Tracy Whelen

Geospatial Consultant and participant in the Business Insights & Analytics Leadership Development Program at Travelers Insurance

Photo of Tracy Whelen

Education: M.S. in Geography (University of South Carolina), B.A. in Geography (Mount Holyoke College)  

The following profile was compiled by Brendan Vander Weil (Texas State University) for the Encoding Geography initiative. To learn more, visit: http://www.ncrge.org/encoding-geography/ 

 


Please describe your job, employer, and the primary tasks you perform in your position. 

I am a Geospatial Consultant and participant in the Business Insights & Analytics Leadership Development Program at Travelers Insurance. Travelers is a leading property and casualty insurance company, offering a wide range of personal and business insurance products primarily in the United States and Canada. 

I recently completed an enterprise rotation in Enterprise Data & Analytics, working on data management and quality assessment of enterprise geospatial datasets and ad-hoc geospatial business consulting requests.   

I am currently in a rotation for Claim Business Intelligence & Analytics. My work includes geospatial information delivery and analysis for Claim senior leadership and field offices. Part catastrophe response, part improving everyday claim handling processes.  

Prior to joining Travelers, I received bachelor’s and master’s degrees in geography (Mount Holyoke College and University of South Carolina, respectively). In between my degrees I worked as a GIS Specialist in a remote sensing lab at University of Massachusetts, Amherst.  

How has your education/background in geography prepared you for this position? 

All the data I work with has a spatial component to it, and we often work with thousands (sometimes millions) of records at a time, necessitating strong geography and computer science skills to efficiently store, process, and analyze data, and to deliver actionable outputs.  

Relevant courses from my education that I use today in my job include: 

Geography 

  • GIS/spatial analysis (intro and advanced)  
  • Remote sensing 
  • Spatial modeling  
  • Web GIS  
  • Basic human and physical geography 
  • Electives: Meteorology, hazards geography, business geography 

Computer Science 

  • Introductory scripting (if statements, loops, functions, etc.) 
  • Python 
  • SQL 
  • Data structures 
  • UI/UX design 
  • Javascript (web app development)  

Math 

  • Discrete math (basic logic and set theory)  
  • Statistics (non-spatial and spatial) 

What geographic skills and information do you use most often in your work?  

Geographic concepts that I use in my daily work are important for things such as asking what business problems have a spatial component to them or analyzing the spatial relationship between two or more datasets (e.g. spatial joins and other geospatial analysis). I also need to understand a wide variety of spatial data formats, how to convert between them, and what formats are most appropriate for a given use case (e.g. basic raster and vector formats, enterprise SQL databases, APIs, published feature services, etc.). Finally, I need to know when to use geographic coordinates versus a projection (and what an appropriate projection might be).  

What is an example of applying geography concepts and skills in order to analyze and solve problems in your work? 

One of the many risks Travelers seeks to mitigate are natural hazard events, such as wildfires and hurricanes that climate change may make more extreme. Sustainability at Travelers means performing today, transforming for tomorrow and fulfilling our promise to our customers, communities and employees. Where these two come together is how our Claims department responds to natural hazard events, especially large wildfires or damaging wind events. The following videos capture the spirit of what we do, and the geospatial component of Claims catastrophe response. 

 

Note that Travelers is organized along an Agile structure, with cross-functional teams continuously delivering improvements. While there are always new products and applications being developed, there are also lots of long-term operational systems being continuously used and improved upon. Often employees build on past work and may not see a large project or system from beginning to end. My team’s catastrophe response work is an example of this type of long-term system, and my answers are on behalf of the team.  

What types of geographic questions did you ask and think about in your project? 

The broad business question underlying this issue is, “How can we optimally respond to catastrophe events, meeting customer needs with the most efficient use of business resources?” Underlying questions include: 

  • What location has been/will be impacted? 
  • What is our exposure in the area? (i.e. number of policies, associated financial exposure) 
  • Where have claims already been reported? 
  • How many claims might we expect? 
  • What types of claims do we expect to see from this event? (e.g., wind, water, fire, etc.) 
  • What types of damage occurred, and how severe is the damage? 
    • Will this impact our ability to respond, either because an area is inaccessible or because local offices or employee homes have been damaged? 
  • Where can we acquire the necessary data from? 
  • Can we develop models to more efficiently review post-event imagery as part of the catastrophe response process? 
    • If so, what features are we trying to spot in the imagery?  
    • How does this vary by event type? 
    • What might be appropriate modeling algorithms to use? 
    • What are some of the challenges the model might encounter? 

My team does not directly answer all these questions, but we need to be able to provide appropriate data to the senior leadership and other decision makers or support staff who can build a final answer.  

What types of data did you acquire to support your project? 

  • Business data (e.g. claims, policies) 
  • Event data – wildfire boundaries, hurricane wind footprints, precipitation measurements, tornado damage reports, etc. 
  • Aerial imagery and derived model output 
  • Property geometry data (e.g. building footprints, parcel boundaries)

What types of content knowledge and skills (both geographic and more general) did you use to evaluate, process, and analyze the data you gathered for your project? 

In the moment skills that we use on this project for responding to a single catastrophe event include: 

  • Querying databases (spatial and nonspatial joins, filters) 
  • Combining and reformatting a variety of data formats  
  • Running models in python scripts 
  • Common sense/data quality checks 

For long-term projects, the output of which gets used in catastrophe response (multiple team effort), the skills we use are: 

  • Internal model development in partnership with data scientists  
    • Curate input data (image locations, image clipping geometry, training data, etc.) using SQL and python 
    • Evaluating model results against other sources of truth 
  • Evaluating new 3rd party datasets (accuracy, timeliness, availability, cost, other potential sources for the same information)

How did you communicate the results of your project (e.g., writing technical reports, making maps and geo-visualizations, creating graphics, data tables, etc.)? Do you have a recent product or publication to share with us as an example? 

We publish web GIS content as both data layers and maps, creating different versions for different user groups in order to control access to sensitive information. We also use frequent email communication, whether it is one on one, small group with specific questions and answers, or larger list-serv communications (with standardized templates) at key time points during catastrophe response (e.g. web map published, imagery collected, etc.). Additionally, we communicate results through spatial SQL data pulls (tabular format)  

See the below list for examples of broader enterprise or external communications about projects and programs mentioned in this interview. 

  • Travelers 2021 Q3 Earnings call. Note CEO Alan Schnitzer’s introductory remarks including, “location intelligence at the parcel level” and our “AI Assisted Claim Damage Detection Model was a key part of our Ida claim response” 
  • Interview of Adam Sobek (Travelers AVP of Geospatial) at NearMap Navig8 Conference 2020 (Travelers’ use of imagery, including for catastrophe events) 

What are the criteria that you use to assess the quality of your results?   

Most important criterium: Has the business need been met?  

Other important criteria: 

  • Validate data quality 
  • Spatial scale and level of accuracy  
  • Minimizing false negatives, minimizing false positives.  
  • Minimizing process (time, number of steps)

The business need at hand dictates which criteria are important, which varies from question to question. Examples include: 

  • Level of address accuracy needed to plot individual policies versus summarize at a zip code level 
  • Some analytics results are only valuable if they can be completed faster than more manual processes out in the field. 

This material is based upon work supported by the National Science Foundation under Grants No. 2031418, 2031407, and 2031380 (Collaborative Research: Encoding Geography – Scaling up an RPP to achieve inclusive geocomputational education). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation 

    Share