Module 2: Visualize results

In a previous video, you were introduced to the idea of scope as it relates to dashboard design. You may have also encountered the word “scope” in terms of project scope. In the business intelligence world, you might find the word scope being used in a variety of contexts. And, as you’ll recall from earlier discussions of context, understanding these contexts is key. In this reading, you’ll get a side-by-side comparison of project scope and dashboard scope and at what stage in a project you will likely encounter these terms. This will help as you encounter scope in different contexts as a BI professional so you know what the expectations are in every situation.

Key takeaways

Often, as a BI professional, you will encounter language that means different things in different contexts. By paying close attention, asking questions, and thinking critically, you can ensure that you and your team stay on the same page. In this case, the difference between project scope and dashboard scope is useful to understand as you communicate with stakeholders about their expectations with the dashboard specifically, and not the entire project.

In this course, you learned about the dashboard-building process. You also became familiar with many of the visualization strategies that BI professionals use to answer stakeholder questions. But how does this process play out in the industry? In this case study, you’ll learn how talent solutions firm Allegis Group, headquartered in Hanover, Maryland, used a visualization to monitor the productivity and effectiveness of their new recruiters.

Company background

Allegis Group is a talent solutions firm comprising several specialized companies that solve business problems for clients across a variety of industries. The firm does this by identifying great candidates for its client companies. Allegis Group recruiters identify talent, and then connect them with career opportunities at client organizations. Each time a candidate is placed, that’s considered a sale for Allegis Group. 

To successfully place candidates in client organizations, Allegis Group needs lots of internal BI and data analytics and a good understanding of its clients’ data needs. Quick and effective BI insights are essential to Allegis Group’s daily operations. The following case study explores how a BI dashboard helped Allegis Group leaders solve a sales tracking problem with a clever dashboard. This end-to-end project is an example of the kind of work you might do as a BI professional for organizations like Allegis Group.

Scenario

The talent recruitment process requires oversight and progress reports. Therefore, Allegis Group conducts its own data analytics and BI research to track the performance of its new recruiters. To challenge assumptions about how long it takes for new hires to meet sales goals, Allegis Group leaders wanted data insights into how each cohort of new recruiters performed within the first year. The goal was to create a dashboard that tracks the sales revenue made by the newly onboarded recruiters. With this tool, Allegis Group would be able to measure the new hires’ ability to bring quality talent to clients. Then, these insights would be communicated to internal company leaders in order to improve processes moving forward.

The business questions

Through internal brainstorming, Allegis Group leaders confirmed that they needed answers to the following questions:

• How are new recruiters performing over time?
• What are the best months of the year to hire new recruiters?
• Are performance goals for new hires correctly aligned?
• Are new recruiters now finding success more quickly than those hired a year ago?

To answer these questions, Allegis Group’s BI team collected data on how many months recruiters had been with the company. They also collected data on the sales revenue generated by recruiters within their first year. The recruiters were divided into cohorts depending on when they were brought onto Allegis Group’s roster. Then, each cohort could be measured by how much revenue they made based on how long they had been with Allegis Group. This would enable Allegis Group to review and evaluate the success of each cohort over the first year of employment. The dashboard would help decision-makers learn about retention rates, the ideal hiring months, and how to improve overall hiring practices.

The solution

With these questions in mind and the data gathered, the team began creating the dashboard. They first visualized the main metric, which was the sales (in dollars) per number of individuals in the initial recruiter cohort population. This metric represented the total sales for a month divided by the number of recruiters that were initially hired in that cohort. Then, they defined the data source and created a data view. Finally, they decided on a monthly schedule to refresh and update the data with the cohort’s monthly progress report.

After the team aligned on these elements, they began creating mockups of a visualization. They modeled the visualization after an existing project template, which used a chart with a triangle arrangement of boxes to represent data for each cohort at a given time. This arrangement represented data within a cohort for each row and within a relative month for each column. The BI team knew that this visual design best suited their cohort analysis needs, so it was a simple decision to build their own visualization based on the template.

The Allegis Group used specific charts to best convey the data. The Financial Times Visual Vocabulary poster includes a guide for choosing chart types based on the relationship you’re trying to visualize; there is also a plain-text version on GitHub.

Then, the team loaded the data into the template and created other charts to answer the business questions. Internally, the team tested the dashboard to ensure its interactive features—such as comparing cohort retention and performance—were working properly. Finally, the team released the dashboard to production. They still had a development version to iterate on, but now they had a fully functional version of the dashboard.

In the BI world, the timeline for completing projects can vary depending on the complexity of the data and the skillset of the team. For this project, the period of time between need-finding and production was about a month.

The results

The resulting dashboard had several strengths:

• The flexibility to answer each of the team’s questions in one place
• A monthly refresh for continued value
• Quick visual comparison within a cohort (rows) and between different cohorts at the same point in time (columns)
• Interactivity for quick comparisons of cohort retention and performance
• Ease of adding new metrics*

*The team began with one metric (sales per number of individuals in the initial recruiter population) but later added percentage of cohort meeting goal, percentage of cohort remaining, average number of starts, and more.

The main chart of the dashboard visualized which cohorts were high-performing and which needed more support. It was easy to find high-performing cohorts based on color. A darker blue indicated a higher number of hires per number of recruiters in the cohort. The triangle arrangement meant that each cohort could be measured horizontally (within cohort), vertically (by time elapsed since hired to Allegis Group), and by color (average starts per recruiter). This gave the long-term year view that they needed for each cohort’s progress, as well as the comparison among cohorts.

Allegis Group leaders use this dashboard to understand the progression of a typical cohort. They can also find out if some months are better to hire Allegis Group recruiters than other months. For example, if recruiters hired in the middle of the year historically perform better compared to cohorts hired toward the end of the year, management can plan accordingly.

Lessons learned

As with any complex project, Allegis Group’s BI team learned several lessons about what worked well and what could have been improved in the dashboard creation process. One strength they found was in their presentation: The team went through each visualization of the dashboard with the recruitment team that was going to use it. They were able to communicate initial trends and insights, as well as how company leaders could best use the tool. This collaboration was instrumental in driving adoption of the dashboard.

A difficulty they encountered was that it was challenging to clearly convey the complexity of their metrics to the audience. Their main metric (sales per number of individuals remaining in the cohort for a given month) was misleading without context. While this initially seems like a logical metric to use, it has a survivorship bias that skews the data.

A survivorship bias focuses on only those data points that make it past a selection process and ignores those that didn’t. In this case, the metric ignores the recruiters who didn’t stay with the company. Because it counts only recruiters who stayed with the company, it skews toward higher-performing recruiters. The Allegis Group BI team introduced a second metric (sales per number of individuals in the initial recruiter cohort population) to address this nuance and made clear the distinction between the two metrics.

Another lesson they learned was to embed a how-to video into the dashboard. This could have given the recruitment team a quick run-through of the dashboard’s tools and capabilities, enabling the BI team to focus on effectively communicating insights.

Conclusion

This example demonstrates how a BI team solved a problem by picking the most appropriate visualization. Choosing a highly specific template enabled the team to visualize the most important data in one chart, rather than in several less effective charts.

This team’s success also illustrates the value of maximizing pre-existing examples in order to create new solutions. Resourcefulness is an important BI skill. Often, the inspiration for a visualization comes from examples from other analyses. Sometimes, a BI professional might have trouble deciding which type of chart to select, so they’ll use a reference sheet, such as The Financial Times Visual Vocabulary poster. It isn’t plagiarism to seek inspiration from other sources, so long as those sources are not proprietary or trademarked.

Conducting research on visualizations is an essential part of being a proactive BI professional. Seeking out community knowledge and continually learning about new BI visualization best practices will help you stay current on evolving trends while building a library of reference materials to guide you throughout your career.

If you completed the Google Data Analytics Certificate, then you have already spent a lot of time considering how to create data visualizations. Many of the same principles will apply as you begin to create visualizations and dashboards for business intelligence work. That’s why, in this reading, you’re going to get a checklist of tips and tricks that you can use to guide your design process.

• Use a visualization framework: Frameworks like the McCandless Method and Kaiser Fung’s Junk Charts Trifecta Checkup can help you organize your thoughts about data visualization and give you a useful checklist to reference.
• Choose the right chart: Part of creating effective charts is choosing which type of data visualization works best for your needs.
• Organize your process with design thinking: Design thinking breaks down the design process into five stages: empathize, define, ideate, prototype, and test.
• Consider pre-attentive attributes: Pre-attentive attributes like marks and channels are the elements of a data visualization that people recognize automatically without conscious effort.
• Avoid misleading or deceptive charts: It’s important that the visualizations you create are communicating your data accurately and truthfully.
• Prioritize accessibility: Make your visualizations accessible and useful to everyone in your audience by using labeling, text alternatives, text-based formats, and distinguishing and simplifying elements.
• Apply design principles: There are nine principles of design that you should consider when designing your visualizations: balance, emphasis, movement, pattern, repetition, proportion, rhythm, variety, and unity.

Resources

If you’re interested in reviewing any of these concepts, you can check out these resources from the Google Data Analytics Certificate and more:

• This reading about effective data visualizations covers some basics about creating visualizations.
• This guide to selecting the right visualization has a detailed encyclopedia of chart types and when to use them.
• The Financial Times Visual Vocabulary poster includes a guide for choosing chart types based on the relationship you’re trying to visualize; there is also a plain-text version on GitHub.
• This reading about design thinking breaks down each step of the design process in more detail.

In a previous video, you learned about dimensions and measures in data visualization. As a refresher, dimensions are qualitative data types that can be used to categorize data. Measures are quantitative data types that can be either discrete or continuous, and encoding is the act of translating dimensions and measures into visualizations. In this reading, you’re going to learn more about dimensions and measures, different encoding techniques, and a guide to choosing the right chart type for your data.

As a BI professional, you will want to create visualization tools that are self-explanatory so that stakeholders can use them to answer their own questions instead of depending on you—which is why thinking about how you are using these elements is so important.

Dimensions and measures

Dimensions are inherently qualitative data—this means that they are subjective and explanatory measures of a quality or characteristic. Basically, this is data that records observations about the quality of the data. For example:

• Customer names
• Product names
• Geographic locations 
• Observations
• Interviews
• Reviews

These examples are descriptive; they indicate characteristics of the data that aren’t necessarily represented by numerical data.

Measures, on the other hand, are quantitative. Measures are what you will use to actually count the data and track changes over time. This data can be discrete or continuous—basically, this means they can be represented by numbers with limited or unlimited values. For example:

• Temperature
• Revenue
• Distance
• Weight
• Time

The kind of data you have can actually help you choose what kind of chart would be best to communicate your findings.  Luckily, there is a helpful guide on choosing which type of data visualization works best for your needs. Note that this optional guide includes some primary chart types; there are many other charts and chart variations not shown in the table. If your data matches more than one description, you can create a combo chart that combines more than one visualization type.

Techniques for encoding

As you have been learning, encoding is the act of translating the information represented by your dimensions and measures into visualizations. The artistic elements you choose communicate things about your data:

• Line: Lines in visualizations can be curved or straight; thick or thin; vertical, horizontal, or diagonal. They add visual form to your data and help build the structure for your visualization.
• Shape: Shapes are a great way to add eye-catching contrast—especially size contrast—to your data story.
• Color: Color can help differentiate different elements of visualization and communicate insights.
• Space: Space is the area between, around, and in objects. There should always be space in data visualizations so that the visualization isn’t too cluttered.
• Movement: Movement is used to create a sense of flow or action in a visualization.

These elements can help you break down the parts of your data visualization and what it is communicating—which will help you develop effective visualizations.

Key takeaways

BI professionals are often focused on empowering their stakeholders and users with access to the data they need to answer their own questions. They do this by visualizing measures and dimensions in charts that complement their properties. These dimensions, measures, and strategies for encoding will help develop visualizations users understand and use.

Completed Exemplar

To review the exemplar for this course item, click the following link.

Link to exemplar: Airbnb Athens Charts Exemplar

Assessment of Exemplar

Compare the exemplar to your completed activity. Review your work using each of the criteria in the exemplar. What did you do well? Where can you improve? Use your answers to these questions to guide you as you continue to progress through the course.

Note: The exemplar represents one possible way to complete the activity. Yours will likely differ in certain ways. What’s important is that your charts answer the business problem in this activity.

As a business intelligence professional, you will often design and create visualizations to showcase data findings to stakeholders. In the previous activity, you used Tableau to create visualizations to help answer two questions for a rental hosting business in Athens, Greece:

• What is the average price per night in each neighborhood?
• Where in the city are the highest concentrations of currently available rentals?

To design visualizations that could answer these questions, you accessed and explored the data to determine what metrics had been captured. Then you took the stakeholder’s questions into consideration as you created a bar chart to determine the average price of each rental per night in each neighborhood. You removed an outlier from the visualization to make sure your findings weren’t skewed in the report. And finally, you built a second visualization to demonstrate the concentration of rental properties on a map.

Your charts don’t have to match the exemplar exactly; with these activities, the goal is for your work to achieve the same results, even if you found another way to solve the problem. Basically, this is an opportunity for you to check for understanding, ensure that you’ve met the activities expectations, and explore one possible solution.

Now that you’ve accessed the exemplars, take a moment to compare them to the charts you created. In the exemplar, you will find that this BI professional followed the steps from the activity to create multiple visualizations that allow stakeholders to clearly answer the two questions established in the activity:

• What is the average price per night in each neighborhood?
• Where in the city are the highest concentrations of currently available rentals?

Your work should achieve the same basic use-case, even if it appears slightly different from this exemplar version.

Key takeaways

Being able to recognize what visualizations are useful for presenting data and creating charts based on the questions your stakeholders ask is a large part of a BI professional’s work. Coming up, you are going to have an opportunity to take your reporting one step further and use these charts to create an interactive dashboard that your stakeholders can access themselves.

Making trade-offs

This scenario describes making trade-offs. A trade-off involves balancing various factors, often by prioritizing one element while sacrificing another, in order to arrive at the best possible result.

Qualitative

A dimension is a qualitative data type that can be used to categorize data. Some examples include customer names, product names, and locations.

Encoding

Encoding is the process of translating dimensions and measures into visual representations of data.

Legend

Legends explain how data is encoded in a dashboard so that users can better interpret its insights. The legend in your dashboard describes what each dot size on the map means.

Symbol | Meaning
------- | --------
Blue bar | Number of students enrolled in a course
Orange bar | Number of students who completed the course

Avoid clutter, Increase processing speed, Highlight certain elements

Grouped bar chart

A bar chart is best for illustrating data with a changing variable. A grouped bar chart will visualize values of two categorical variables, so it is easier to make comparisons. Therefore, you could use a grouped horizontal bar chart to compare the average price per night in each neighborhood to the number of currently available rentals in those neighborhoods.

Completed Exemplar

To review the exemplar for this course item, click the following link below

Link to exemplar: Airbnb Athens Dashboard

Assessment of Exemplar

Compare the exemplar to your completed activity. Review your work using each of the criteria in the exemplar. What did you do well? Where can you improve? Use your answers to these questions to guide you as you continue to progress through the course.

Note: The exemplar represents one possible way to complete the activity. Yours will likely differ in certain ways. What’s important is that your dashboard answers the business problem in this activity.

In the previous activity, you used Tableau to lay out elements on a dashboard. You built on your previous work by designing charts for a rental hosting business in Athens, Greece. The goal was to create an interactive version of the dashboard so that stakeholders could access the data and use it to answer their primary questions:

• What is the average price per night in each neighborhood?
• Where in the city are the highest concentrations of currently-available rentals?

To do this, you accessed the charts you made earlier and the appropriate data, organized your dashboard, added charts to the interface, and tested the user experience by using filters. You might have even added additional visualizations to help build out your dashboard further! In this reading, you will be provided with an exemplar of this activity to compare your work to. 

Your dashboard doesn’t have to match the exemplar. The goal is for your work to achieve the same results, even if you found another way to solve the problem. Basically, this is an opportunity for you to check for understanding, ensure that you’ve met the activity’s expectations, and explore a possible solution.

Now that you’ve accessed the exemplar, take a moment to compare it to the dashboard you created. In the exemplar, you will find that this BI professional followed the steps from the activity to create multiple visualizations that allow stakeholders to clearly answer the two questions established in the activity:

• What is the average price per night in each neighborhood?
• Where in the city are the highest concentrations of currently-available rentals?

Your work should answer the same basic questions, even if it differs slightly from this exemplar version.

Key takeaways

As a BI professional, a large part of your job will be designing dashboards that empower users with meaningful data that allows them to answer their own questions and make informed decisions. These insights are key to business intelligence, so designing dashboards that incorporate interactive visualizations that guide business decisions will be a key skill you’ll use a lot in your future career!

Previously, you learned about the importance of optimizing processing speed for dashboards. Processing speed describes how quickly a program can update and load a specified amount of data. Basically, it’s how fast your dashboard can deliver answers to users. Processing speed is usually determined by the volume of data, the number of measures, and the number of dimensions. This is another example of a trade-off: you have to balance various factors, often prioritizing one element while sacrificing another, in order to arrive at the best possible result. In this case, there is a trade-off between processing speed and workload. This reading will offer solutions that enable you to reduce processing load while maintaining the effectiveness of your dashboard.

Reduce processing load

One of the primary ways you can work to optimize your processing speed is by reducing the processing load. You can do this by:

• Pre-aggregating: This is the process of performing calculations on data while it is still in the database. Pre-aggregating data will transform data into a state that’s closer to what you ultimately need because some necessary calculations will happen before the data is sent to the data visualization tool. The trade-off is that your pipeline will involve more steps and your dataset uploaded into the visualization tool will be less flexible , but your users will get the information they need more quickly.
• Using JOINs: JOINS are used to combine rows from two or more tables based on a related column. This basically merges tables together before they’re ever used in the dashboard. This can save a lot of processing load in the actual dashboard. However, if you are trying to join a full table, it can be more of a burden to the system. This is caused by the dimensionality of the tables. For example, joining a one million row table with a 100 million row table will most likely generate a lot of overhead every time the dash is updated. So it’s important to think carefully about how you use JOINs to reduce processing load!
• Filtering: Filtering is the process of showing only the data that meets a specified criteria while hiding the rest. Filtering the data early in your dashboard’s processing means that it doesn’t have to sort through data that isn’t actually going to be used. The tradeoff of this is that this means less data is available for your users to view on their own.
• Linking to external locations: In cases where you have data in your dashboard that you can provide context for outside of the dashboard and which can help cut down on the processing load, you can link out to that location for users to explore on their own.
• Avoiding user-defined functions: Users making requests of your dashboard can add a lot of load to the processing work it’s doing. Consider the kinds of questions that users might have when designing the dashboard so that you can address them without the users themselves having to input functions repeatedly.
• Deciding between data views and tables: Tables contain actual data. Data views are the result of a stored data query that preserves business logic and can be queried like a database. Data views often require much less processing load because they don’t contain actual data, just a view of the data. This makes them less flexible, so you’ll want to consider how interactive you need the data in your dashboard to be.

Key takeaways

When you are considering dashboard design, you’ll have to consider processing speed and load and decide how to best balance them to deliver the answers your stakeholders need as quickly as possible. This can be challenging, but you can apply the strategies described in this reading to reduce processing load and improve performance.

As a business intelligence professional, you won’t just create dashboards and visualizations. You will also share these tools with stakeholders so that they can access the data to get up-to-date information and make informed decisions. You empower stakeholders with the ability to answer their own questions—but you also want to ensure that only the people who are supposed to access that information can do so. This has to do with data privacy and security. In this reading, you will learn about some of the privacy restrictions that are already included with Tableau as well as other common BI tools.

Privacy settings in Tableau

Incorporating privacy settings in your dashboard helps ensure that the data remains secure, even when people from across your organization need to access it for different purposes. Throughout this program, you will be using Tableau to practice key concepts and get familiar with sharing BI insights. Luckily, Tableau already has a variety of privacy and security settings built in that you can take advantage of.

Setting permissions

Tableau gives you the power to set permissions to control how users are interacting with your dashboards and data sources; you can even use permissions to determine which users can access which parts of a workbook. Tableau organizes permissions into projects and groups. Basically, this means you can determine permissions depending on project needs, or by groups of users instead of person-by-person.

You can also use permission settings to choose what metrics users can interact with, show or hide different sheet tabs, or even add explanations of the data that can be seen by different users depending on their specific needs.

To learn more about permissions and how to set them yourself in Tableau, you can check out the Tableau Online Help article about permissions.

Managing user visibility

In addition to allowing you to determine what permissions users have as they interact with your Tableau dashboards, you can also manage how users are able to interact with each other. Usually, all users can view other users’ aliases, project ownership, and comments by other users by default. But in cases where you have created a tool that’s being used by multiple clients, teams, or users who don’t need to interact, you can actually determine how much visibility users have of each other.

To learn more about user visibility settings and how to set them yourself in Tableau, you can check out the Tableau Online Help article about managing user visibility.

Row-level restrictions and filtering

Finally, Tableau allows you  to filter the actual rows of data so users can access the data relevant to their role without having to create an entirely separate view for them. This is especially useful when working with live data sources or extracts that use multiple tables.

To learn more about user visibility settings and how to set them yourself  in Tableau, you can check out the Tableau Online Help article about user filters and row-level restrictions.

Privacy settings in other tools

Other tools that you might encounter as a professional also use privacy settings that allow you to determine what data different users can access and view. Here are some resources you can use to learn more about those tools:

• Data Studio: Sharing, access permissions, and data credentials
• Looker: Access control and permission management
• MicroStrategy: Restricting access to data: security filters
• PowerBI: Power BI Desktop privacy levels

Public availability

To create a dashboard that is accessible to anyone, the public availability privacy setting should be used.

A single item

They want to control the availability of a single item, such as a table, dataset, or single visualization. This is the purpose of object-level privacy permission.

In the database

A business intelligence professional sets up row-level permissions in the database. This privacy setting controls the availability of specific rows of a table or dataset in a dashboard.

Dimension (visualization): A qualitative data type that can be used to categorize data

Encoding: The process of translating dimensions and measures into visual representations of the data

Measure: A quantitative data type that can be either discrete or continuous

Object-level permission: A privacy setting that controls the availability of a single item in a dashboard

Pre-aggregation: The process of performing calculations on data while it is still in the database

Processing speed: How quickly a program can update and load a specified amount of data 

Public availability: A privacy setting that allows anyone to access a dashboard

Row-level permission: A privacy setting that controls the availability of specific rows of a table or dataset in a dashboard

Trade-off: Balancing various factors, often by prioritizing one element while sacrificing another, in order to arrive at the best possible result

• Consider which trade-offs best align with business goals.
• Determine what to prioritize and what to sacrifice in order to produce the best possible result for all users.

making calculations

discrete

Encoding

It prioritizes accessibility

• Prioritization and hierarchy
• Hiding elements to spotlight others
• Use of negative space

• Pre-aggregate the data
• Include fewer measures

row-level permission