Latest / MBA Training Data / Self-service analytics at scale: what CDOs keep getting wrong
Transcript
- 0:00Welcome to the Leaders Insights podcast. Today's episode, self-service analytics at scale,
- 0:10what CDOs keep getting wrong. Most companies bought the self-service analytics dream five
- 0:15or six years ago. In 2026, the majority are still watching 80% of their Tableau or Power BI licenses go
- 0:21unused. What went wrong?
- 0:24They confused tool deployment with capability building. A CDO buys Tableau, sends an email
- 0:29saying go explore your data, and calls it a strategy. What they actually created is an
- 0:35expensive shelf product. The organizations that made self-service work, I think of Heineken's
- 0:41data division, and what Marsk did with their logistics intelligence program, they treated
- 0:46it as a change management problem first, a technology problem second.
- 0:51But change management is the answer everyone gives. What does it actually look like in practice
- 0:56at scale?
- 0:57It means you have to embed what I call data translators inside each business unit, people
- 1:02who are fluent enough in both the domain and the data to sit next to the finance director
- 1:08or the supply chain VP and say, here's the question you're actually asking, and here's
- 1:13how to build it. Without that human bridge, you get two failure modes. Analysts who build
- 1:19beautiful dashboards nobody uses, or business users who build dashboards that answer the
- 1:24wrong question with bad data.
- 1:27Let's talk about data quality, because that's where I see self-service collapse most visibly.
- 1:32Business users pull a report, get three different numbers for the same metric, and they stop
- 1:37trusting the tool entirely.
- 1:39That's the trust cliff. And once you fall off it, recovery is brutal. Gartner, and
- 1:44I'll flag this as their research, estimated that organizations lose an average of around
- 1:49$15 million annually from poor data quality decisions.
- 1:53The root cause in self-service environments is almost always the absence of a certified
- 1:58metrics layer. Companies like Spotify built their own internal semantic layer years ago.
- 2:05Today you have tools like DBT semantic layer or at scale that can impose that discipline,
- 2:11but most CDOs haven't made it a non-negotiable architectural requirement. They made it optional,
- 2:17so nobody did it.
- 2:18This is the other flashpoint. Too little and you get data anarchy. Too much and you've
- 2:23just recreated the IT bottleneck you were trying to eliminate. How do you calibrate
- 2:28that?
- 2:29You govern the definitions, not the exploration. Lock down what revenue means, what active
- 2:34customer means, who can certify a dataset as production grade. Then give people genuine
- 2:40freedom within those guardrails. What I've seen fail repeatedly is CDOs who govern access.
- 2:46Who can see what? Who can publish what? Without ever governing the semantic layer, so you
- 2:51have perfectly secured garbage. The permission model is airtight. The numbers are wrong.
- 2:57If a CDO is listening to this and they have one thing to fix this quarter, what is it?
- 3:02Audit your top 20 most used dashboards and find out how many of them share a single
- 3:07certified definition of your core business metrics. If the answer is fewer than half,
- 3:13it's your problem. And that's where you start.
- 3:16To go further, Leaders Insights offers strategic training programs, diagnostic tools, and an
- 3:22executive self-assessment designed for CMOs, CDOs, CFOs, and leaders aiming for board-level
- 3:29performance. Explore everything at MBA-training.com. Thank you for listening to the Leaders Insights
- 3:35podcast. See you tomorrow for another episode.