Drafting data analysis guidelines
Where and how to start creating guidelines and a foundation for data analysis
Roles and Responsibilities
Goal and project summary
Working on the analysis platform for Getfeedback I faced the challenge of bringing the best options for customers to visualize their data for specific use cases of question types. Understanding what are the use cases for those. How can we start drafting a guideline for data analysis?
Context and background
Our tool offers customers the flexibility to ask multiple types of questions or gather various types of data in their surveys. However, when it comes to analysis, we need to have a clear understanding of the main use cases for each question or data type so that we can plot relevant data visualizations. Additionally, our tool supports multi-survey analysis, allowing users to combine data from various touchpoints and see how they relate to each other. Due to this, we do not permit data filtering at the tile level, only at the workspace level.
In the majority of our user research, users express a desire for additional types of visualizations. However, upon further investigation, we discovered that the root issue they are trying to address is related to data correlation, which is not yet available within our tool. Additionally, users often struggle with setting up their workspace to work effectively with filters, which is another area we are exploring ways to improve.
The CX market is rapidly evolving and new professionals are entering the field, often tasked with building a CX program from scratch. These professionals face the challenge of proving the value of their program with data and sharing insights with stakeholders across their organization. To address this challenge, I aim to provide easy, meaningful, and actionable insights to customers without overwhelming them with a plethora of data visualization options.
So, what problem are we trying to solve?
How can you define standards for that visualization and make sure that we are bringing easy, meaningful, and actionable insights for these users?
How can we don't overload users with useless visualization options and help them with guidance across the flow?
Definition of the process for Data Visualization in Workspaces
Understanding the context and the problem, I’ve come out with a process to define visualizations in steps:
In essence, for each question type, I aim to develop a workflow for tile creation on workspaces, constantly testing different options with customers to find the best approach. For instance, with text analysis, we found that a mind map was not useful for our customers, who instead preferred a list with percentages that still showed the relationship between the keywords and key phrases. We also implemented drill-down options to support their analysis. We have more examples to come.
By following our main proposal of providing easy, meaningful, and actionable insights through the right visualizations and clear data selection, we can meet the needs of our customers without overwhelming them with unnecessary data. Our approach enables them to drill down and see the relationships in the data for a complete analysis.
As mentioned before, we work with different personas, for different use cases on our tool ( such as specialists of CX, product professionals, and UX designers) but we need to keep in mind one main point:
Customers need the data to be right on the spot to be able to share across stakeholders. It has to tell a compelling and easy story for those who are not so involved in the CX program building;
Example of the process
Now that we have a clear process let’s go over an example of how we work on it:
Data Type: Meta Data
What type of data does the user collect and why? Operational system types, devices, and browsers are the main ones. With that, they can analyze the performance of the product in each type, not only in finding bugs, but features performance, and how each one impacts the overall scores (NPS, CSAT). Right now we can’t compare those results on tile, but how can we allow users to have that in another way?
Some constraints of the data and questions to be answered on analysis:
We have a lot of versions of each type of data. What is the level of the version that we need to show in the analysis? When it is relevant for these customers.
What type of analysis do we think would be easier for customers?
Change over time
Per previous analysis of usage of our tool and research with customers over time is a trend that 80% of customers look for it. The main challenges for this visualization are the number of versions and how to make it easier for users to visualize it.
To begin with, we need to determine the appropriate visualization options based on the data to be analyzed. Our research has shown that customers want to see the top items for each data type. Therefore, we should display the top items prominently in the visualization. If users want to explore further, they can filter the data based on their specific needs.
When users have a lot of variables it is meaningful for customers to understand what is the comparison between these items. The main challenge for this type is how to show this comparison and breakdown of versions on the same tile because for this type of viz it is expected that you can understand all the comparisons.
To specify visualization options, users should determine the item they want to analyze and how they want the breakdown to be presented, providing flexibility for comparison. One effective visualization option is to utilize expandable bar charts that allow users to easily view an overview and breakdown of the data at a glance.
This is the type of visualization that helps customers with easy insights: it will be straightforward and give a glance at the part of the whole of the data. For that reason, we are using donut charts and limiting the number of parts under 7, and grouping lower results.
To define visualization options for a specific type of data, we can display only the top results and group smaller results under an "others" category. This approach provides users with a clear understanding of the most significant data points while still allowing them to access additional information if necessary.
Results and Reflections
As we are still in implementation, the metrics for results are not available for this specific solution, but for the overall process, we can say that helps structure the work of bringing all the data and not overloading developers with building multiple visualizations that don’t reflect user needs. From the design perspective:
• This process makes the design process easier and more productive. It’s almost like a guide and checkpoints for a good experience in analysis;
• We keep consistency in the experience;
• We align better on requirements with product and engineering;
• Sets foundation for an analytics design system;