Data’s Use in Process Mapping & Modeling Today for AI Tomorrow

Editor’s Note: This article is the continuation of an earlier article about how credit unions can prepare for the fourth industrial revolution – artificial intelligence. In response to that article, readers asked how a credit union cleans up its data to prepare for the onset of AI and simply run a better credit union. This article, the second of a two-part follow-up, helps credit union leaders get started on their path to data hygiene after they’ve embarked on their path to data quality.  

External market pressures demand that the credit union industry do more with less, while constantly improving the member experience. As the use of data grows, it has a positive impact on operational efficiency and member satisfaction that can be realized by any credit union. Through data use, process mapping and modeling, these benefits are obtainable and simultaneously help credit unions prepare for artificial intelligence.

Process mapping is documenting all the steps in a process to deliver a service to a member. Through each step of process mapping there is an opportunity to integrate modern data science techniques with the deployment of models and algorithms. Process mapping models and algorithms are tools to capture critical data, indicate information flow and calculate.

For complex processes, it is beneficial to use an iterative approach when mapping processes. This starts with sketching a process map at the highest level. Next, establish the cost and value generated of each step. Then prioritize each major step for deeper process mapping as needed. Deeper process mapping is best reserved for the steps in which a lot of complexity or teams are involved.

Process Mapping Example: Loan Origination

A credit union that is interested in maximizing the value of its loan origination process would first identify all the major processes and decision points. Then, it would assign a value and cost for each step in the process. Valuation can be tricky, but it is best to remember the whole value chain of the process cannot be greater than the value of a loan, so each process is a fraction. Now it is easy to identify which processes require a deeper dive by looking at the value distribution of the process map.

In the example process map above, the credit union found five major steps in its loan origination process. It might be striking that the values assigned for Processing, Loan Decision and Loan Closing are so low, but it is important to recall the principle of value in Lean. Lean instructs that only processes that are of direct consequence to the member are what matters – the rest is waste. Why do these internal processes have any value? Because they are necessary to produce the loan and for the credit union to stay compliant, which has value.

Seeing the average cost of each step is important as a composite of a credit union’s overall operational efficiency ratio. This tells a leader that if an average loan generates $3,250 in revenue, it will break even. For this sample credit union, this process needs refinement. The process map also shows the days in process or inventory. It shows a maximum of 36 business days and a minimum of eight business days. In the age of Rocket Mortgage, inventory hold times like this are likely not going to be acceptable to members.

Increasing Operational Efficiency

There are several algorithms a credit union can select in its attempt to increase operational efficiency. A waste-sensitive algorithm is the Value Stream Mapping Algorithm (VSMA), which examines every step of a process. It creates a data box, which records the statistics of a given step, such as time to complete, time in queue, number of people involved, errors, volume in, volume out and so on. This data box is completed every time the process runs.

VSMA then looks for the Seven Wastes of Lean that could be contributing to a less-than-optimized process. These seven critical factors are:

• “Defects” or errors in the process;
• “Inventory” or better use of resources elsewhere;
• “Motion” or redundant human tasks;
• “Over-processing” or work that does not return value;
• “Overproduction” or producing excess that is not used;
• “Transport” or unnecessary notifications or handoffs; and
• “Waiting,” the most common form of waste found.

In the example process map, the credit union chose to apply VSMA during the Processing step because of its high cost and low value to the member. Creating the data box for the Processing step as illustrated below allows the credit union to start collecting data to understand the sources of waste. Each time a loan application moves through, an analyst starts a new data box and records the data to create a data-driven view of the process.

The data box at left suggests that the credit union is losing the battle of flow. There are 50 loan applications processed, but six remain in queue. There were errors in which several members input their annual income incorrectly, which took time to correct. Lastly, an average of 3.2 people were required for processing: Two people to print and sort, another to review the documents and a supervisor to sign off on the corrected income error.

Using the VSMA, a credit union leader can apply the 7 Wastes of Lean and take direct action to reduce waste. In the example above, that means preventing income input errors on the application by simply prompting the member to make a correction if they really mean $4,800 per year when it is most likely $48,000 per year. This solves several problems, including reducing the average time to complete and number of people involved. It seems straightforward, but if no one evaluates a credit union’s processes, this waste remains hidden.

Preparing for the Future

These approaches are sensible and easy to use for improving operational efficiency, and thereby increasing member satisfaction. However, they have promises for the future. A credit union that maps its processes now can focus on building models of its business. Models are the foundation for machine learning and AI, but a credit union can employ them today in Excel formulas and process streamlining.

For instance, interview an underwriter and ask them what the key inputs to their lending decision are. Ask them what documents are required, nice to have and unimportant. Take the learning and document it. Then take those data-derived points like credit score, length of membership and annual salary, and create an Excel formula. Use the formula on each lending decision, continuously improving the input and formula, until it is reliable. That is when this model is ready for conversion to an AI process. Before then, it will be a waste of time and money.

For any credit union, it is easy to see how basic process mapping and algorithms can produce tangible results. These results are impactful on operational efficiency, which, as external pressures increase, will no longer be optional but required. Improving operational efficiency will lead to increased member satisfaction, as back-office functions return quicker results to the member. And, these steps are requisite for any credit union interested in positioning itself to be ready for the fourth industrial revolution – AI.

Tim Strasser is the Co-founder of Clear Core and Senior Performance Manager for CUNA Mutual Group’s analytics division. He can be reached at 608-665-4082 or timothy.strasser@cunamutual.com.

Ray K. Ragan is Assistant Vice President, Project Management for Vantage West CU. He can be reached at 520-617-4014 or raymond.ragan@vantagewest.org.