Pro tips on growing data participation, protecting your data, increasing diversity, and more
A few weeks ago, Fivetran hosted the Modern Data Stack Conference (MDSCON) 2021, a virtual conference to empower data-driven decisions that transform businesses, teams, and careers. We at Atlan were stoked to attend and shared notes from our favorite sessions throughout the conference’s two days and 40+ sessions.
Since then, while mulling over the stories and insights we heard, we realized that there were a few common themes. Across sessions, different speakers approached the same set of questions in different ways: How can we increase data participation? How can we protect the data team and its data? How can we rethink traditional ideas about data today?
For those who missed the conference, and for those who were there but couldn’t attend every session, here are five key ideas and takeaways from MDSCON 2021.
Grow data participation
Multiple speakers spoke about the importance of increasing data literacy within your organization. Every company that wants to become data-driven knows this is a problem they will need to tackle, but actually solving it is a huge challenge.
Callie White and Jacob Frackson from Montreal Analytics emphasized the importance of creating a great information architecture before you even think about data literacy. As you build your data system, think about what different types of people need out of your company’s data.
- What is the user experience?
- What is the maintainer experience?
- How can we simplify or improve those with information architecture?
Bake these personas and their needs into the planning of your information architecture, rather than trying to meld the system around them after it’s built.
Nelson Davis from Analytic Vizion also emphasized the importance of thinking beyond data knowledge. Rather than focusing on the number of people who understand your organization’s data, focus on people’s ability to interact with and act on that data.
Nelson pointed out that people adopt data just like they adopt new technology. Enthusiasts will latch onto data early, but for an organization to be truly fluent in data, it has to penetrate past that 16–18% of early adopters. True data participation requires targeting the remaining majority who may be more data-hesitant.
Create a culture where the majority of the people are using data to make decisions, not just using data.Nelson Davis, Analytic Vizion
- Add data landing pages with top reports, new releases, and important links.
- Demo new tools and offer targeted trainings to relevant teams.
- Promote and incentivize internal learning sessions — e.g. Office Hours to help business people learn how to generate quick insights.
- Partner with power users, who can champion self-service analytics within their teams.
- Identify “citizen data analysts”, a community of subject matter experts who will socialize their findings for other stakeholders and teams.
Protect the company’s data
As speakers talked about opening up an organization’s data, they also emphasized the importance of keeping that data safe. Giving the team authority and autonomy is important, but make sure that it comes with guardrails.
We want centralized governance but to keep federated analytics.Rashmi Agrawal, Oldcastle APG
As Nelson Davis and Rashmi Agrawal explained, don’t just let people download data from dashboards. That’s a security risk, because people can take the data, put it in Excel, and create their own numbers. This creates multiple versions of “the truth” as data and data users proliferate.
Instead, make sure that data tools like dashboards are well-designed so people can get the insights they need. Enable people with tools to let them explore and innovate, but pair those tools with guardrails to ensure safety and consistency.
Protect the data team
Brittany City from Asurion spoke from her personal experience about analysts’ struggle with wearing multiple hats in their role. From organizing data to creating dashboards to coordinating with other stakeholders, analysts can often feel like middlemen.
Emilie Schario from Netlify called this the “service trap”, where a team is trapped in never-ending requests for creating stats and proving their work, rather than focusing on driving impact. Servicing requests is part of a data teams’ work, but it’s not their core mission — to create insights and drive impact.
If we spend all of our time answering questions, we will never deliver insights.Emilie Schario, Netlify
This is part of why it’s so important to empower everyone to become analysts — it protects your data team’s time.
Rather than becoming the middleman between a company’s people and its data, focus on making data easy to access, identifying what data can be self-serve, surfacing insights publicly, and growing data knowledge. Only then should the data team serve requests that aren’t addressed by these processes.
Reverse traditional patterns
Another thing that multiple speakers emphasized was the importance of flipping our data challenges and assumptions on their head. Here are three great examples from David Cohen at WW International and Nelson Davis at Analytic Vizion.
First, don’t forget about the simplest solutions. David pointed out the irony of how we always look for technical solutions to help data teams move faster, “overlooking the power of humans talking to each other”. In his team, they decided to meet once a week and “just talk about the data”. The team may not have had all the data they needed, but it allowed them to immediately align on what decisions could and couldn’t be made.
Second, David also pointed out a common fallacy of so-called data-driven teams — using data to confirm what they already think and reinforce existing decision-making. Instead, truly data-driven organizations need to fall out of love with their ideas and look for “uncomfortable” data, because being uncomfortable is what makes organizations grow.
Third, Nelson explained the value of pushing authority downward, rather than data upward. Today the typical decision-making model is that an analyst creates data, passes it up the chain of command, and a leader makes the decision. This isn’t ideal, since it separates those with the most context on the problem (leaders) from those with the most knowledge about the data (analysts). Instead, leaders should flip this decision-making model to push authority and context down to those who actually have the data, rather than pushing data and reports upward.
As data becomes more and ubiquitous, the decisions are going to become more data-driven, regardless of whether we hold onto those decisions as decision-makers.Nelson Davis, Analytic Vizion
Never forget diversity
One of the conference’s most powerful talks was from Sadiqah Musa and Devina Nembhard, both analysts and co-founders of Black in Data. Their talk was deeply personal, based on the exclusion they often felt within multiple groups in the data industry.
They reminded us that the future of the data industry will also be found in its people, not just in the industry itself. Diversity in teams has repeatedly proven to be a driver of desirable, lucrative business outcomes — from better teams to better financial performance.
There is no “silver bullet” or overnight solution, but here are some of their tips for building a more diverse and inclusive data team:
- Attract diverse talent by working with specialized recruitment groups, publishing your ethnicity pay gap and transparent salary bands, and building inclusivity and representation into your brand and careers materials.
- Hire diverse talent by incorporating diverse interview panels throughout recruitment, ensuring that people of color meet their team before they accept the job offer, and creating objective hiring practices and objectives.
- Retain diverse talent by building a welcoming onboarding process, constructing a truly inclusive environment, conducting authentic anti-racism training, creating measurable diversity-focused business objectives, and being open about the organization’s ethnicity make-up and plans for improvement.
Loved an insight or session that we didn’t cover here? Drop a comment with your favorite takeaway from the conference!
This article was originally published on Towards Data Science.