Analysts are ecosystem-builders
How to understand purpose and the vectors by which we pursue it
👋 Hello! I’m Robert, CPO of Hyperquery and former data scientist. Welcome to Win With Data, where we talk weekly about maximizing the impact of data. As always, find me on LinkedIn or Twitter — I’m always happy to chat. For those who have been following me, this is a more measured take following my last article on truth. 🙂
For the last couple years, I’ve felt the prevailing narratives around analytics have grown increasingly out of touch with day-to-day practitioner needs. We've been distracted by hype trains that encourage us to build novelties of questionable value, we've built and rebuilt labyrinthine nests of data models in an endless pursuit to align with the ideals of the Modern Data Stack, we've ridden the bullwhip and succumbed to our hermit IC minds, rather than building things that optimally solve real problems. And in due course, we've found ourselves buried in maintenance work and half-assed proof-of-concepts, doubt finally seeping into the minds of our stakeholders. We wanted to avoid updating that spreadsheet once a week, and so we requisitioned cathedrals.
But all of this data navel-gazing begs a single eigenquestion: what is our purpose, and how should we measure it? We must not know, otherwise why would we spend so many resources on excess? Perhaps we’re just getting nerd-sniped, but even so, let’s spend some time discussing purpose so we can make sure it doesn’t happen again.
So what is our purpose?
Our purpose, plain and simple, should be impact. If we are not generating business value, we are not generating business value.
The problem, however, is that value is both difficult to measure and wildly heterogeneous in practice. Owing to the former, we seek proxy metrics. Yet owing to the latter, no single proxy metric is ever comprehensive -- our yardstick varies from analytics-attributable decisions, to team alignment and inspiration, down to visibility metrics (sorry, though I do think this is generally a good thing). We argue about which of these should have primacy, but the problem is that none of them ever will, because none are synonymous with impact. As much as I want to believe that our objective should be truth, this isn't always right.
That said, I've realized there is a way of viewing our charter that makes things quite clear:
We are ecosystem-builders. 🌱
We craft environments that help others in the organization make decisions, set strategy, develop intuition, be more data-driven. And there are three primary ecosystems which we analysts can construct and through which we can provide value:
📊 Data ecosystems.
By defining and giving access to raw data. Dashboards, data apps, self-service BI. This might lead simply to basic visibility into core metrics, it may spur folks to toe the party line. But at the end of the day, these all commonly come by simply serving raw data with little to no story-telling.
📕 Knowledge ecosystems.
By providing interpretations of data and access to these interpretations. Analyses, notebooks, reports, not dashboards. Data, but with interpretation.
🧠 Truth-seeking ecosystems.
By being the voice of intellectual honesty in key conversations, to ensure reasoning around both data and its interpretations is correctly incorporated into decisions, strategy.
And here, data primitives are a subset of knowledge primitives are a subset of truth-seeking primitives (arguments).
Should we try to figure it out? Well, dashboards aren’t dead, but having a million trashboards isn’t good for your mental health. On the other hand, while seeking truth sates our rational minds, I’ll admit, truth isn’t what everyone always needs — sometimes a rallying cry, a story around an uptick in a dashboard can drive a business forward more than stringent honesty.
And, this ambiguity is the core of the problem — the pathways by which data can create value are nonlinear, unpredictable. We quibble over which of these pathways is paramount, but the truth is that there is no hard line truth here. Sure, I’m a strong advocate for building knowledge and truth-seeking ecosystems, but dashboards can have incredible value as well.
So the best I’m able to give you today: it depends. But here are some things to consider:
Good analytics always looks different. B2B companies, for example, have less data than consumer businesses, so the primacy of data ecosystems may make sense.
“[If] large amounts of context live in the heads of stakeholders, not data people, transferring it over takes a lot more overhead than in other industries. Self-service and dashboards are therefore critical, to minimize context transfer.”
The growth stage of the company
Growing companies have different needs from steady ones. Growth means new surface areas, unpredictable questions, sometimes even adjustments to top-line metrics. And dashboards often can’t keep pace. Analyses will inevitably be your bread and butter. On the other hand, stable businesses often have needs that are understood and predictable, and a culture of dashboard-first could make sense. If most key interpretations of your data have been etched into the minds of decision-makers, data may be self-explanatory, and you just need to expose it.
The final piece of the puzzle — your data, your knowledge, your truth-seeking efforts won’t matter without stakeholder buy-in. Analytics is a domain where our value is predicated on its reception and incorporation in the decision-making process, and so you are going to be inevitably beholden to the rest of the company. But make your case, of course.
Why this is important?
Because it can shape everything. If you’re selling microwaveable foods (dashboards) vs. cooking meals for folks (analyses), the problems you face, how you optimize for scalability — it all changes.
One obvious ramification, for instance, would be around choice of tooling: as much as I tout that Hyperquery can do everything, we’re going to be a less feature-rich BI tool than something like Tableau or Superset (for now), where dashboarding is their entire focus. That said, if you’re drowning in custom work and see a need to be more involved in strategic decision-making, Hyperquery could reinforce better practices and allow you to be more nimble, pushing you towards the knowledge and truth-seeking layers.
Another ramification: knowledge and data are maintained and scaled differently. In a dashboard-first world, for example, you might care deeply about data quality, reusable data models, cost optimization. Conversely, in a knowledge-first world, you’d worry about visibility and comprehension. The pathways to excellence, and thereby, maximal business value, are different.
Thanks for reading Win With Data! Subscribe for free to receive new posts and support my work.