EKOS employs a variety of research methodologies, including traditional telephone surveys (using
random digit dialing – RDD),
interactive voice response (IVR) technology, and its proprietary hybrid online/telephone research panel, Probit. EKOS utilizes IVR technology for political polling, which uses automated phone systems rather than live interviewers. In recent years, EKOS has emphasized its hybrid methodology (Probit), which combines random digit dialing telephone recruitment with online panel participation. This approach aims to provide exhaustive coverage of the population (Internet, landline, cell phone) using probability-based sampling. In 2018, EKOS introduced refinements to its methodology, including enhanced cell phone sampling and more sophisticated weighting techniques to better represent younger demographics and mobile-only households.
Advanced research methodologies EKOS utilizes its Probit panel, described as a hybrid online/telephone research panel constructed using probability-based methods. Recruitment for the panel is conducted randomly via telephone (using random digit dialing for both landline and cell phones), and respondents participate either online or by phone. This hybrid approach aims to provide exhaustive coverage of the Canadian population and support margin of error estimates, unlike non-probability opt-in panels. These adjustments typically account for demographic variables including age, gender, region, and educational attainment to ensure the sample's composition reflects population benchmarks, often based on Census data. EKOS has noted that weighting by education can be appropriate when response rates appear correlated with this variable, though acknowledging the complexities involved in comparing survey samples to Census data on education due to measurement and coverage factors. In its qualitative research, EKOS employs methods such as focus groups, key informant interviews, ethnographic research, and
content analysis to gain deeper understanding of public attitudes and motivations. The available sources do not specify whether their content analysis techniques involve computational methods like
natural language processing (NLP) or proprietary algorithms. == Polling accuracy ==