Data teams are often run like service organizations: a request goes in, an answer comes out. Can I get last week’s revenue numbers? What percentage of our website traffic comes from Instagram vs. Tiktok? And to stem the ceaseless flow of questions, we do what any reasonable analyst would: we answer questions faster, we build dashboards on dashboards, we push back, and, failing all that, we jump the chasm and try to teach our stakeholders SQL.

Yet this particular cocktail of solutions is both all-too-common and all-too-commonly ineffective. Your half-answers lead to more questions, your relationship with stakeholders grows increasingly transactional, and your stakeholders’ attempts at writing SQL leave you with refrains of “why isn’t this select * from query taking forever?”. And one day, you’ll realize that you and your colleagues have become SQL monkeys, and you’re drowning in a trashboard cesspool of your own creation.

The root cause of this chaos: we’ve failed to define best practices around how we should be collaborating around analytics. In this post, I’d like to nail down once and for all:

• How we should be thinking about analytics collaboration (what is it, even?)
• What good analytics collaboration entails
• What you can do to get there

The key lessons: align better and find a better place for your work.

What is analytics collaboration?

I’d like to start by defining analytics collaboration, perhaps a bit more broadly than you might expect:

Analytics collaboration is the process of collaborating on analyses, and it can manifest in one of two ways: as synchronous analytics work or as asynchronous consumption of an analysis.

We traditionally think of collaboration only as synchronous collaboration — the act of directly working with others towards some decision. The work is produced by you, then delivered. Impact is circumscribed within a relatively small circle of stakeholders and ensuing decisions. Sometimes the work will be a joint effort with other analysts or your stakeholders themselves, but the pathways of collaboration are predictably linear.

But this narrow view of the world leaves a whole class of collaborative endeavors unspoken for: asynchronous collaborations spurred when others find your work. And this is where analytics work can be truly scalable. We endeavor to attain this level of scale — it’s generally why we build so many dashboards — but we rarely execute on this well. While dashboards enable a particular type of consumption-at-scale (raw data consumption), we often fail to consider how we can scale our insights and learnings (interpreted data consumption).

Any strategy for improving how analytics collaboration occurs, then, should incorporate both aspects of collaboration, addressing the pitfalls of each.

What makes for good collaboration, and what we’re doing wrong

Good synchronous analytics collaboration hinges on alignment.

Good asynchronous collaboration hinges on documentation.

The problem with data organizations is that we often do neither of these things well.

Good synchronous collaboration: alignment

The first pillar of collaboration is alignment. Alignment with stakeholders should be a prerequisite for every piece of analytics work. It ensures that the analysis you produce, the dashboard you create, or the request you field are intimately coupled to the intended business impact. Unfortunately, analytics organizations tend to find themselves in loops where alignment falls by the wayside.

Consider, for example, the ad hoc request and the collaborative pathways it engenders. Ad hoc requests are direct, inbound requests from stakeholders. It’s the tendency of overworked analytics organizations to try to minimize the amount of time we spend doing ad-hoc work — and for good reason. Stakeholders view ad hoc requests as “quick questions”, and analysts hate this kind of work. We thus attempt to optimize this work by making it smaller — answering it as quickly as possible. The standard approach looks something like this:

The problem with this approach, though, is that:

1. It rarely saves time, while producing a lot more frustration.
2. A lack of alignment means the wrong question is often answered. The most common symptom of this problem: lots of follow-up requests. One request turns into 20.
3. We act as SQL monkeys, therefore we become SQL monkeys.
4. By answering reactively, you’re setting expectations around the kind of value you’re providing, and that value is minimal. Behind every question is a more strategic question, and rarely are stakeholders so data-proficient that the latter will surface first. If you’re able to surface these, you can raise your status as a thought partner. If not, you’ll be forever relegated to acting as a human interface for your organization’s data warehouse.

This story plays out even more disastrously for other flavors of data work. Dashboards without prior alignment lead to floods of unused dashboards. More proactive work — longer-term analytics initiatives — can suffer from the same alignment problems, but worse, analysts can end up working on these for months before realizing the solution doesn’t provide the value that was assumed.

Good asynchronous collaboration: documentation

The second pillar of analytics collaboration is documentation. Good documentation enables your work to scale past the single decision it was intended for.

Consider the figure from the first section of this post — it depicts a world (below, left) where your work has been properly documented and placed in a central, discoverable place. Numerous other stakeholders and analysts can find this, draw from it.

Now consider a world where your work disappears (below, right). Those stakeholders now have to turn to their analysts for help. Analysts have to re-do all the work you’ve already done, leading to an immense amount of duplicate work and wasted effort.

This hornet’s nest is notoriously difficult for analytics organizations to internalize, let alone untangle because it’s effectively invisible. No individual analyst feels the pain, because it’s nearly impossible to detect duplicate work. But consider the infrequent instances where you discover your coworker did the same analysis you’ve done, and recognize that the number of times it happened and you didn’t notice it are substantially higher.

Asynchronous collaboration is the only scalable form of collaboration, and it’s impossible without a strong culture of documentation.

Your new workflow

The solution, at this point, should be clear: make alignment and write-ups standard practice for your organization. Below, I’ve indicated what your new analytics workflow should look like, were collaboration to operate more optimally.

Note I’ve added “Research” in as part of the new collaborative workflow. Once work is written up, research into past work is possible and should be the first step before the start of any new work.

The final step: consider better tooling

Let’s take one more step back. There are actually two levers to improving collaboration:

• The process by which it happens.
• The platform over which it happens.

While you’ve heard me discuss process quite extensively at this point and how to improve it (alignment & documentation), choosing the right platform for collaboration is equally, if not more, important. Forcing alignment on business objectives and writing up your work are seemingly simple process changes, but the greatest barrier to success here isn’t setting these processes - it’s getting people to do it. And in our experience, the culprit here is always the use of high-friction tools.

If you’re using an IDE as your main point of communication and Slack as your main point for delivery, no amount of forced process change will save you from poor collaboration practices. These tools are not made for documentation or establishing alignment. A Google doc or Notion are better, but what you gain in ease of alignment you lose in reproducibility.

We’ve built Hyperquery — a workspace for data collaboration — with these problems in mind. Hyperquery is our attempt to make a platform that’s built from the ground up for collaboration. Alternatively, you’re welcome to cobble together existing tools to make this work, but don’t say I didn’t warn you.

Conclusion

Strong analytics collaboration practices are key to growing an analytics-driven, high-leverage organization. And encouraging proper alignment with business objectives and a strong culture of documentation are critical in ensuring that analytics organizations operate as advisors, not help desk technicians.

And if you want to learn more about how hyperquery can drive adoption of these processes and effortlessly elevate analytics collaboration within your organization, check us out at hyperquery.ai or contact me at robert@hyperquery.ai.

‍