Using Data Science to Trust but Verify

Demand for insights about customer experience is more urgent than ever. Whether it’s in-depth strategic research led by insights professionals or customer experience feedback collected by product managers, marketers, or customer care teams, clients need trustworthy knowledge to make data-driven decisions.

Look at David Zaslav, the newly minted CEO of Warner Bros. Discovery. According to The WSJ, he emailed this to his leadership team, “As we build this new company, we need to be guided by data and insights to understand what’s working and what’s not.”

A media merger like the one he’s stewarding is the highest of stakes. However, today’s inflation and supply chain shortages mean that bad decisions by any size or type of business are more expensive and difficult to rectify.

Data quality must be a primary focus; and, our industry must invest more and innovate faster in fraud-fighting technologies.

First, what is Data Science?

Data science combines math and statistics, specialized programming, advanced analytics, artificial intelligence (AI), and machine learning with specific subject matter expertise to uncover actionable insights hidden in an organization’s data. These insights can be used to guide decision making and strategic planning.

Deepening industry-wide investments in data science, artificial intelligence, and automation – and sharing our learnings – are imperative to raising the bar on the quality of the data collected.

 

What can be done about data quality?

Given the volume and rising sophistication of research fraud, we simply won’t outpace the thieves without technological solutions. And it’s equally important that we use these tools to preserve the experience of good respondents.

 

Consider a tiered approach to identity validation

What leading providers find works best are tiered approaches to identity validation that combine passive and active measures. In this type of approach, validation requirements can ramp up based on the likelihood that a potential audience member may be invalid.

Machine learning models are key to this. With these tools that can improve over time, a provider can canvass thousands of data points, including respondent-supplied information, third-party data sources, and later… the behavior within the survey platform, scrupulously looking for suspicious signals.

By flagging suspicious activity, the machine learning model can tell you how vigorous to be in verification and very importantly — when to usher someone authentic and sought after by clients into their projects. Technology can help us to nurture the “babies” while discarding the bathwater and the “soap scum.”

 

Use machine learning across the respondent’s life cycle

In a “trust but verify” approach, you can also check multiple fraud indicators at the outset and assign tags to respondents for use in other models later. With a combination of supervised and unsupervised learning models, you can efficiently monitor a respondent’s full journey within your environment. When done right, the models learn emerging and suspicious behavioral patterns and prompt intervention. This empowers you to catch a fraudster who may have made it through your initial defenses.

When you add automation on top of machine learning, this empowers you to speed intervention when required, and to ensure continuous audience grooming. This must be done very carefully to protect the respondent’s experience. Some respondents may simply need to take additional steps, like biometric identity verification, to prove they are who say they are… and that they are qualified for a particular study.

Technology solutions are simply imperative to meet the sophistication of today’s bad actors. For example, we’ve seen groups of synthetic profiles with similar behaviors that targeted specific surveys and manipulated the links. Providers have also seen fraudsters signup with hundreds of duplicate accounts predicated on different disposable domains.

 

Stay vigilant

I talk to clients about fraud every day. It’s top of mind for them. Not long ago, there were daily headlines about fraud in programmatic advertising. Clients learned then that the “people” on the other side of an ad impression were not all the same – and that investing in quality audience supply and tech defenses was worthwhile. The parallels to our corner of the industry are very clear.

It’s incumbent on all of us to invest, innovate, and do more.