# Data Storytelling 101 | Think Like a Data Analyst

**Channel:** Christine Jiang
**Source:** https://www.youtube.com/watch?v=H79S8YDuYUU
**Transcript page:** https://www.withtranscript.ai/video/H79S8YDuYUU

## Chapters

- 0:00 — Why Data Storytelling Matters
- 0:56 — Defining Storytelling Through Examples
- 4:24 — Drilling Down to Find Insights
- 7:39 — Choosing Charts for Storytelling
- 9:26 — Best Practices and Common Mistakes

## Transcript

**[0:00]:** So, let's be real. Most data analysts know how to make charts, but not all data analysts know how to actually tell compelling stories with the data. And this is one of the skills that I see over and over again really helps my students stand out in the job hunt because they can impress hire managers. And in the age of AI, where AI can spit out a graph in like seconds, the analysts who are going to be able to communicate their insights across a variety of different audiences are always going to be the analysts who not only get the job the fastest, but stand out in their careers more. So, in today's video, I want to walk you through some game-changing steps to immediately elevate your data storytelling skills so that you can also make your portfolio stand out, feel more confident in interviews, and be able to impress

**[0:39]:** more hiring managers. If you're new here, I'm Christine. I'm a former data hiring manager and a data mentor who takes people on the most direct path to working in data. And in this channel, I share some of the insights, the strategies, and the frameworks to help my students stand out in the job hunt and on the actual job.

**[0:56]:** So let's make something clear first. Data storytelling is not just making charts and visualizations. That's data visualization which is the process of using tools like Excel, PowerBI, Looker, Python, Tableau to create graphs that actually surface the trends and the patterns in the data. But data storytelling takes this a step further. It actually interprets what is going on in the numbers and in the visualizations to bring these numbers to life and helps someone who is maybe not from a data background understand not only what's going on in the data but what they should do about it. So for example data visualization is something like here's a graph of sales trends for the last 6 months whereas data storytelling is saying something like sales have dropped by 30% in the last 6 months. This is actually primarily driven by the dip in Apple iPhone sales in the Asia region and we should look into this further to understand why.

**[1:47]:** So now let's do a quick demo of how to actually do a little bit of data storytelling. Let's say I'm starting out a portfolio project where I'm doing some exploratory analysis based on a sample data set from an e-commerce company. So let's say my data set looks something like this where I have things like the customer ID, the delivery timestamp, um whether or not or what kind of product was actually ordered, when the product was ordered, and of course most importantly how expensive the product was. Now, because it's a portfolio project where I don't actually have stakeholders to ask questions to, I'm going to use chatbt to help me pose some stakeholder relevant questions. So here are five stakeholder relevant questions you could explore in a portfolio project. Marketing performance. Which marketing channels drive the highest revenue and most valuable customers? Regional performance. Which regions generate the most sales and how does pricing vary by region?

**[2:36]:** What are purchasing behaviors of different customer segments? And how do sales trends fluctuate over time? And are there seasonal peaks and customer lifetime value? What is the expected revenue from a customer over time? Now, based on my experience working in data for the last 10 years, I will say that every company I have worked at cares about these kinds of questions. Now, I really like question number four. How do sales trends fluctuate over time? Because this is a question where we're taking a key northstar metric and we're looking at one of the main dimensions data analysts will work with, which is the element of time. So, let's actually take this question and start exploring in our data set and see what stories we can find. So, I'm using Tableau where I've actually loaded this data set into Tableau already. It's a free account made on Tableau public which has most the same functionality as the paid version of Tableau.

**[3:21]:** So, I highly recommend you start out here if you don't already know a visualization tool. So, I'm going to start exploring this question of how do sales trend over time. The first thing I want to do, of course, is actually pull in the purchase timestamp and then I'm going to pull in the US dollar price because that is going to show me the total number of sales. Now, this is flipped. So, I'm going to flip this so that I have the dollar price over time. And um because we want to look at trends, looking at trends on a yearly value is a little bit too aggregated. I actually want to roll this down to the monthly level. So, I'm going to break this out by month. And already I'm starting to see some interesting patterns here. I'm observing that for this e-commerce company there was a huge spike in sales in 2020.

**[4:03]:** And then since then there's been a pretty steady decline across the last well from 2020 to 2023. But there does seem to be some spikes in seasonality because there are peaks in values that I observe. But from 2022 to 2023 there was a pretty significant decline throughout that entire year. So that is already a key story that I'm telling there.

**[4:24]:** However, someone on the leadership team or my proposed stakeholder for this project would likely ask, well, can you help me understand why that is actually happening? And this is where we would start to drill down into different dimensions. So, let's actually dive into that. Let's say I take marketing channel and I split it by color. I'm seeing here that there's a huge decline mostly in direct, which is actually the number one channel that brings customers to our platform. Whereas all other channels seem to be pretty stable, but they are also lowest volume. I'm also going to look at the product names and see if there's one product that's really driving the change. Now, this is where it gets a little bit interesting. It looks like there were dips across all the different products. However, I am also observing here that the top three products, which is responsible for the most of the sales in the first place, also exhibit the biggest dips.

**[5:12]:** So, there's something to look into there. And you can see how with that simple exploration of just a few clicks in Tableau, I'm already starting to pull out an insightful story. But if I start to apply the same approach to a number of different metrics and a number of different dimensions in my data set, I will start to see the story line come through as I start to piece together the headlines of the overall sales trends. And then I'm also going to look at the region. So this is what I'm seeing now. What's the story here across all four regions? There is a dip. The dip is the most steep in North America and AMIA because those were the regions that had the most sales in the first place. However, because the dip is in all four regions, it doesn't really point to a key problem in one specific region.

**[5:53]:** But if we were to investigate where the dips were, it would probably first be in North America and AMIA. What you want to get at with data storytelling is giving someone a big picture overview of what's happening and then also giving them a more detailed understanding of why that's happening. And so that's what we were doing just then. I looked at the top level of sales trends. I saw that there was a big peak in 2020. I asked myself why and I started drilling down into the different dimensions and then reporting back up which of those different dimensions, which regions, which marketing platforms, which products were responsible for the dip or whether or not the dip was exhibited across all of those different dimensions. Then the stakeholders bring their business context to the picture and together as a team, you often decide what to do next.

**[6:36]:** Here are some standout examples of storytelling from some of my students. Here we're looking at sales trends over a few years and the so what is actually in the title where monthly sales revenue spiked in 2020 and has steadily declined since. The average line also helps us benchmark which months did well and which months did badly. Now this second example from one of my current students. She's working on this visualization where she's actually doing storytelling directly in the graph and she's talking about whether or not a newly launched loyalty program is working well. And this is what she's observing. At first, non-loy loyalty members spent about 30% more. This spiked to about 50% more on average. But in more recent years, this trend actually flipped. Loyalty program members started purchasing relatively more expensive products. So perhaps the program is actually doing better than earlier. And this might be something worthwhile continuing for a few months to see how this trend progresses.

**[7:27]:** The more you can do this, the more data storytelling becomes an intuition and an instinct so that when you get a new data set, you're already asking yourself the questions that we talked about.

**[7:39]:** So effective storytelling usually sits on top of effective visualization. Which is why picking the right visualization is actually a prerequisite to even mastering storytelling. You can think of storytelling as belonging to a few main categories. Value reporting trends over time looking at distribution or mix and making comparisons. Now let's break this down a little bit more. So value reporting is where you're just reporting what the actual values of a metric are. For example, what are the values for clickthrough rate? What are the values for signups? What are the values for sales? In this case, you want to use a table most of the time because this is actually what's reporting the physical values. For trends over time, this is where you're trying to understand how a metric has actually progressed throughout its history. For example, what are the overall trends in signups or what are the overall trends in sales or what are the overall trends in revenue?

**[8:25]:** And of course, you can split these into different dimensions. Most of the time, we're going to use something like a line graph here. You might have a singular line graph which just has one line or you might actually split the visualization into a series where you actually see the line graph split into the main dimensions for distribution or mix. Now this is a bit more of an advanced topic where you want to see how a key metric is actually split across a specific dimension. So for example, what is the distribution of signups over category? And this is where we would most likely use something like an area graph where we can actually see the split of that key metric across that dimension over time. And then for comparisons, a lot of times as data analysts, we want to understand how one metric compares to another or how one dimension compares to another.

**[9:09]:** So for example, how does clickthrough rate compare across regions or compare across different platforms? Most of the time here we would use a bar graph where we can actually see the values across the different categories and visually compare them and see them next to each other.

**[9:26]:** Yeah, flashy charts are really impressive looking, but practical visualizations usually are the most clear and clarity is what actually makes data storytelling compelling. So in order to really master data storytelling, we also should understand what are the industry best practices for dos and don'ts. So one of the first don'ts is to not use pie charts. I see a lot of pie charts in portfolio projects for some reason. And the thing is these are really not industry best practice. We almost never use them at work. The main reason is because pie charts don't actually show you that much. But usually stakeholders need to make decisions based off of trends over time and how things have changed most recently and what they look like they're going to be in the future. But a pie chart doesn't actually tell you any of that. It just squishes everything together into one value.

**[10:12]:** So for example here, if I looked at a pie chart of sales broken out by product on the data set that I was showing you guys before, I would see that the monitor, AirPods, and the MacBook make up about 20 to 30% of sales. But that doesn't really tell me that much. Was that in 2019? was that in 2022? How has that been trending recently? However, if I take a pie chart and I instead translate it to an area graph, then I see a slightly different story. I see that AirPods have consistently made up 25% of sales, while gaming monitors spiked during the pandemic and have normalized at 33% of sales. But there was a period of time at which the MacBook Air actually took up slightly more sales than the gaming monitor. If you think about it, an area graph is just a pie chart that is smooshed over time.

**[10:56]:** The second main don't is again to not prioritize cool over clarity. A lot of times on Tableau Public you will see these visualizations and dashboards that look really fancy. They have tons of cool colors and they also use really cool looking visualizations. But the reality is that these are not really practical for the workplace. Oftent times the fancier the visualization is the more obscure the real insight is. Now in terms of dues here are my two favorite tips. One of the easiest ways to bring storytelling to the surface is to actually relabel the titles of your charts and your graphs in your projects and presentations to instead of being a descriptive title, use a storydriven title that actually relays the main insight or the main so what of that graph. And the second tip is to use color for purpose, not decoration. There's an important principle here, which is that the more color you use, the more you actually neutralize the effect of how much that color actually stands out.

**[11:49]:** And so if you're minimal with your use of color, then when you do use color, it really helps something stand out. It is very important to be intentional about using the right tool. If you use Excel or if you use Python just for data visualization, this is usually going to make your life a little bit harder because the visualization defaults in both of these tools are often quite clunky. And so you want to use a visualization tool for effective visualization like Tableau, PowerBI or Looker because these were built to have aesthetic looking defaults and also to be interactive for stakeholders. And also do prioritize simplicity. Just remember the quote perfection is not achieved when there's nothing left to add but when there is nothing left to take away. So you can think of something called the data to ink ratio here. Think about all the grid lines, the tick marks, the titles, the axes labels, and ask yourself, is this really necessary to actually bring the data to the surface?

**[12:38]:** If you want to take your data storytelling skills to the next level with me in a real world setting, check out my mentorship program below. We're going through the applications now and sending out invites for info sessions in the next two weeks. So, I would love to meet you guys there. And if you find this video helpful, don't forget to subscribe. And I will see you guys in the next