10 Power BI mistakes to avoid
Power BI is increasingly bringing the power of data to business users, but you may need more guardrails to get the most value from it. These are the most common things organizations get wrong.
As a leading business intelligence tool, Power BI offers business users power and flexibility in dealing with data. The Microsoft tool provides everything from Excel integration to enterprise reporting and an increasing number of AI features that simplify getting deeper insights. In fact, the latest Forrester Wave report on augmented BI goes as far as to say “it is hard not to consider Power BI as your top choice for an enterprise BI platform.”
But with so much power and so many features, you need to understand how to work with Power BI to get the most out of it. These are some common mistakes enterprises make with Power BI and how to avoid them.
Going wild with Power BI
Much like Excel, the power of Power BI gives it grass roots popularity. There’s probably already more usage in your organization than you know about. Just leaving people to find out how to get the best from Power BI on their own means they may not make the most of it. Even worse, you can end up with so much data that’s been uploaded (and duplicated by different people) and so many data artefacts that have been created that no one knows who’s doing what and where the useful reports are.
Instead, curate and endorse trusted data sets to make it easy for people to find quality data. Provide training and consider building a center of excellence that includes BI architects, data experts, and people who can help business users be effective with Power BI. You may even want to restrict them from publishing data until they’ve had some training to help get the value without the chaos.
Forgetting data security
The most valuable data in your organization might be confidential, and its usage likely needs to be audited. Turn on integration with Microsoft Information Protection so that you can manage and track data usage and control how it’s used even if it gets exported from Power BI to Excel.
Being too restrictive
The real key to successful Power BI adoption is getting the right balance between empowering users and effective governance. One sure way to fail is to lock down usage to only a few business analysts or data scientists who will look at the data and issue official guidance. You probably can’t hire enough of those experts, and far more people in your organization likely have the data literacy to find valuable insights with Power BI that allow them to react to the changing environment based on data rather than hunches.
Don’t try to control Power BI usage so tightly that you get in their way, but do help them get the best experience with training and good examples.
Ignoring performance
When you think about the value of business intelligence it’s easy to focus on the intelligence — the insights you can get from quality data with the right questions — and the business — the fact that executives will use the tools and act on the insights. But you also need to make sure the basics are right: Unless reports load quickly and interactive visuals are fast, people just won’t use it.
According to Microsoft, successful Power BI adoptions correlate very strongly with good report performance — and not paying attention to performance is the top reason business intelligence projects fail.
Use the Query Diagnostics and Performance Analyzer tools in Power BI Desktop to understand and improve how queries and reports run (SQL Server Profiler can also be useful for finding out which queries cause bottlenecks). You can use Azure Log Analytics with Power BI Premium to dig into your telemetry. Start with the built-in Usage Metrics report to see what reports are popular and make sure they’re well optimized. For large organizations, Microsoft will warn you (in an anonymized report) if Power BI is getting low satisfaction scores from users, suggest which reports are causing problems and, for the largest customers, provide free consultation to help you optimize those reports.
Over-relying on visuals
A picture is worth a thousand words, as the cliché goes, and what Power BI does well is show you a picture that reveals what’s going on inside your data. But not every picture is a useful picture; too many charts and visualizations can produce visual clutter where you don’t notice the most important measures. They also generate a lot of queries that can slow down report performance. Find the right visuals for each report that focus on what’s really important, or use AI visuals such as Key Influencers that try to automatically identify the most significant data.
Failing to make the most of mobile
Not everyone who will find Power BI useful sits at a desk and avoiding overcrowded reports is particularly important when the insights help frontline workers. The Power BI Mobile app is helpful here, as are embedding reports into Power Apps or setting up OKR boards and scorecards with notifications for specific goals that employees will see wherever they are.
In the future, the mobile app will use spatial anchors to bring data and reports to the physical space where the data is actually generated, with an AR view that pins reports to physical objects. That could involve seeing usage state for exercise equipment in a gym, sales figures for the products on a supermarket shelf, throughput and other statistics for machinery on a factory floor, or cars at a dealership: anything where having the right data immediately visible could help you make the best decision.
Overlooking DAX — or its performance
Data Analysis Expressions (DAX) are powerful functions that enable you to use complex filters, conditional logic, and aggregations for more advanced analysis. Poorly written DAX functions can also slow down reports and use up capacity. Use the Power BI Performance Analyzer to find DAX measures that need improvement, and make sure you have resources and training to help users understand how to use DAX well.
Assuming your back-end system can handle DirectQuery
Not having to replicate data into Power BI to work with it can be very convenient, but you need to verify that your back-end system can handle the load. Microsoft suggests that the response needs to be less than 5 seconds for a typical aggregate query, but remember to multiply that by the number of visuals in reports. If there are 10 or 20 visuals, loading that report will generate 10 or 20 queries, which means users will be waiting for one or two minutes just for the report to load.
Not using incremental updates and aggregations
Loading data into Power BI to take advantage of the high-performance query engine is very straightforward, with connection packs from a wide range of data sources and a gateway for on-premises data. It’s so easy that users will frequently reload all the data every day to make sure they’re up to date — but there’s no need to reload old data that hasn’t changed.
You can speed up refreshes by using incremental updates that only load the data that’s changed; this will also reduce load on the back-end system. For DirectQuery, enable aggregations that automatically preload and aggregate data in a cache to speed up reports and reduce the load on the external system you’re querying.
Not building a data culture
The most successful Power BI adoptions bring data tools to a wide range of employees, not just people whose job is working with data. But that means making it second nature for your business users to rely on data as a way of making decisions.
For this to happen, you have to develop a data culture where, CTO of Microsoft Analytics Amir Netz suggests, “every conversation starts with what does the data say, every suggestion includes what does the data support. Move from what you think and who you know to what you know and what the facts are saying and it makes a huge difference. A data culture moves from operating on hunches and who speaks louder or who has more power to being an objective organization that is constantly looking at and measuring itself, and constantly optimising to get better results.”
Courtesy of: Mary Branscombe
