Charlie Allieri, CEO of Imperium, joins Sima on today’s show to discuss data quality in the market research space.
Imperium offers a full suite of solutions, but their goal is to help customers with data quality issues that they’re dealing with. Data quality in the market research space usually aligns with people and machines who are intentionally providing bad data (also known as fraud).
A Device Solution
RelevantID® is a solution that focuses on the device, ensuring that the device isn’t coming back multiple times trying to take a survey. This involves digital fingerprinting and de-duping and represents the first level of defense.
On top of this, there are other indicators and data points that can be used to triangulate in order to get a more precise determination on whether that person is really fraudulent or not.
Measuring the Bad and the Good
The Verity® service they offer has two parts: the first part is making sure the respondents are not bad, and the second part is verifying the respondents are good.
“Fraudsters have an incentive to misbehave, and the barrier to technology to do these things are getting lower and lower.”
There are behaviors around the community that they see in their data. That’s the first solution they were able to create by mining all the data that they have access to, and seeing behaviors across all of their customers and being able to indicate if a respondent’s quality is not going to be very good or it’s outright fraud.
Another Imperium solution involves analyzing the behaviors in open-end responses: Are they speeding through? Are they cutting and pasting? And the latest contextual awareness: Does that answer actually apply to the question being answered?
“The ability to deliver value within that cost requirement is key to this market.”
Traditionally, humans would be used to look at all the responses, but Imperium can look at the responses in real-time and kick it out, or give a score to customers that indicate they should not allow that data to get into their systems.
Patterns in Data
One of the biggest indicators of somebody’s next behavior is their past behavior.
Being able to mine data to find behaviors and patterns of behaviors allow the creation of models that are predictive of bad data, or indicative of an expected outcome.
“I think there’s an exorbitant interest rate on bad data once it gets into your systems, and the longer it sits there, the more it costs you.”
The most useable solution is an IP Blacklist, and this is shared across all customers.
Fascinating Data Insight
Charlie shares an interesting tidbit that has been gleaned from Imperium’s huge repository of data: From a contextual nature, the words “hope” and “joy” are two of the most popular words used.
Quick links to connect with Charlie Allieri: