on - edited on by
DRay
Introduction
The following article discusses how to take a good functional dashboard to the next level. It mentions the relevant Sisense features available to help achieve your next goal.
Table of Contents
- Introduction
- Table of Contents
- Dashboard Complexity
- The "Descriptive BI" Approach
- The "Diagnostic BI" Approach
- The "Predictive BI" Approach
- The "Prescriptive BI" Approach
- The "Assisted Intelligence BI" Approach
Dashboard Complexity
In previous articles, I've discussed how you should plan your dashboard (Link), and later, how to assess and monitor its performance (Link). But once one is ready and being used - Can you still somehow enhance it?
The "Dashboard Complexity" term refers to the added value a dashboard provides aside from answering the business question. The name suggests that the dashboard should be more complex. However, the "Complexity" term refers to the amount of data and transformation required to create it.
A "More Complex" dashboard:
- Provides more valuable insights
- Required more data for processing/training
- Requires more dashboard design and data modeling work
To emphasize this, think about a BI Dashboard designed to answer the business question of "Financially, How is the company doing." The following three insights answer the business question, however, provide different benefits:
- Insight #1: Company sales dropped by 15% in the last quarter
- Insight #2: Company sales usually drop by 25% during wintertime. However, they only dropped by 15% this year
- Insight #3: Company sales are currently dropping, you should increase your investment in export by 20K$
While the first insight seems negative, insight #2 provides a perspective and lets you understand the real numbers are positive. The last insight provides actionable instructions on what to do next.
The article will take you through a journey of five different approaches to "Dashboard Complexity." Each provides you with a different focus, challenges, and insights:
- Descriptive BI
- Diagnostic BI
- Predictive BI
- Prescriptive BI
- Assisted Intelligence BI
The "Descriptive BI" Approach
The "Descriptive BI" approach (a.k.a. "Reporting") focuses on the question of "What Happened."
This type of dashboard can be generated by:
- Displaying simple figures (such as an annual revenue)
- Displaying simple facts (such as a month-over-month revenue drop)
- Displaying anomalies (such as a sudden revenue drop in the last quarter)
- Displaying simple trends (such as a gradual growth in monthly revenue)
Let's assess the "Descriptive BI" dashboard:
- Data Complexity: The information required for generating these statements is very easy to generate
- Data Transformation: Data requires simple aggregations (sum/average/etc.)
- User Benefit: The information answers the business question. However, It can't help a user explain the figures shown or help them decide what should happen next.
Available Tools for Implementation
Sisense offers the following tools to generate "Descriptive BI" dashboards:
| Feature | Description | |
| Simple Out-Of-The-Box visualizations | Visualize data graphically (a.k.a. widgets) | Link |
| Dashboard and Widget Filters | Enable focusing on the data of interest | Link |
| Conditional Formatting | Add visual aids to help convey a message about the data | Link |
| AI Exploration Paths | Automatically generate visualizations and insights that anticipate your Viewers' questions. | Link |
| Narratives | Use natural language generation (NLG) to make your data more accessible and easy to understand | Link |
| Trend Lines | Highlight tendencies in your data and get insights quickly | Link |
| Break By | Segment data into groups of interest | |
| The "Switchable Dimensions" add-on | Easily toggle between dimensions displayed in a single widget | Link |
| The "Blox" add-on | Create custom data visualizations by the use of code | Link |
The "Diagnostic BI" Approach
The "Diagnostic BI" approach (a.k.a. "Analysis") focuses on the question of "Why did it happen."
This type of dashboard can be generated by:
- Displaying a root cause (the "A" event took place due to the "B" event)
- Displaying a correlation between variables (the variable "A" has a positive correlation to "B")
- Displaying an insight ("A" is better than "B")
Let's assess the "Diagnostic BI" dashboard:
- Data Complexity: The information required for generating these statements is harder to generate
- Data Transformation: In addition to earlier requirements - Requires teaching Sisense business rules of how to draw a conclusion
- User Benefit: The information answers the business question. However, it provides minimal benefit as it only shows simple correlations (not the "full picture"). A user cannot make educated decisions based on this minimal data.
Available Tools for Implementation
Sisense offers the following tools to generate "Diagnostic BI" dashboards:
| Feature | Description | |
| Drilling | Get an in-depth view of a selected value | Link |
| Hierarchies | Create logical hierarchy paths for an in-depth view of a selected value | Link |
| Break By | Segment data into groups of interest | |
| Explanations | Find the root causes that contributed to a data anomaly | Link |
| Narratives | Use natural language generation (NLG) to make your data more accessible and easy to understand | Link |
| The "Blox" add-on | Create custom data visualizations by the use of code | Link |
The "Predictive BI" Approach
The "Predictive BI" approach focuses on the question of "What will happen next."
This type of dashboard can be generated by:
- Displaying an interactive What-If analysis
- Displaying a forecast based on historical data
- Displaying an optimization recommendation
Let's assess the "Predictive BI" dashboard:
- Data Complexity: The information required for generating these statements is harder to generate (more history, more variables, etc.)
- Data Transformation: In addition to earlier requirements - Requires AI/ML modules to predict forecasts and interactive dashboard widgets to allow a what-if analysis
- User Benefit: The information answers the business question. In addition, it allows simulating various scenarios to help the end-user make educated decisions on what has to happen next.
Available Tools for Implementation
Sisense offers the following tools to generate "Predictive BI" dashboards:
| Feature | Description | |
| Forecast | Forecast future values based on historical data | Link |
| The "Advanced Parameters" add-on | Add user input boxes to enable "What-If" analysis | Link |
| Custom Code | Run Python code from your Jupyter Notebooks to transform and cleanse data | Link |
| The "Blox" add-on | Create custom data visualizations by the use of code | Link |
The "Prescriptive BI" Approach
The "Prescriptive BI" approach focuses on the question of "What should be done next."
This type of dashboard can be generated by:
- Displaying an actionable operation
Let's assess the "Prescriptive BI" dashboard:
- Data Complexity: The information required for generating the right action plan is hard to generate (more history, more variables, more business rules, etc.). The person designing this dashboard will have to conduct strict due diligence and be familiar with every factor and consideration of the decision-making. In addition, the dashboard will require serious UAT.
- Data Transformation: In addition to earlier requirements - Requires teaching Sisense all possible business logic.
- User Benefit: The information answers the business question and tells the user what to do next. However, with power comes responsibility. By providing a firm action plan, the dashboard would become accountable for the result of that action. A positive outcome will result in stickiness, and people will start to depend on it. A negative outcome may result in a lack of trust or complete neglect of the dashboard.
This type of system requires constant maintenance and testing to make sure its data is correct, and its decision-making rules are up to date.
Available Tools for Implementation
Sisense offers the following tools to generate "Prescriptive BI" dashboards:
| Feature | Description | |
| Custom Code | Run Python code from your Jupyter Notebooks to transform and cleanse data | Link |
| The "Blox" add-on | Create custom data visualizations by the use of code | Link |
| Pulse Alerts | Create alerts to notify you when certain thresholds are met, or anomalies in your data are detected | Link |
The "Assisted Intelligence BI" Approach
The "Assisted Intelligence BI" approach focuses on automated processing.
This approach doesn't necessarily require a dashboard as it will trigger the actionable operation on its own. The system will collect the data, process it, generate insights, and act upon it.
Let's assess the "Prescriptive BI" dashboard:
- Data Complexity: The information required for generating the right action plan is the same as in the "Prescriptive BI" approach.
- Data Transformation: In addition to earlier requirements - Requires building interfaces to communicate with external systems (such as an email server, slack bot, or an application via a webhook).
- User Benefit: This approach is fully automated and provides the best user benefit - The user has to do nothing! However, it is also very dangerous as a person learns to trust the system and might neglect to verify the data or update the business rules.
Available Tools for Implementation
Sisense offers the following tools to generate "Assisted Intelligence BI" dashboards by interacting with external applications:
| Feature | Description | |
| Pulse Webhooks | Send notifications through additional 3rd party channels | Link |
