Having created the Cake Works dataset and commenting on its capabilities in answering the company’s business questions, we are finally ready to jump into Tableau and see how the overview dashboard looks like. Let the data storytelling begin!

Before we begin, it is important to reiterate a point that was made towards the end of the previous article. That is, the Cake Works dataset is entirely random. This matters because randomized data may not look sensible or insightful when visualized. To avoid this issue, it would be necessary to modify the Cake Works dataset (or make an entirely new version) that closely resembles a representative dataset on baked goods. This could be done by scraping data from bakery websites and consolidating that information into a dataset. That way, the data is both realistic and historical, which is more likely to translate to sensible visualizations when such data is brought into Tableau.

Here’s what the overview dashboard looks like. There are three distinct sections. First, there is the title and the filter buttons (images of different types of baked goods). Second, the dynamic KPI section (the KPIs reflect your selection when you click on a filter button). Third, four different charts that answer some of the basic business questions. These charts also change based on what filter button is clicked. Additionally, the chart about Regions can also be interacted with. That is, when you click on a bar of a specific Region, the remaining charts and KPIs will also reflect that selection.

At its core, the dashboard is summative and meant to give a high-level overview of how the sales of different product categories (baked goods) perform across different regions, cities and periods of time (month-by-month). Interactive and explorative elements were intentionally designed to be as simple as possible so that end users could avoid getting lost in any rabbit holes.

This dashboard answers a handful of high-level questions, such as what were the average profits for each of the baked goods and how those profits were broken down based on region and city. We can also see, from a month-by-month level, how the amount of orders and average price of orders changed over time. These graphs can help stakeholders identify, from an “overall perspective”, which regions and cities may need further support in order to drive up sales. In theory, it could also be possible to see if there are any seasonal patterns, which may inform inventory management decisions. For example, maybe pies tend to sell better during November and December than cakes. If so, then it would be a good idea for stores to shift production away from cakes and towards pies instead. If the data was more comprehensive, it could eventually be possible to create visualizations that may better serve to answer questions involving store expansion. For example, foot traffic-related data could be used to make informed inferences regarding whether or not it could make sense to develop a new storefront in a particular area, given that there are more people passing through that area.

With all of that said, let’s take a look at an experimental dashboard that I created that was meant to give a more detailed look at how specific items, such as Rainbow Cakes or Chocolate Chip Cookies, performed.

In the image below, you’ll see the current version of the experimental dashboard. I ultimately decided not to include it as part of the “final product” because I ran into some issues with consolidating the two dashboards. More than that, the dashboard is quite busy (this is especially true if you set the “Location” option to be “City” instead of “Region” – you’d see twice the amount of scatter plots). As it currently stands, it’s difficult to immediately glean impactful insights – it’s more of a tool to help draw out further investigative questions than a tool that can answer “what now?” or “so what?” kinds of questions. Besides this, I realized that the structure and level-of-detail for the Cake Works dataset didn’t sufficiently convey information about different storefronts, so it would be difficult (if not impossible) to meaningfully compare how different storefronts performed, which was the original intent of this dashboard. Instead, this dashboard looks at differences between individual items.

Whatever the shortcomings of this dashboard, it could eventually be further developed to help draw out more pinpoint analysis. That is, this dashboard allows a stakeholder to look at different location levels (Region, State, City) and different characteristics of every sold item (Average Customer Rating, Average Calorie Count, Average Price). With more comprehensive data, it would be possible to include more metrics to compare across the different items. With this dashboard, we could easily identify how items performed and even pick out outliers (items that were either recorded incorrectly or performed exceptionally poor or well). This dashboard could help stakeholders redesign product lines across different cities. Perhaps the success from one city could be replicated to another city which was previously underperforming.

Ultimately, besides the downside of having completely random data, the Cake Works dataset could be improved by adding more relevant data so that there is more opportunity for insights to be discovered. With that added data, data exploration would enable more probing questions to be asked so that insights could be generated, potentially leading to more impactful business decisions for Cake Works.

If you’d like to dive into the current version of the overview dashboard, please check it out here! I hope to rethink what the other dashboard could be and then properly integrate it with the overview dashboard so as to better meet the original intent of this project.