1. AI visualizations in Power BI
Welcome back! In this screencast, we will explore revenue with some of Power BI’s AI-driven visuals.
First, we’ll start with a forecast for the next six months of revenue. To do this, we must use a line chart, and the dates must be on the x-axis, or it will not work. So we will add a line chart visual and select Order Date and Revenue from the Sales_Fact table.
Next, we move to the analytics pane and flip the Forecast to “On”. Now we can edit some of the options here.
Notice that it has created a forecast for the next ten years. This is because the units are set to points, which are the points on the line chart. If we drill down into the line chart, the forecast length will also change.
We want this to stay at six months no matter what, so we will select months as the units, then six as the forecast length.
Power BI will automatically look for seasonality in our data; however, we can set this manually. Typing a number under Seasonality will tell Power BI how long the seasonal cycle is. By typing 12, we are saying the cycle is 12 months long. It’s usually best not to type anything here unless we’re confident with the dataset’s seasonality.
Finally, we can adjust the confidence interval. A confidence interval is basically a range in which the expected values should land. So a 95% confidence interval says, “95% of expected values will fall within this range”. Likewise, a 99% confidence interval contains 99% of all expected values. Notice that as we increase the confidence interval, the range gets wider.
That was easy! In less than two minutes, we’ve got a sophisticated forecast.
Another AI visual is the decomposition tree. This is great for splitting up data into different verticals to understand the root causes and relationships between data.
We’ll start by analyzing revenue, and we want to explain it by Product_ID and Customer_ID.
Now that we have the fields selected, we can split up revenue within the visual. We’ll start by adding an AI Split, distinguished by the lightbulb. This is a split where the AI can find the next highest or lowest value in our exploration.
We don’t have to choose an AI split, though. We can just simply choose the next field we want to see, so we will select Customer_ID. Nice!
As we click on different nodes, the visual will split the data accordingly. How cool!
Finally, the last visual is probably one of the coolest AI features in Power BI. The Q and A Visual automatically looks through all the data in the model and will answer any question to the best of its ability. For example, we could ask, “What is the profit margin by country” and it will show me.
Let’s say that in our company, we are more likely to say “place” instead of “country”. If we type this in, Q and A will not understand.
We can teach the visual to understand this term by adding synonyms in the Q and A setup. The easiest way is to go to Teach Q and A and define the relationship here. It will highlight the term it doesn’t understand and ask me to define it. So we will type “country” and then save.
This saves the new term in our field synonyms list, and we can find all of our created terms in the manage terms section.
Now Q and A knows that “place” and “country” are the same, and it can answer our question.
Awesome! Now that you know how to harness the power of AI, it’s your turn to practice.
2. Let's practice!