From Numbers to Charts in Minutes
1. From Numbers to Charts in Minutes
Uncovering patterns and insights during the analysis phase is often made easier with data visualizations, like charts or dashboards. In this video, we'll use AI to help us choose and design visualizations for effectively exploring and understanding data.2. Example: Reducing Operational Costs
Imagine we're advising a mid-size European manufacturer. Operational costs are too high, and the data clearly points to headcount reductions, vendor renegotiations, and the closure of underused offices. Nevertheless, such significant decisions must be firmly supported by the data, and we need an effective way to present those results.3. Example: Reducing Operational Costs
More concretely, our goal is to use AI to help design a dashboard that clearly shows the current cost structure, the savings levers and their impact, and the post-restructuring steady-state costs and headcount.4. Example: Reducing Operational Costs
Before we start, let's take a look at the data we'll be working with. First, we have a table showing the cost breakdown by business unit and category, with annual run-rate values in millions of euros. We also have a second table detailing restructuring initiatives, including annual savings and other key indicators.5. Generating Charts
AI systems can help us design dashboards by proposing the most suitable charts to communicate our goals. Remember that when crafting the prompt, it's important to upload the required data and provide any relevant context about it. In our case, the model suggests a stacked bar chart to show current costs, a waterfall chart to illustrate annual savings by initiative, and a grouped bar chart to compare steady-state costs and headcount before and after the restructuring.6. Generating Charts
Next, we can ask the AI system to help us generate the proposed charts. Here we have two options: we can either ask the model to generate all the charts at once, or7. Generating Charts
we can request them one by one. A good practice is to generate one chart at a time, as this makes it easier to apply follow-up modifications.8. Generating Charts
Let's ask the model to generate the first stacked bar chart! Normally, models write code to generate the charts. It's not important to understand this code, as you'll be reviewing the charts visually, and providing any modifications in natural language for the model to make changes. In just a few seconds, the model generates our stacked bar chart!9. Generating Charts
Now onto the waterfall chart. The chart looks good, but there's a catch. Our consultancy policy is to only use waterfall plots when initiatives have both positive and negative impacts, so we can clearly see how each one contributes to the total savings. When all values are positive, we prefer a classic pie chart.10. Refining Charts
In this case, we ask the model to turn the waterfall into a pie chart instead.11. Refining Charts
Now, let's say we want each initiative to have its own color, instead of coloring based on risk. We want to use the palette that best matches our consultancy's branding. To do so, we can provide the model with the exact color codes we want in the chart.12. Assembling the Dashboard
Finally, once we have all the necessary charts, we can bring them together into a single dashboard. Some AI systems can also assist in this aggregation, suggesting titles or descriptions for each visual in the dashboard.13. Let's practice!
Now it's your turn to derive insights through data visualizations.Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.