top of page

Visual Data Analysis & Exploration




🥅 Know your goals


There are two general types of data visualization we’d like to distinguish, first exploration, which has the goal to use visualizations to understand your data and determine what questions you can answer with it. Second, visual data analysis whose goal it is to focus on comprehensively addressing a specific question, then to visually present your findings and tell the story around the data.


This can highlight why a certain outcome is happening or identify what is working and what isn’t. Sequence wise, you initially need to do a proper job in your data exploration before you can start working on an insightful analysis. That being said, visual data exploration and visual data analysis are closely linked and the line can blur during iterative processes.



🔦 Explore and understand

Visual exploration can be an extensive process and an important step to help you ensure data quality and detect feature correlation, anomalies and trends in your data. One way to achieve some of these aims is the pairgrid method: If you have a set of numerical data which you want to compare, it plots pairwise relationships and thus makes them more easily visible.


Later, during visual data analysis you select the specific findings that help you address key questions that you want to answer (to yourself or an audience). This way your previously gained insights enable a coherent storytelling. Here, choosing a fitting visualization style (e.g. charts, tables, dashboards) is crucial. For instance: You can use a time series plot to highlight a trend in your data, e.g. showing the performance of your marketing campaign over a specific period of time and visually pointing out the impact of specific events on that campaign.



🛠 Know your tools

It is important to select tools that work well for your data exploration efforts but also make it easy to share the results with stakeholders. For initial visual exploration we often use static graphics packages like seaborn (in Python) or ggplot2 (in R).


For presentations, the multi-language Plotly library makes interactive and visually aesthetic graphs and better engages the less technical people in the organization and can easily be integrated into slide decks. For ongoing reporting, tools like Holistics or Looker can enable end users to do sophisticated visual analyses with a high degree of customizability and ease of use.


Comentários


bottom of page