top of page

Our Data-driven Polling

In an era where polling is so heavily scrutinized by the media and political insiders, it takes a very careful and precise approach to rise above the rest.  At Ascend Action, our elite team of analysts and data scientists have fully modernized the research process.  Incorporating predictive modeling into the process, we can guarantee our clients accurate results with a methodology designed for a whole new generation of voters.

We have also designed a whole new suite of research and polling products with competative pricing for smaller campaigns.  Advanced polling and data tools shouldn't only be reserved for multi-million dollar races.  At Ascend, we've made it our mission to bring high value tools to campaigns at every level, from state legislature to presidential.


In the 2016 and 2020 presidential elections, pollsters were hammered for their inaccurate election predictions.  This was mostly caused by outdated sampling methodologies.  During this same time, Ascend was well within the margin of error on every poll conducted for our clients.


We use a proportionate stratification technique that allows us to select a representative sample most closely mirroring the demographic makeup of the voting population. Instead of dialing through a true random sample of likely voters, we separate voters into sub-groups (age, gender, geo, etc.) and select a proportionate sub-sample from each.  This technique, paired with quota-based dialing, significantly reduces or eliminates post survey weighting – increasing accuracy and reliability.



We select the ”likely voters” for our sample based on our predicted turnout models (when available) instead of vote history.  Ascend invests heavily in predictive voter modeling – regularly scoring voters on their likelihood to turn out in each election.



Traditional pollsters only stratify the overall survey results, meaning only the topline results are demographically proportionate.  This could cause your overall results to be accurate while still being 3-4% in each geographic region when analyzing sub-groups.  At Ascend, we use a dynamic mode of stratification that changes the weighting quotas for each geo individually.  





The traditional mode is to collect a random sampling of registered “likely voters” selected based on exclusively vote history.  This sampling method does not account for outlying factors and behavioral trends that can have a significant effect on who turns out in each election.


If your sample is demographically proportionate but includes voters who will not turn out, results will be undoubtedly skewed.  True random sampling also relies heavily on post survey weighting to achieve proportionality – a mathematical control that can significantly reduce accuracy and reliability.

bottom of page