Behind the Headline: Fraud Detection Methods
Blog | February 27, 2024
By Terence McCarron
Can Survey Fraud Detection be enhanced by Facial Recognition?
OpinionRoute breaks down headlines from industry news that relate to online surveys, online sample, data quality and relevant topics through our lens into the ecosystem.
Headline: KANTAR AND REALEYES PARTNER ON ANTI-SURVEY FRAUD TOOL
Kantar recently announced a new partnership to “Fix survey fraud”. These reports share findings of a validation done on Kantar’s panels and third-party partners. The results argue the tool is effective in preventing a couple of core things- bots and duplicates.
We continue to love all things Data Quality, so we dug in to see if this could be a nice add-on to our approach. Here are a few points to consider when you contemplate this more headline more deeply.
Here are some Pro-Tips to break all this down.
- Facial Recognition Requires Consent of the Respondent & Sample Vendor.
Using a facial recognition feature requires informed consent for privacy compliance. This would require notifying the respondent before it’s run (likely as a prescreener) and then getting permission.
Over the years, consent for facial recognition has dropped significantly. As a result, vendors often resist these approaches.
- Let’s be clear on what fraud is and isn’t in this context.
By OpinionRoute’s analysis of surveys across hundreds of research agencies, bots are a minimal part of the current fraud problem in MRX. In fact, combatting bots is one of many industry joint successes over the last few years. Duplicates are a larger issue, consistently representing between 2-10% on any given project. However, this figure is tied closely to the vendor blend chosen by the researcher. Certain companies produce more dupes than others.
- Data Quality solutions should calibrate against researcher expertise, not vendor expectations.
At OpinionRoute, we do not believe quality performance should be measured against our own definition of quality.
Instead, we believe researchers should calibrate all quality systems against expert survey data reviews. In other words, if you invest in a fraud system that doesn’t meaningfully improve your cleaning outcomes, then is it really effective?
- Fraud Detection should not be selectively applied.
If you give the respondent a chance to opt out of fraud prevention, won’t the bad guys say no too?
So does it work?
We love constant advancements in authenticating respondents. But in this case, relying too heavily on “selective” facial recognition technology creates huge gaps in your projects’ fraud and dupe mitigation and results often do not align with whether the survey data is quality. We do however like this idea for panel registration validation.
Don’t just take my word for it…. Contact us now if you’re interested in reading the summary of findings from our recent test study.