How do you create a data visualization plan?

Ingredients (Data): Gather your data, like ingredients for cooking. Decide what information you want to show – numbers, categories, trends, etc. Recipe (Purpose): Know why you're making the visualization, just like you cook for a reason. Are you explaining a trend, comparing things, or telling a story? This helps you pick the right way to show the data. Flavors (Visuals): Choose the best way to present your data. It's like picking the right spices for your

Define the Problem Understand the goal and objectives you are trying to achieve. Anything from identifying the number of titles read in a certain period to visualizing number of users per region.

This is the perfect place to start, understand the insights the clients would be drawing from the visualizations. Remember this scope and try to build a consensus with your stakeholders to ensure that your understanding is correct. More often than not, this will keep shifting as time progresses (scope creep), but the essence of the insights should remain the same. Let me exemplify this, if you are looking at customer buying behavior for a particular marketing campaign, then you can be asked to bring in the customer loyalty data to add insights but that should not turn into a loyalty points earn-burn analysis. The scope should be limited to customer buying behavior. Otherwise, this may quickly turn into a wild goose chase.

The FACT framework serves as an integral roadmap guiding us through the intricate landscape of data analytics. The journey commences with "Framing the question", an essential step that requires both precision and simplicity. This well-defined problem statement not only streamlines our approach but also paves the way for subsequent phases, including data collection, processing/aggregation, and executive communication. Embarking on a data analytics venture without a concise and accurate problem definition is akin to blindly scouring the Pacific Ocean for a singular ring without any navigation aids – its endpoint remains uncertain and elusive.

Understand what the client business is and more specifically the view your audience will want on the area of the business they work in. Someone in sales and someone in inventory in the same company would want to know different pieces of same data sliced up differently.

UML diagrams, specifically a use case diagram, can really help in articulating what actions your users should have with your systems and data repositories. Building a use case diagram for them with their involvement will reveal the various data touch points and how those touch points influence their processes.

Understand the various stakeholders and their interests. Everyone might be looking at the same data and viz but drawing different kinds of insights. Next, categorize these users and keep the categories as small as possible. Understand the key aspects from these users and provide them with customized filters and even customized datasets. Let me give you an example- There is a sales dashboard that the leadership team is looking at. They would use this to provide incentives to their sales representatives. The views might look like this: 1. Leadership team - Overall achievement, payout curves and total payout 2. District Managers - Specific districts 3. Sales representatives - data for all the leads and their conversions

Understanding your audience, including their experience and background, is pivotal to engaging their attention in the narrative you're presenting. Please think of it as crafting a pitch specifically for potential investors. Certain audiences might focus solely on the end result, bypassing the underlying causes. In such instances, it's prudent to adopt a "prescriptive analytics" approach, emphasizing forward-looking insights. Conversely, other audiences seek a deeper comprehension of the factors leading to specific outcomes. For them, it's essential to integrate both "diagnostic and descriptive analytics", shedding light on the underlying processes and historical context.

80% of any project is data cleaning and preparation for every analysis that I have ever done. If anyone else says otherwise, either they have figured out a genius way to mind-read the variables in the dataset or you are talking to an AI (which is quite possible nowadays). Exploratory analysis as the name suggests is a getting into the dataset and explore every nook and cranny of the dataset and map out the trends within the data. This would provide an appropriate understanding of the relationships within data and provide a holistic view of the reporting as you would know what variables should be looked at and its conjunction with other variables.

In the FACT framework, the stages "Assembling the Data" and "Calculating the Outcomes/Results" stand out. The process starts with identifying suitable datasets for analysis. This data is processed using the ETL (Extract, Transform, Load) method. Extraction consolidates data from various sources into a central location. During transformation, initial tasks like de-duplication and outlier detection are performed. After cleansing, business rules enrich the dataset. This enriched data undergoes operations such as aggregation and filtering. Finally, exploratory data analysis reveals patterns within the dataset.

Gather the data Collect data you need to solve the problem. This may involve using APIs to access data or use data already available. Clean the Data Data cleaning involves removing unwanted data, dealing with missing values and transforming data into a format suitable for visualization.

The last stage in the FACT framework is "Telling others". Choosing the appropriate presentation style should also be strategically designed keeping the audience type, experience and background in mind. Every visualization should start with a executive summary that explains the outcome of the data analytics process along with explaining the emerging data patterns. Identifying the intent of the message that needs to be communicated will complement the process of establishing the visualization elements such as the type of charts, graphs, etc that should be used for communicating the message. Using colors, sizes to differentiate the outcomes such as "increase vs decrease", 'met vs missed" etc would help audience focus on attention.

Retroactive data analysis is a critical process in the business analytics world! The data patterns that once identified, may change in the future due to various business/economic situations and therefore a retroactive, periodical study is required to ensure that accuracy of the entire data analytics flow. Receiving feedbacks from the audience, is a gamechanger. If possible, prompt to receive audience feedback immediately after they have seen the visualization. This will help obtain accurate feedback vs receiving a feedback after a couple of days.

Sharing your visualization enables the circular loop of learning from you and your audience. It has been proven that sharing induce appreciation and possible improvements of your visualization

Training. Providing training to end-users is a good way for them to understand how to interact and interpret the visuals you created. As an additional step, an evaluation process of your dashboard is helpful to compare the original objectives set at the beginning and the overall outcomes to measure success.