5 Ways to tame the chaos of data democratization
Data democratization remains a hot topic. More organizations want to break down data silos and give business users the data they need to make key decisions.
But in the rush to more openness, are too many companies creating an unregulated data anarchy?
Granting access to all data for all users may sound nice. But it isn’t a practical or scalable approach in practice. Some data—such as Personally Identifiable Information (PII), or organizational financials—is simply too sensitive to share broadly. Data exists in a hierarchy. Some data needs to be more rigorously controlled, filtered, and verified than other data.
dbt Labs held a webinar with three experts in the industry to answer a burning question: Is a “data democracy” actually obtainable? And if so, how do you prevent it from becoming a data anarchy?
Our host, dbt Labs Sales Director Zola Petkovic, moderated a lively discussion with Natalie Greenwood, Managing Director at Analytics8; Michael Colella, Senior Director at AXS; and Liz Connors, Senior Director at Mission Lane.
Our panel shared their insights on how to enable a true self-service analytics with strong governance. Their takeaway? It’s possible—but takes a different way of thinking. We’ve distilled their insights for you below into five principles to guide your company as it seeks to tame the data democratization beast. To learn more, watch the full webinar recording here.
1. Don’t fear adding roadblocks
A more open, democratized approach to data can lead to greater data discoverability and better decision-making. But it can also go too far to the opposite extreme.
Some orgs have grown so tired of a “data monarchy” that they move to a free-for-all model more akin to what Connors called a “data anarchy.” Teams go off on their own and create datasets, views, and reports that duplicate each other.
“You look at their dashboards or their tables,” Connors said, “and you’re like, I think that 60% of this could be eliminated, because it’s really just a slightly different view” of the same data. That breeds confusion and mistrust in data, as no one knows which derivative is the correct or complete view.
One solution to the problem of data confusion is certification. Certified data sets and dashboards provide a company’s official imprimatur for critical data. This can be data that’s widely used across the company. Or it can be “high-stakes” data used in critical decision-making or regulatory filings, where accuracy is paramount.
Processes like certification may lead to the roadblocks and slowdowns that advocates of data democratization fear. But Connors said that’s not always a bad thing.
“I’ve seen some million dollar decisions made on bad data—which is way more expensive than if we had taken the time to certify.”
Once you have your data, says Colella, understanding what data users need and building out the requisite dashboards “takes care of 80% of those self-service needs.”
2. Choose easy-to-use tools that offer transparency
Webinar viewers asked our panel what roles tools play in data governance. Specifically, how do you choose which technology investments to make?
Colella said he strives to find tools that are easy to use and serve almost as a kind of “staff augmentation.” “We don’t want to default to just hiring more and more people.”
Connors also encourages prospective software buyers to look at what training and certifications are available for them, and how many experts exist within the tech marketplace. She shared her own experiences with acquiring tools only to discover it took five to six months to find people who could develop with them.
Greenwood stressed the importance of transparency in toolsets. As a baseline, data tools should support features such as data lineage, proactive alerting, version control, audit logs, and others that grant visibility into the mobility of data.
“If you don’t have that transparency,” Greenwood said, “you’re already at a disadvantage.”
3. Start with the problem, not the solution
However, on the flip side of the coin, companies have a tendency to reduce governance and data democratization to a tools problem. That leads to making premature investments in technology that don’t solve the organization’s actual issues.
When asked for examples of companies that have “tamed” data democratization successfully, Greenwood shared that she hasn’t seen many. Too many orgs, she said, buy a tool, get it up and running, and leave it at that.
Greenwood said she encourages her own customers to step back and “understand the use cases” first before making a technology investment. This way, business units can verify that the tech in which they’re investing solves their actual problems.
Connors also stressed the importance of identifying the right problems. If your problem is that it’s too slow to access data, that’s a different problem than having conflicting metrics. Both are distinct issues with often very different solutions.
4. Get everyone aligned
One critical component of any successful data transformation project is alignment. Having everyone on the same page—from leadership on down—must come before more detailed decisions such as tooling.
Greenwood said the best way to go from a data anarchy to a more governed approach is to align goals to company objectives. She advocated for defining your organization’s use cases and determining how they drive your strategic objectives forward.
“If they don’t,” she said, “de-prioritize them.”
A data-driven prioritization approach can also help foster alignment. Without data to determine where you should make investments, said Colella, you’re left making decisions based on “who shouts the loudest.”
Top-down leadership may sound antithetical to data democratization. But Connors argued that it’s one of the best ways to avoid ending up with a dozen different implementations and focusing everyone on key objectives.
Objectives will drive the metrics your organization uses for measuring success of your self-service data democratization effort. For example, if the problem is that no one can find data, then the number of Daily Active Users or Monthly Active Users may serve as initial success metrics.
Conversely, if your goal is to reduce redundant effort in data quality and report generation, a working metric could be the amount of code obsoleted, or a reduction in the number of hours that non-data engineering personnel spend transforming data.
5. Accept that change is boring
A 2021 study by McKinsey found that 69 percent of all digital transformation projects fail. There are many reasons for that. But one reason is that, too often, teams try to do too much, too soon.
Our experts all agreed on not trying to bite too much off of the data transformation apple in one go.
“Identify what your biggest pain point is,” said Greenwood, “and solve for that. Don’t try to solve all of [your issues]. Solve something, get that one percent, and go from there.”
Connors also emphasized the importance of starting small and getting “one win” under your belt. “Even if it’s a small one, that one win gains a lot of traction in your org.” Start with one thing to fix—be it a faulty dataset, dashboard, or metric—to get the data transformation snowball rolling down the hill.
From data anarchy to a governed data democracy
So is data democratization possible—or is it a pipe dream? Can companies go from a data anarchy to a governed democracy?
Our panel thinks so—but it’s a marathon, not a sprint. Data democratization, Colella said, takes “time and resources.”
“A lot of people don’t realize the amount of work that goes into it,” said Connors. “The ones who get it right actually put in the work and follow the procedures.”
Done too quickly, data democratization may create more problems than it solves. By aligning on objectives, starting from one win, investing in the right tools and training, and slowly scaling out, you can reap its benefits while avoiding its potential perils.
Last modified on: Nov 22, 2023