At Credit Human, System Dynamics is Part of the Mission

Businesses around the world engage in system dynamics projects. It’s less common to find a business that builds system dynamics into its managerial toolkit and rarer still for a business to incorporate system dynamics in its mission statement. Credit Human, a San Antonio-based credit union serving over 200,000 members, does all three.

Credit Human’s embrace of system dynamics as an operational practice began, as it does in many organizations, with a critical project – the launch of manufactured housing lending. “In the early 2000s, there was a huge demand for manufactured housing but so many lenders had gone through boom and bust cycles,” says Steve Hennigan, CEO. “We wanted to participate in the market in a more sustainable way.”

Hennigan had been introduced to system dynamics through an MIT executive program taught by Dr. Jim Hines. He continued to read and study system dynamics and knew that causal loops and modeling would help Credit Union analyze the manufactured housing market its potential role in it. He asked Hines, who had left MIT, to serve as a consultant at Credit Human to build a system dynamics practice, a long-term effort that began with the manufactured housing market investigation.

“Lenders sell services through dealers who are selling manufactured housing,” says Hines. “We thought that we would need a sales force time-allocation model in order to understand how many people were needed to sign dealers. Unexpectedly, simulations showed one of our three manufactured housing regions running out of prospects. That aha moment came with critical policy implications, namely, that the sales force in that region should stop prospecting. The market was already nearing saturation.”

A separate model uncovered a problem in the manufactured housing market’s credit cycle. “Banks base lending decisions on average FICO score standards that oscillate or cycle,” says Hines. “When defaults are high, the FICO standard is raised. When defaults are low, the standard is lowered. In each cycle, standards rise to levels that are too high and then fall to levels that are too low.”

“Based on credit cycle modeling, model analysis, and real-world testing, we developed a policy for the entire credit union that changes the information used to change credit standards,” says Hines. “The overall result of our work is a reduction in the amplitude – the high highs and the low lows – of the credit cycle we experience. That’s both better for our borrowers and makes it easier to manage lending overall. In the manufactured housing market, in particular, we are a steadier lending partner to dealers which has made us the third largest manufactured housing lender in the country.”

“We turned the model into a credit cycle policy for all our product lines,” says Hennigan. “It’s documented and part of our day-to-day operation that receives institutional revisiting, review, and compliance checks.”

Using causal loops and models helps [managers] understand what’s happening inside and outside of work so they can find better solutions.

Credit Human is also using system dynamics to investigate service quality erosion with the assistance of Dr. Rogelio Oliva, professor at the University of Texas A&M and winner of the 2019 Forrester Award. “Service quality suffers when you don’t properly resource and fatigue sets in,” says Hennigan. “The typical reaction is to overhire, which creates other problems. Managers don’t always understand the work that’s happening below them, so they don’t understand the real cause of service quality erosion. Maybe the cause is outside of work. Using causal loops and models helps them understand what’s happening inside and outside of work so they can find better solutions.”

A system dynamics look at productivity is very tied to the Credit Human mission to serve people of modest means. “In lending, productivity doesn’t increase over time,” says Hennigan. “There are only so many loans that can be processed in a day. If you want to grow and increase wages, you have to process more loans and that leads to corner cutting. We’ve seen what happens when lending institutions cut corners. So, for-profit banks are pursuing larger loans, which leaves out lower income families. That’s our market and it’s a huge opportunity.”

Credit Human’s intention to successfully serve that market by employing systems dynamics as an operational practice is laid out in a revitalized mission document that was, itself, developed with a model. It begins by stating that Credit Human’s mission “is to help people create, build, and maintain financial slack”, especially those “who either experience or at risk of experiencing financial stress because they lack desired slack.” The second key principle in delivering on that mission is “use a system dynamics approach to understand and solve the challenges of our mission.”

“Moving forward, we’re making system dynamics a part of our problem-solving toolkit,” says Hennigan. “Mid-level managers will apply system dynamics to concrete, day-to-day problems and higher -level executives will look at bigger, more abstract issues. They’ll understand what’s going on before forwarding a solution.”

To help managers build their own system dynamics toolkits, Hines is working with Michael Friedman, a consultant who is helping Credit Human apply Requisite Organization principles. “Managers and executives don’t have time to become a system dynamics expert in a couple of years,” says Hines. “I started thinking about a longer learning process and, working with Michael, I got an idea.”

Hines and Friedman thought about their own system dynamics and Requisite Organization learning experiences. They realized that their practice area was their primary mental model (Mental Model 1 or MM1) and other things they had learned were secondary mental models (MM2) that supported MM1. “For me, MM1 is system dynamics,” says Hines. “Math, computer programming, behavioral economics, and organizational evolution are MM2s that extend and deepen my understanding and use of system dynamics. For example, my background in system dynamics enabled me to learn aspects of differential equations that would help me in my system dynamics work.”

For Credit Human managers, their job is MM1. System dynamics is MM2. “Rather than teaching system dynamics as an MM1, we’ll teach it as an MM2,” says Hines. “People will learn at a time that’s relevant to their own job requirements or career development.”

Requisite Organization provides a framework by creating management levels that Hennigan has used to define system dynamics adoption and use expectations. “At lower management levels, people might learn how to use reference modes or simple causal loops to think through a problem,” says Hines. “At higher levels, they’ll create full models and simulations to investigate broader, more strategic issues.”

Hines points out that every step in the system dynamics method is useful. “Just listing variables helps managers calm down and focus,” says Hines. “Going on to create reference modes adds to understanding and causal loop diagrams add insight. Simulation modeling deepens prior insights which uncover new ones. Applying system dynamics at any level helps people see their own role in a problem which helps eliminate arrogance and broaden viewpoints. Perhaps most important, system dynamics requires managers to separate the problem from its causes and possible solutions. Often those things are discussed in a confusing jumble.”

Credit Human is compiling a library of system dynamics work and models to support employee learning. Sharing causal loops and models will allow peers to learn from one another, apply work done in one area to new issues, or use prior work as a starting point for new models and simulations. And, like the credit cycle policy, Credit Human will continue to use models to formulate policies that direct on-going decision making.

This is a hard journey, but once you change the way you work it gets easier, and it’s nice to solve problems once rather than again and again.

“Increasing the use of system dynamics throughout management levels is a big project,” says Hennigan. “Jim and I work together five to six hours a week. People know that I appreciate it when they think about and explain issues using causal loops. This is a hard journey but once you change the way you work it gets easier and it’s nice to solve problems once rather than again and again.”

“When I was a doctoral student at MIT, studying under Jay Forrester, I believed that system dynamics would become part of the managerial fabric of most businesses,” says Hines. “I was enthusiastic to help make that happen but, sadly, it hasn’t. Credit Human is the most remarkable company I’ve worked with, the most managerially innovative. It shows that when a company gets its act together with system dynamics, it will succeed.”

Modeling Water Supply and Demand in Northeast Texas

Philosopher Adam Smith’s thoughts on supply and demand relationships have inspired generations of thinkers, Ed Weaver among them. “I was sitting in a systems dynamics class where we were talking about supply and demand looking at John Sterman’s causal loop representation of Smith’s philosophy and thought ‘That’s it!’” says Weaver. “I need to look at supply and demand in our water system from Smith’s supply and demand model.”

A PhD student at Colorado State University, Weaver is also Integrated Pipeline Manager for the Tarrant Regional Water District (TRWD), one of three major raw water suppliers to customers in the Dallas/Fort Worth, Texas area. Weaver joined TRWD right out of high school and, over the last 45 years, has worked his way through every level of the operation, welding to operations to construction to engineering. Now he’s looking for ways to ensure that the district’s customers continue to have enough water as the population increases and climate changes.

Texas currently has 16 water supply planning regions, each containing around 16 counties. Each region is responsible for providing its own planning group that is in charge of developing and updating their respective 50-year horizon water plans on five-year recurring cycles. Within Region C, the primary raw water suppliers are the City of Dallas Water Utilities, North Texas Municipal Water District, and Tarrant Regional Water District. Together, those suppliers serve roughly seven million people, a number that is projected to double by 2070. Weaver’s academic work in supply and demand inspired him to think about Region C as a whole – a single system rather than three separate suppliers – an exercise that would benefit water suppliers and consumers in and around Dallas/Fort Worth and serve as his PhD dissertation project.

“I decided to look at the Region C water resource plan from a system engineering perspective,” says Weaver. “Texas is one of the fastest growing states in the country in terms of population and the Dallas/Fort Worth area is one of the fastest growing in Texas. The 2016 state water plan anticipates that the growing population will boost demand 71% by 2070. We also know that supply is impacted by increased use and climate variation. Air, water, and food, in that order, are needed for survival so we have to know that we’ll have the water we need as conditions change.”

Water suppliers in Region C could add supply by building new reservoirs (TRWD currently manages four reservoirs and one constructed wetland), but that’s a complex, expensive proposition. “Reservoir building is environmentally, socially, and economically challenging,” says Weaver. “State and federal laws and regulations in place to protect wetlands, historical and cultural areas, wildlife, and certain industries limit the areas where reservoirs can be sited.”

“I started thinking that, rather than building new reservoirs, maybe we just need a new conveyance between the current reservoirs in the Region,” says Weaver. “If the three water suppliers in Region C can collectively meet growing demand, we can avoid or at least defer adding new reservoirs.”

Currently, Weaver is working on the Integrated Pipeline Project (IPL). The IPL is a joint 350 million gallon per day, 150-mile, $2.3 billion pipeline project in partnership with the City of Dallas. Initially, each supplier had plans for individual pipelines from existing reservoirs to meet their respective demands. The driver for a joint system was physical proximity of TRWD and Dallas water sources and projected supply shortages. In the IPL system, the conveyance is shared while the sources are independent by supplier. Preliminary business case analyses showed that there was a potential $500 million capital cost savings and a one-billion-dollar life-cycle cost savings shared between TRWD and Dallas.

He made that discovery through model building exercises using Stella® and some spreadsheet models. “We had to get a handle on how the integrated pipeline would work and our consultants suggested we initially use Stella to model the system,” says Weaver. “We put together a rudimentary model and then played around with our ideas.” The primary benefit for Weaver was the dynamic visualization of the system’s behavior and functionality.

Of course, expanding or integrating an already complicated water resource planning project across a larger area just adds to its complexity. “We add to the already multiple water sources, we increase the number of sources and resource owners, and we add new political considerations,” says Weaver. “There are intrastate compacts and rules and laws based on the water resources needed. There is also a natural reluctance to share among people who are used to taking a single agency point of view. How do you get people to think about resources across the region from a single system perspective?”

Weaver’s approach to answering that question is a model that will help suppliers and users see Region C water districts as a single system. They’ll understand current and projected supply and demand and how sharing resources via regional interconnections of existing sources benefits their customers. “I want anyone to be able to use the model, even if they don’t have any experience in systems engineering,” says Weaver. “Anyone who is charged with delivering raw water can look at the Region and say, ‘Here is what we’ve got and here is where we need it to go.’”

Stella is easy to learn and it does what I need it to do.

While he had experience using another model building application in conjunction with his PhD program, Weaver continued using Stella Architect to build his Region C model. “We already use Stella at TRWD and, while I like the model building features in other applications, I really like Stella’s equation editor which highlights problems,” says Weaver. “That’s much easier than having to go back and figure out why something isn’t working. Stella is easy to learn and it does what I need it to do.”

The isee systems public workshop Introduction to Dynamic Modeling with Stella gave Weaver a chance to work on his model under the guidance of Bob Eberlein, Co-President of isee systems. “Bob led us through stocks, flows, arrays, and equations and gave us time to work through exercises,” says Weaver. “I used that time to build the reservoir system model. My biggest challenge was getting the relationships right in equations and, when Bob came by, he gave me one-on-one assistance.”

He’s now taking advantage of eight hours of individualized model building support. “I use the eight hours in two-hour blocks and work with Bob online. He has helped me visualize supply and demand and fit them together,” says Weaver. “In our next session, we’ll work on a dashboard that will allow end users to run what-ifs by changing ambient temperature, precipitation amounts, and other variables.”

You can’t beat one-on-one, hands-on, applied problem solving from someone with Bob’s experience. Together we got more done in two hours that I would in a week by myself using manuals.

The training course and consulting hours have given Weaver a big assist in model building and improvement. “Trying to build a converter, for example, isn’t hard once you know how to do it,” says Weaver. “You can’t beat one-on-one, hands-on, applied problem solving from someone with Bob’s experience. Together, we get more done in two hours that I would in a week by myself using manuals and help.”

When complete, Weaver’s model should provide a new, region-wide view of water resource supply and demand and consider usage rates over shorter durations. “Each water supplier already looks at the drought of record, population growth, and water use year to year,” says Weaver. “With the model, they’ll be able to look at month-to-month demand periods and determine if they need additional resources. By taking a system-wide view, they’ll find those resources short of constructing a new source.” That meets the primary goal of Weaver’s dissertation project: to determine any possibility to defer the building of new reservoirs by interconnecting existing regional supply systems, regardless of their ownership and in consideration of conservation and climate variability impacts across the system.

Having gathered data on the supply side surface water (primary sources for region C), Weaver is now collecting demand data from each of Region C’s water suppliers. In addition, he has built in equations to account for predicted evaporation. Data from the National Oceanic and Atmospheric Administration and National Aeronautics and Space Administration provide starting points for those adjustments. Population projection data is coming from the 2016 Region C Plan, the Texas Water Development Board, and the 2010 US Census.

With considerable work under his belt, Weaver is hoping to complete the model in the next two years. “When the model is finished, I’ll write it up for my dissertation, hopefully pass, and hand it off to planning directors so they can use it for long-term planning,” says Weaver. “Casual, multidiscipline users will use a dashboard to select or adjust demand, population, ambient temperature, precipitation, and/or conservation variables to test and see water supply and demand relationships in a dynamic model. That’s much more effective than handing them a spreadsheet.”

Tips and Tricks by Jon Darkow

Converters: Leveraging Engaging Interfaces

Designing an interface is different than designing a system dynamics model. When designing a system dynamics model I try to in use the fewest number of primitives that will generate the range of patterns of behaviors that I am interested in understanding. However, when designing interfaces I try to maximize playfulness, experimentation, and understanding for the end-users, my students. When designing features for interfaces to my model, I use many converters. Stocks and flows form the architecture of the system's behavior, and the converters are levers by which users will evaluate the system’s structure. I will always have users manipulating converters that are then, in turn, manipulating stocks, flows, or other converters. For example, I will create converters to represent growth and decomposition rates, effects from temperature, and the initial quantities of stocks.

One way I like my students to interact with models is by setting the initial quantity of a stock. I create a converter and then, instead of setting the initial value of a stock to a number, I set the initial value of the stock to the converter. In the interface, a slider of the converter can be inserted so users can easily manipulate the starting value of a stock.

Most of my models have the RANDOM function so my students can practice statistical reasoning. Users can then perform descriptive statistics, like averages and standard error of the mean, and inferential statistics, like Chi-square and the student T. Adding random functions as converters allow my science students to see how increasing sample sizes account for natural variation and encourages them to run many, many simulation runs.

In addition to perturbing the parameters of the model, I use converters to help communicate outputs of a model. Now that scalable vector graphics (SVGs) can be animated using quantities in a Stella model, qualitative data can be added to an interface. For example, I have added color changing graphics to illustrate chemical changes during a simulation. Adding an SVG of a test tube as an “Animation Object” into a Stella interface, users can easily simultaneously examine how the color of the test tube changes over time, and see in a line graph of the chemical changes over time.

Converters help create playful simulations. They can be used to manipulate stocks and flows and add variation to simulation runs. Additionally, converters can help add multiple representations to interfaces by adding qualitative changes to SVGs. Playful, variable, and illustrative interfaces help engage my students. Converters help bring these designs into an interface.

Making Connections 2020

We are excited to announce that we will be hosting our next user conference, Making Connections, on October 1st & 2nd at the Holiday Inn in Concord, New Hampshire. Come join us for this great chance to network with other Stella users and the isee systems staff while enjoying the beauty of New England's foliage season. Watch for further information and a call for papers in early 2020. We hope to see you there!

On the Road

This fall has been a busy for Co-President Karim Chichakly. In mid-November, Karim traveled down to Bogotá, Columbia for the 17th annual Latin American Conference on System Dynamics. He gave a talk on “Systems Thinking for Public Policy: A Case Study in Brazil”, taught a workshop, and participated in a panel discussion on developing a career in System Dynamics.

Following CLADS, Karim continued his southward journey to Brazil, where he worked with the Brazilian Federal Government to help coordinate efforts in using Systems Thinking and Dynamic Modeling to better implement policies that benefit society. Karim conducted 8 workshops to incorporate Systems Thinking into several projects Brazil will be conducting in 2020. Karim also gave a presentation at ENAP on “Systems Thinking and Public Policy, which can be viewed here.

Making it back stateside, Karim spent a week in early December in Annapolis, MD supporting a group working to link “feedbacks between the social and climate systems. The impacts of climate change are likely to alter human perceptions of climate change and associated greenhouse gas (GHG) emission behaviors that, in turn, influence the magnitude of climate change.”

Early February will see Co-President Bob Eberlein traveling to Brisbane, Australia to attend the 3rd Asia-Pacific System Dynamics Conference. This conference will showcase System Dynamic applications within the Asia-Pacific region. Bob will be giving a pre-conference Intermediate Workshop that will cover more advanced topics such as sensitivity testing and calibration. We look forward to connecting with Stella users in this region and seeing all their great projects.

New Online Courses

We are partway through updating our six-course series From Systems Thinking to Dynamic Modeling. The first three courses, Introduction to Dynamic Modeling I, Introduction to Dynamic Modeling II, and Dynamic Modeling I have now been recorded and are available online for purchase. These are now prerecorded, self-paced courses. The final three courses, Dynamic Modeling II, Intermediate Dynamic Modeling I, and Intermediate Dynamic Modeling II will be offered as live courses in the beginning of 2020. To learn more about each of these courses, visit our bundle page.

Barry Richmond Award

Established in 2007 to honor and continue the legacy of isee systems’ founder, this award is presented to a deserving Systems Thinking practitioner whose work demonstrates a desire to expand the field or apply Systems Thinking to current social issues. This year’s winner was Dartmouth student Rachel Matsumoto for her paper “The Dynamics of Social Movements.” Her paper illustrated why so many social movements fail, as well as the dynamics required to have a successful social movement. The award was presented to her at the International System Dynamics Conference in Albuquerque, New Mexico. If you are interested in submitting your work for the Barry Richmond Award, you can learn more about the requirements here.