The Clearly Podcast

Andy: Gentlemen, it feels like we've been apart for weeks.

Shailan: It does.

Tom: Yes, it's been a couple of weeks since we last recorded.

Andy: I've missed it. Welcome back to our Power BI therapy.

Shailan: To our listeners, this is a weekly podcast.

Andy: I just want to say how much I've missed you two. Anyway, I got an email from Helen Wall, a LinkedIn Power BI instructor. She shared a job description for a senior Power BI developer in Houston, TX. One of the requirements was to convert all many-to-many relationships in the Power BI model to a more normalized structure.

Tom: That sounds terrifying. Enough to make a job out of it.

Andy: It was the third bullet point. Many people are still sold on Power BI as an end-user tool. I looked back at one of my first models and was horrified. You learn over time what not to do.

Shailan: Many-to-many relationships were introduced in one of the updates. If it hadn't been introduced, it wouldn't be an issue.

Tom: It's not necessarily wrong but shouldn't be commonplace.

Andy: There are times you need to use it to make something work, but don’t make it a habit.

Shailan: Often, a proof of concept (POC) becomes production, which leads to issues.

Andy: Exactly. Let's move on to today's topic: The Four Levels of Data Analytics—descriptive, diagnostic, predictive, and prescriptive. Often, the foundational elements like data quality, data architecture, and data governance are overlooked, but they're crucial.

Tom: Before diving in, let’s define the levels. Descriptive analytics answers, "What happened?" It's crucial because without knowing what happened, further analysis is pointless. Diagnostic analytics answers, "Why did it happen?" Predictive analytics answers, "What will happen?" And prescriptive analytics answers, "How can we make a certain outcome occur?"

Andy: Descriptive analytics is often called basic, but it's fundamental. Without accurate, transparent, timely, and automated data, you can't get the rest right.

Tom: Descriptive analytics needs to be accurate, timely, and transparent to build trust and ensure use. Automation is also key for timeliness.

Andy: BI and analytics are tough. There are essential practices that often get overlooked. For example, some clients want to reduce dependency on Excel in favor of Power BI for its structured, robust approach. But we need to think about the right hygiene factors.

Shailan: Data quality is crucial. Without good data, even AI and ML models can't deliver predictive analytics. People often try to mask issues instead of fixing them at the source, which is problematic.

Andy: Masking issues is common but problematic. You need to correct data at the source for reliable analytics.

Shailan: Culture plays a big role in data governance. People often don't want to show negative data and try to put a positive spin on it, leading to masked problems.

Andy: Some clients have good governance and architecture, but often there's a lack of recognition of what’s needed.

Tom: Most companies think they're further ahead than they are. Rarely do they have the ambition to go all the way to prescriptive analytics because they haven't thought about using data science to drive outcomes.

Shailan: From a C-level perspective, there are two extremes. Some think they have everything in place, while others realize their reports aren't good enough and seek strategic advice.

Andy: If you want to understand your data foundation, look at how long it takes to produce reports, whether they are automated, and how often data corrections are needed. Governance is crucial for long-term success.

Shailan: Ask why it takes so long to produce reports. It could indicate underlying data issues. Governance is often not explicitly mentioned but is essential for data quality.

Tom: Descriptive analytics often drives the need for better governance. Without accurate data, nothing else works.

Andy: It's a process, not an event. You need to make data quality an ongoing practice, not just a one-time fix.

Tom: Absolutely. Treat the different levels of BI as an incremental journey. Focus on getting descriptive analytics right to make the higher levels easier.

Andy: Gentlemen, thank you. And a special thanks to Jebs for producing and publishing our podcast. Next week, we’ll dive into data quality. Thanks for listening.

Shailan: Thank you.

Tom: Cheers.