Gartner BI Bake-Offs offer prospective buyers side-by-side comparisons of analytics/BI vendors based on scripted demos using common datasets in a controlled setting. This year’s focus was disaster/flood analysis based on data from the Organization for Economic Cooperation and Development (OECD) and Federal Emergency Management Agency (FEMA). This post demonstrates the Tellius approach to analyzing flooding in both video and write-up form. Check it out!
In This Post
- What Is Tellius? What Are We Analyzing?
- Key Findings
- Ad-Hoc Exploration of Disaster-Affected Populations in Natural Language
- Bringing in More Data: U.S. Flood Insurance Claims
- Disaster-Affected Population Change Root Cause Analysis via Automated Insights
- Predictive Analytics
- Sharing Analysis Results
- Conclusion / Next Steps
What Is Tellius? What Are We Analyzing?
What is Tellius? Tellius enables organizations to get faster insights from their data using AI-powered automation. Any user can ask ad hoc questions across billions of records via a Google-like interface, understand “why” metrics change via automated insights that surface hidden key drivers and trends, and get predictive recommendations—without relying on data experts. Tellius complements traditional BI tools by helping users go beyond static dashboards/reports to gain ad-hoc answers, insights, and business-friendly advanced analytics—fueling better decisions.
What are we analyzing? We’re analyzing data related to flooding from the OECD, FEMA, and U.S.-based flood insurance claims to identify key trends and recommendations to share with policymakers.
- After analyzing global disasters across the past 30 years to understand the impact of floods vs. other natural disasters, we discovered that from 1990 to 1991, there were a lot of land and mudslides in eastern Asia that led to a 388% increase in total affected population due to floods caused by heavy rains. Put simply, indirect flood-related disasters have proven devastating.
- When we explored river and coastal flooding in the U.S. on a city level, we found that river flooding areas had a drastic increase starting in 2015.
- Finally, by enriching the data with flood Insurance claims, we found that the most recent dollar claim value impact occurred in Ada, Okla.
Ad-Hoc Exploration of Disaster-Affected Populations in Natural Language
How did we get to these insights? We start by querying the data in natural language—a strength of the Tellius platform as the only provider of true natural language search (compared to simple keyword search)—specifically, “show me affected population by disaster type”…
…which automatically parses the relevant answer from the underlying datasets in a fraction of a second and returns an answer in the form of a best-fit visualization. In this case, we see that the top three most impactful disaster types are flooding, drought, and storms.
We can hone our analysis by exploring within flooding the monthly coastal and river flooding exposure…
…to reveal that the river flooding outpaced coastal flooding around 2015, likely due to increased rainfall and human erosion activities.
The Tellius Difference: Natural language search allows end users to simply input questions in a Google-like interface to automatically receive recommended visualizations, saving time and unlocking the value of data, rather than time-consuming cycles of requesting answers from analysts, waiting for custom reports, etc.
Bringing in More Data: U.S. Flood Insurance Claims
From this global view, let’s zoom in to the United States of America to see which areas OF the U.S. are most prone to flooding. To do this, we decided to bring in additional data to the OECD and FEMA datasets: U.S. flood insurance claims. A strength of the Tellius analytics platform is the ability to quickly tap into multiple data sources to expedite ad hoc data analysis. In this case, we brought in additional flood insurance claims data from across the U.S. in the last few years, which we joined and prepped through a few simple no-code point-and-click steps, as well as some code-based approaches visible below in the form of a visual data pipeline, for full transparency around every transformation performed to the data:
Now we can ask the data to reveal the amount paid on claims for the last three years by state, which is auto-visualized onto a geographic heat map, revealing that most flood insurance claims were paid out in the South/Southeast states of Oklahoma, Texas, and Louisiana, where storm and flooding activities are prone.
The Tellius Difference: Rather than having you spend time picking measures, dimensions, and chart types, Tellius’ automated visualizations pick best-fit visualizations based on the context and question asked, dramatically cutting down on the time necessary to create visual data stories.
Disaster-Affected Population Change Root Cause Analysis via Automated Insights
Looking back on the global flooding levels of the past 30 years, we see an interesting upward trend in the amount of population impacted by floods, with a particularly massive jump (388%!) from 1990-1991. Why did this happen?
To understand the key drivers, we utilized Tellius’ built-in Automated Insights to perform root cause analysis.
Top change contributors are given in plain English at the top, with key drivers laid out on the left-hand side such as land, mud, snow, and rock slides; heavy rains; and the Yangtze river basin region of China. Other change contributors are outlined in the automated analysis, further pinpointing the “why” behind the “what” that we identified in our ad hoc analysis earlier.
Tellius also includes several other types of automated insights, including comparison drivers—where we could compare two cohorts of population to understand what’s different between these two worlds—as well as key driver analysis.
Using key driver analysis, we could explore the key drivers of disaster casualties in 2021, in which our system would analyze the drivers of the top 20% of casualties rates, revealing that the specific disaster type and population density are top drivers for increased death tolls.
The Tellius Difference: Traditional BI tools help pinpoint what metrics have changed, but they fall short when it comes to identifying the drivers, trends, and root causes of those changes. Tellius’ Automated Insights provide a robust, automated approach to diagnostic analytics in an intuitive and business-friendly manner.
In addition to the ad-hoc analysis and insights generation capabilities outlined above, our platform also allows for predictive analytics. For example, it is possible to train a machine learning model to perform a wide variety of ML tasks on this data, such as:
- Predicting the number of deaths by using a regression model;
- Classifying more/less impactful variables using a binary or multi-class classification model; and
- Performing clustering analysis on different regions where these disasters may occur.
Machine learning modeling is available in a point-and-click or autoML interface, in which Tellius chooses the ML algorithms and hyperparameters, compares multiple models, and picks the best-fit model for your needs.
The Tellius Difference: Advanced analytics often require different workbenches and domain experts, causing time-consuming delays in analysis/decision-making. Tellius offers business-friendly predictive analytics right out of the box, allowing for analytics flexibility and agility.
Sharing Analysis Results
As each step of analysis is completed (or at the end, once the story comes together), it is very easy to share and collaborate in Tellius. Any analytic object—visualizations, insights, predictions, dashboards, etc.—can be shared with specific user groups or individuals either via URL (the receiving user doesn’t need to have access to the entire platform, but they could just access the specific analysis and still drill as needed), in a shareable dashboard, or by exporting into a number of different file formats, including native PowerPoint files, which retain the underlying data so end users can utilize company colors/templates for better personalization.
The Tellius Difference: Analytics is a team sport. By offering a highly collaborative and shareable/exportable single pane of glass for analytics, Tellius allows people with various domain and technical expertise to benefit from data.
Conclusion / Next Steps
As you have seen, Tellius makes understanding “what” is happening and “why” flooding is occurring simple, using natural language and automated insights that automate complex data analysis to uncover root causes. Ultimately, a similar analysis could be useful for:
- Disaster preparedness and response: Helping policymakers identify areas that are most susceptible to floods and develop strategies to prepare/respond to flood disasters.
- Infrastructure planning: Informing policymakers’ decisions about infrastructure planning, such as building codes and standards, as well as the design of stormwater management systems and flood control measures.
- Climate change adaptation: Helping policymakers understand the potential impacts of climate change on flood risk and develop policies to adapt to these changes.
Policy decisions, in turn, may include zoning regulations, insurance requirements, and infrastructure investment decisions, to name a few areas. Whether it’s flood analysis or customer lifetime value analysis or physician targeting or anything in between, Tellius helps expedite the journey from data to decisions.
Check out our analysis of last year’s Gartner bake-off here or explore how Tellius could benefit your business regardless of industry here.