Errors occur in spite of careful study design, conduct, and implementation of error-prevention strategies. Research For Good’s best practices, outlined below, identify, correct and minimize any errors or impact on study results.
Machine Learning & Artificial Intelligence
Fraudulent cyber machines can generate “clean” versions of a device system as a way to bypass digital fingerprinting. With machine learning and artificial intelligence, we analyze the patterns in attributes like locations/ip addresses/accounts of known fraudsters to create a set of identifying rules that our system follows. Using these rules, and machine intuition, we can identify if the transaction will be high-risk or not. What’s great is that the more data linked with the A.I. the more it will be able to do.
Device Recognition & Linked Accounts
Devices have thousands of attributes linked to them. Things like IP addresses and Apps all combine to make a unique code for that device. With device recognition, we are able to track this specific device and should that device log into an account associated with fraud, it gets flagged. On a deeper level, we utilize hidden connections between devices and accounts so that any device associated with an already flagged fraudulent account gets referenced against a database containing over 5 billion device codes to ensure only the most trusted user can access our system.
Digital Fingerprinting enables us to create a unique identity for each computer. Using this method, no single computer (regardless of how many people may share that computer) is allowed to enter any single survey more than once. The Digital Fingerprint itself is a proprietary combination of many characteristics of the user’s browser and computer. This creates a statistically large basis set of fingerprints, sufficiently larger than any single survey would ever require. Digital Fingerprinting effectively eliminates the possibility of a survey being taken twice on the same computer, even if its IP address changes and even if the cookies/cache have been cleared.
Our proprietary SecureGeoIP (TM) technology is based on a proprietary algorithm that uses IP Geo-location to evaluate the truthfulness of a respondent’s declared postal code. It effectively cross-references IP & Postal code distances from one another. Thorough and in-depth analysis has indicated that this measure is an excellent predictor of fraud, while still allowing legitimate, quality respondents through to client surveys. The addition of SecureGeoIPTM to our arsenal of quality control measures, therefore, results in a higher level of data quality without sacrificing response and conversion rates through false positives.
I am not a robot. Sounds simple enough, but as we know, it can be hard to detect ever more sophisticated bots attempting to defraud surveys. To combat this, we are using very advanced captcha technology.
Active Detection of Connections from Proxy Servers
While the known proxy server list is large and always being updated, new proxy servers are always being turned on as fast as proxy server lists can be updated. Fraudulent entities may also be running their own proxy servers or using VPNs to hide their identity and location. RFG work with the leaders in detecting fake proxies and ensures that the IP of every respondent is analyzed and scanned for the presence of a proxy server. Any respondents using proxy servers are blocked from entering surveys.
We use fingerprinting, IP and other proprietary methods to guarantee that no respondent will be able to attempt the same survey more than once.
IP Authentication Through Known Suspect Online Activity Database
Research For Good authenticates all respondents through a database of known suspect online activity IP addresses and blocks those who do not meet our stringent threshold of quality. This allows us to prevent users with a history of suspicious activity from participating in our surveys, and we can take advantage of shared history to prevent these people from entering our client’s survey the very first time we see them, and not having to wait for undesirable behavior within our own system.
Suggested Additional Data Quality Measures
In addition to the measures Research For Good has implemented, we also suggest that clients continue to take their own steps to further improve survey data quality. Some strategies may include:
- Actively running your own digital fingerprinting technology on all surveys
- Any surveys showing respondents proprietary and confidential information (movie trailers, images, product feature descriptions) should include client’s Privacy and Terms & Conditions verbiage within the survey itself and require respondents to agree.
Research For Good continues to proactively enhance our security measures to consistently offer the highest quality online sample possible.
For more information, please contact us.