Knowing is half the battle. This one phrase is the driving force behind every decision made, every action taken and every idea envisioned.But what does this have to do with dashboards? Well, the biggest reason why building dashboards or a data visualization platform is so important is that they provide a driving indicator to what we have accomplished so far and what should we plan for in the future.A good data visual model is once that can answer several questions with just a glance. It makes it easier for stakeholders to quantify metrics and evaluate key metrics that can be used to drive decisions. From a programming perspective, data visualization provides a solid platform to kick start your Exploratory Data Analysis (EDA) in such a way that you will be able to pick and choose which factors in your dataset make you go “Yes! This needs to be a part of my data!”
So how do I visualize?
The first thing that you would do when dealing with a visualization model is to ask yourself, “Does the data that I currently have do justice to my model?” Most datasets require you to clean before you can use them. Each dataset will have a different type and level of detail that you need to come across before starting with the visualization. Some of the biggest factors that are commonly found in datasets and how to handle them are as follows:
- Missing values: Depending on the number of values missing, you can either impute (i.e. fill in a value such as the average) or delete the row or column entirely
- Outliers: Although it is common to remove them, it shouldn’t be recklessly deleted as there might be an interesting story behind this point (E.g. A company paying exorbitantly high premiums for insurance while everyone else is opting for the cheapest)
- Granularity: Sometimes we would like to group values into distinct categories (this is known as binning) or break them down into a more granular analysis (E.g. Breaking down a timestamp to bin them as early or late arrivals)
Depending on the programming language or application that you are working on, this will eat away the most of your time. Once we are satisfied with the data that we have, we get to the fun part.
Show and Tell
Now that the stage has been set, it is time to dive deeper and visually deliver the answers to your questions. Which brings us to a series of questions: What are the most meaningful questions that should be answered? What the most relevant questions that should be addressed? How does anyone looking at my visual models understand what is going on?These are important questions you need to ask yourself before visualizing. Depending on your audience and their requirements, it stands to reason that different demographics will require different metrics to look into.Business stakeholders are more likely to require comparison plots between competitors and granular metrics that explain hiccups in their workflow. The public audience will want to see how an organization is performing over time and what are the date’s most eye-catching features.Since the workflow of a model depends on the audience and the dataset involved, this visual dashboard will differ. Some of the more common takeaways that can be applied to most datasets are followed:
- When comparing two or more quantitative features, it is easier to list as a bar chart.
- A trend line will come in handy when showing growth over time
- Less is more. Keep things simple
- Use the right chart to tell the right story
- Remember to use headers wisely
- Do not overpopulate the visual models with excess fonts or distracting elements
- Try as much as possible to maintain a standard set of colour scheme to denote different items
- Maintain evenness and arrange data to look visually ordered when applicable.
These are a few tips that would go a long way in building amazing models. There will always be more ideas to be added. With constant practice and domain knowledge, I believe that you can also build these ideas to convey a great story through your visual presentations.~ Shawn Rahul D'Souza