How does Latana prevent false data?

Providing very high quality data that is precise and representative of the real world is central to Latana’s methodology. While our non-incentivised sampling method already mitigates common survey biases and naturally protects against fraud, there are still some behaviours in our surveys that can produce false results, such as accidental misclicking or not being attentive.

We have developed a process that identifies such behaviours with a high level of precision which allows us to remove those respondents from the sample. Each of our respondents is assigned a “Quality score” which grades respondents from 0 to 1. This score is assigned by a machine-learning-trained model which takes into account indicators like the respondents’ speed, positive/negative answer patterns, screen touch patterns and answers to trap questions. Respondents below a healthy score are cleaned from the pool while the higher quality respondents remain.

Respondents who have previously received a low quality score are blocked from participating in our surveys again and we block respondents from participating in the same tracker for 6 months to maximise the amount of unique respondents in our sample and avoid respondents answering differently because of having previously seen the same survey.