like relative to present state?

On-call hand off meetings are wonderful for this. The outgoing on-call just spent however long primarily looking at how things are going. On my teams, primary on-call is absolved of their project responsibilities during their shift. I recommend this for many reasons, not the least of which is establishing a mechanism for the whole team to look long and deep in turn at your domain across the year.

Whatever the details look like for your org, the most important thing is for you to explain what you're asking for, that it's okay to provide basically any answer, and that the org will seriously consider what folks raise in response to such asks.

In the beginning, especially if your org is deeply entrenched in quantitative justification, take the time to explicitly qualify these asks as not veiled metric-seeking. Remind your team that there is value in their perspective and that the team can benefit from it. The flip side is equally important. When you have an intuition that something isn't right, pursue it! If you frame your investigation in the language of intuition, your team will see that it's okay.

When you start seeing your people ask each other such questions, you've probably succeeded in normalizing and internalizing intuition as a tool for farming non-obvious ops inputs.

Another thing that I find profitable to do a couple of times per year, is to ask in one-on-ones what docs they've written or know about that never went anywhere but feels like they ought to have. Doing the same when talking to partners, leadership, peers, and so forth can also produce excellent results. In particular I find that, coming from the tech side, these questions can do a great deal of work for me in refining my understanding of what my non-technical partners actually need.

Most critically, however, is acting on intuition, investing a little capacity here and there in things you don't have an a priori case for. Not being able to invest in every such thing is no argument for opting for none. Without action, you not only won't affect culture change, you'll expend the good faith of your teams for naught.

For ICs

This varies wildly by level and org. Tenured ICs with a lot of moral authority in the org can proceed very similarly to managers. Less tenured and early-career folks have less agency to do so in many places.

What you can do, however, is pursue your intuitions anyway and use the exercise to demonstrate the value of doing so back to the team. This might look like putting a doc together about what troubles your mind and socializing it to your seniors and management.

If you are in a context wherein saying "no metric but take me seriously" just doesn't work, I recommend framing your doc as seeking wisdom on how to measure the dynamic that interests you. This keeps you in the fold but knocks on the door of your concerns anyway.

Wrapping Up

The intuition that has gotten you this far in life matters, even in sterile, heavily instrumented technical environments. Integrating intuition into operations provides local, practical benefits for teams and can counter the dehumanizing impulse of administration at scale toward better net outcomes across the board.

I particularly love that embracing intuition often has the knock-on effect of encouraging more voices to speak up. I find that a greater diversity of my team participates actively in a broader swath of situations in spaces that embrace intuition as a first-class concern.

"Only human" is a common euphemism for imperfect, flawed, irrational, and messy, lacking. We so easily forget that we are the most complex, incredible, and completely unparalleled constituents of the known universe. The highest achievements of the most advanced technologies don't even remotely compare to the hardware between your ears.

Make room for a little messy humanity.