More Than Numbers

Top lessons from top data teams

No items found.
Taylor Brownlow
April 30, 2024
May 20, 2024
7 min read
Share this post

It’s healthy to step back and take stock every so often, which is what I did for today’s article. I’ve done my best to distill the top lessons I’ve learned over 100s of conversations with data leaders over the last 6 months. Unsurprisingly, this was a challenging exercise that I hope is as useful for you as it was rewarding for me.

Learning # 1: Have a purpose beyond ‘helping the business make better decisions’

As I wrote in this article, the default mission statement for most data teams is ‘to help the business make better decisions.’ The best data teams I see can get far more specific than that when defining why they exist.

“You can have that high level objective, but that can be interpreted in a million ways, so you need to have an understanding whether implicitly or explicitly what you own and what you don’t own, what you do and what you don’t do.” - Conor O’Kane, Head of Analytics at Cleo

Some examples of this are:

  • A data science team at a fintech company has assigned themselves 2 core business metrics to manage. This is only possible because the models they build directly impact those metrics, and is therefore not possible for more centralized data teams.
  • A product analytics team fully aligns themselves with the goals and outcomes of the product organization
  • Creating a subset of the central data team whose mission is to solve the company’s biggest problems
  • A central data team with embedded analysts aligns themselves with the objectives of all of their partner organizations, detailing how data contributes to their goals

The most important distinction here is that these teams are able to say specifically how data is contributing to the wider business objectives.

Learning #2: Take responsibility for operational clarity

I’ve heard this described in many different ways. One data leader referred to it as cultivating ‘connection’ in which the entire organization has a shared understanding of where the business is at, and where the biggest problems are.

Often this manifests in sitting down with business leaders and mapping out their world - what is the growth funnel of the business? What are the key metrics you’re looking at, and how do those break down into sub-metrics? How does the lead qualification process work at scale?

Data teams that take ownership of this for the rest of the organization benefit from a common understanding of the existing world, and the biggest problems, which helps alleviate so many downstream issues many data teams run into (e.g. different definitions of the same metric). These teams also have very strong relationships with the rest of the business and seem more immune to the ‘do I even need a data team?’ line of questioning we see today.

“You might have this moment where you think ‘Why am I doing this? Shouldn’t it be marketing’s job to understand how to measure their own market? Why do I have to do it?’ That voice will always be there. Part of our role is really understanding other people’s challenges and helping them do that better with the expertise we bring. … Maybe it is someone else’s job to do these things, but sometimes we have to do that work because without it, you can’t succeed.” - Alan Cruickshank, Insights Director at
Metric tree made in Count. Learn more at this event.

Some examples of this are:

  • Metric trees show not just key metrics but the drivers behind those metrics
  • process flow diagrams help clarify how systems work and identify gaps
  • customer journey maps are commonly used in product domains, but paired with data, have the power to drive decisions incredibly quickly

Learning #3: Ask ‘So what?’

Prioritization is a common challenge for data teams, and the most direct approach I’ve seen to that problem is by simply asking ‘So what?’

“I could see people were working on really exciting projects but my question was ‘So what?’ We create this model, but what is the value to this to the business?” - Anonymous

Being ruthlessly focused on what is delivering tangible value to the business prevents teams from falling into the trap of optimising the wrong things like improving the performance of data pipelines because it can’t hurt vs optimizing those models so you can increase time to decision for another team.

Some examples of this are:

  • One team replaced all of their data scientists with decision scientists to encourage everyone to think about the ‘So what’ of their work rather than just the technical aspect
  • Another team separated ‘BI and dashboarding’ from the main analytics team, and gave it a headcount of 20% the size of the core analytical team - making it clear that dashboarding wasn’t the way they were going to drive value
  • During team meetings, make sure everyone can articulate the business impact of the work they’re doing, and if they can’t work with them to find out what it is
  • Yordan’s story of looking into cost savings not as a ‘why not’ exercise, but using the project to make sure he was better serving the business

Learning #4: Don’t run from working with the business

I’ll never forget a talk I attended once in which the speaker advised everyone to make the data requests process as painful as possible in order to discourage people from doing it, thus giving you more time to work on your dbt models.

This idea of putting up walls between data and the rest of the business is still pretty common, but over the last few months, I’ve noticed that the most successful teams lean into that discomfort.

“I realized we weren’t following this. We were just going along with whatever was asked of us. Someone would ask for something and we’d just say ‘sure, we’ll do that for you now.’ We never pushed back and asked if that was even the right question to ask. We didn’t interrogate the data or the approach we were going to take. We just acted. And if we were going to start building up trust again, and trying to hit our timelines, we were going to have to change that.” - Anonymous

Some examples of this are:

  • Having regular touchpoints with the business, like weekly KPI reviews
  • Encouraging iterative, agile workflows that have more touchpoints (and are usually faster), like spikes
  • Teaching everyone on your team the soft skills needed to go from ‘Can I have a dashboard’ to ‘Tan you help me with this big problem?’

Learning #5: Creating a team people want to be on

Lastly, these data teams go above and beyond in creating a culture that people want to come back to. Interestingly, few data leaders realize they’re doing this. It comes through in the questions they ask like ‘How are other data leaders making sure their team goes to conferences and has space to develop?, or from practices like this one:

“At the end of every week, I ask my team 4 questions: (1) Did I create value today for myself, my team, or, preferably, the company? (2) Did I feel successful today? Did I achieve what I set out to do? (3) Did I learn something or teach something to somebody else? (4) Most importantly, did I ensure I did not take it all too seriously and broke my neck while wanting to achieve the former three?” - Anders Kring, Head of IT Development, Data & Analytics at European Energy

Some examples of this are:

  • Creating a clear career progression for all of your data roles (e.g. an analyst isn’t just an entry-level data scientist, they have their own path forward)
  • Creating a team that values experimentation, trying new approaches, and sees mistakes as lessons rather than failures
  • Celebrating success by sharing data team projects and stories across the business
  • Leaders that give credit to their team rather than taking it for themselves

Lesson #6: Don’t be afraid to ask for help

Each organization is different, each team is an assembly of different personalities, strengths, and challenges. No one I’ve spoken to believes their work is done, or that they’ve tapped into the full potential of what a data team can be.

And yet data leaders are often tasked with this daunting task all on their own. Finding spaces to connect with other data leaders is something that helps these leaders try different approaches, learn what mistakes to avoid, and just have a place to vent about things that have gone wrong.

What this looks like:

  • Finding local data leadership events (We run some in Europe - you can see the calendar here)
  • Finding a virtual community of data leaders (We have started a small one recently. If you’d like to learn more, check out the details here!)
  • Messaging a data leader in your local area and asking them to grab a coffee

Note: If you’re looking for a data leader to get advice from on a certain topic, feel free to reach out to me and I’ll see if there’s anyone I know that can help!