Our Quality Program
You can’t make sound business decisions without solid data backing them up. Your information must be properly sourced, accurately reported, and closely reviewed before you act. To ensure the information we gather is both reliable and actionable, Harmon Research employs a robust quality assurance program called DataDefense that reviews our data before, during, and after the survey. From the moment you propose your survey project to our delivery of final data, DataDefense ensures we never lose our focus on seeking, gathering, and analyzing high-quality data.
1. Before the survey
For our own internal panels and our partners:
We confirm the identity of each panelist and only allow each member to have one profile. They must complete a double opt-in process when they sign up.We identify and deduplicate respondents based on combination of geo-IP address, cookie Identifiers, device Identifiers and duplicate fingerprint.
We have additional security checks such as:
- User Emulator
- VPN usage detected
- TOR network detected
- Public or web proxy server detected
- Web crawler usage detected
- Internet fraudster detected
- Recent Abuse detected
2. During the survey
- We employ gate-keeping tactics to filter out bots and low-quality responses.
- Any respondent who can’t pass a CAPTCHA test is immediately terminated from the study.
- We also have safe guards such as time checks, attention checks, IP geo-checks and cut and paste blocking.
- We do quota monitoring to ensure completes match invites and bot farms are not attacking.
3. After the survey
Once the survey is completed, and we’ve collected the responses. One of our researchers conducts a manual, line-by-line review of the data as a final check to make sure your analysis is based solely on quality answers.
Our experts examine the responses to ensure that:
Our experts examine the responses to ensure that:
- Each open end is valid.
- Someone who writes open-ended nonsensical answers like “good” or “I like it” is eliminated.
- They address the survey questions and are of an appropriate length.
- Respondents are in the targeted region.
- Duplicate responses or respondents are removed.
- Perfect open ends (written by AI) are removed