Purposeful Sampling vs Random Sampling
Every company faces decision-making risks. Neuro-ID™ technology can dramatically improve your company’s decision making, allowing you to better identify both risk and opportunity and empower you with new actionable insight to focus resources more efficiently.
When you ask a customer a set of online questions, especially risk relevant questions, you can never be sure about the confidence of their responses. Risk relevant questions are those relating to an activity or event where there are meaningful negative stakes for not disclosing the truth. Did you accurately state your income?
When answering risk relevant questions, people either quickly respond because they know the answer and are confident in sharing their response. Or they delay in responding due to low-confidence in their answer. A lack of confidence can be attributed to not understanding the question, not knowing the answer or not wanting to disclose the answer.
Whatever the reason, a low confidence answer puts your organization at risk and is precisely what Neuro-ID™ technology identifies.
Due to limited resources, companies can’t investigate every answer on every online form completed. So how do you decide which individuals or questions you’re going to look into further?
Suppose a company receives 1,000 applications per day and 10% of those applications contain low confidence answers.
That means, on a typical day, roughly 100 applications contain suspect answers.
It’s impractical to investigate every application submitted. Suppose that the company is only able to investigate 10% of the applications or 100 per day.
The big question is how do you decide which to investigate?
One way is to randomly sample 100 applications from the 1,000 received. You find, on average, that 10 out of the one-hundred sampled have low confidence answers. Unfortunately, using this approach, you will miss on average, 90% of the applications containing low confidence responses.
Unconfident people utilize additional cognitive resources to contemplate, decide, and generate an appropriate response.
Neuroscience research shows this lack of confidence causes simultaneous and uncontrolled changes in a person’s motor-nervous system which influences how they move a mouse, or touch and hold a smartphone.
To detect this low confidence, we sample data at millisecond precision from all modern computing devices; including mice, touch pads, tablets and smartphones. Our patent-pending algorithms interpret this data immediately, so the technology can be deployed into any context where real-time decisions must be made.
Each customer will then be given a confidence score called Neuro-CS™. This score is the purposeful criteria used in selecting those to look at more closely.
What does this mean for an organization’s bottom line?
Suppose that 25% of those customers with a low Neuro-CS™ are indeed answering deceptively. This results in a 150% improvement over random sampling. At 50%, a 400% improvement is realized.
In controlled studies, low Neuro-CS™ has identified fraudulent responses at greater than 90%; resulting in an 800% improvement over random sampling.
Indeed, this is sampling with a purpose!
While every context is different, and detection rates will vary from case to case, this is a powerful tool. It saves your company time while increasing good decisions and decreasing bad decisions.
Smarter questions…better bottom line.
Jeff Jenkins, CTO Neuro-ID™
Joe Valacich, CSO Neuro-ID™
Credit: The authors thank Jackie Valacich for writing this blog post on their behalf.
Neuro-ID™ provides an advanced behavioral and prescriptive analytics platform that can easily be integrated into your current online questions and forms. Guided by years of peer-reviewed science, we capture real-time, millisecond signals based on behavioral and biometric monitoring while a customer fills out an online form. We are then able to provide you with immediate and previously unattainable insight about your customer’s intent.
Prescriptive analytics (PA) is an emerging area of data science dedicated to finding the best course of action for a given situation. PA is one of three general types of analytics which also includes descriptive and predictive analytics.
Descriptive analytics utilizes data mining to help businesses gain hindsight into what has happened in the past in order to predict possible future outcomes.
For example, the FICO credit score is a descriptive analytic approach used by lenders to assess a person’s credit risk based on their financial past. Has this been a good customer?
Predictive analytics provides a probability about what might happen in the future by utilizing relevant current and past information.
For example, one of the most well-known predictive analytic approaches is credit scoring used to predict a person’s likelihood of repaying or defaulting on a loan. Such predictive models combine historical factors, such as credit history, and current factors such as employment status, income and debt level. Using predictive analytics, organizations gain insight into what might happen in the future. Is this a good customer?
Prescriptive analytics is a rapidly emerging field that attempts to gain foresight into what will happen, when it will happen and why it will happen. PA helps you determine your best options in order to take advantage of future opportunities and minimize future risks. Will this be a good customer?
Using PA, Neuro-ID™ determines a customer’s confidence when answering your critical questions. Does the customer show strong confidence when asked about income or expenses?Did the customer accurately represent his/her income on the loan application? AND/OR Does the customer anticipate any changes that will affect his/her ability to pay back this loan?
Gaining this type of insight about your online customers’ needs, wants and intentions, will help guide your best course of action. You will move faster and make better decisions. PA suggests how to take advantage of this predicted future without compromising other priorities.
Neuro-ID’s™ prescriptive analytics compares a user only to themselves. All noise is filtered out while global behavior patterns are analyzed. Indicators of low confidence are then detected and grouped, resulting in a Neuro-CS™ or confidence score. Below is an example of a low confidence response automatically detect by Neuro-ID™.
Data mining is great for understanding what happened in the past.
Predictive analytics brings additional value by combining historic data with current data to predict what might happen in the future.
Prescriptive analytics allows your business to better understand your customer, your risk, your best course forward.
Imagine the value of having foresight into what lies ahead.
Dr. Joe Valacich, CSO Neuro-ID™ (@valacich)
Dr. Jeff Jenkins, CTO Neuro-ID™ (@jljenk)