Data, Analytics, Modeling, Polling
December 18, 2020
Polling and modeling are both important tools in the world of data-driven campaigns. While polling and modeling seem similar in nature, there are some key differences, both in methodology and in conclusions, that can help determine which tool is best suited to specific situations.
Please welcome Causeway Solutions’ Chief Analytics Officer Molly Rutledge to the Majority Strategies #StrategySession as our guest blogger.
Polling and modeling are both important tools in the world of data-driven campaigns.
While polling and modeling seem similar in nature, there are some key differences, both in methodology and in conclusions, that can help determine which tool is best suited to specific situations.
Polling is a thermometer.
Polling allows you to take the temperature of a population at any given point in time.
Polls are typically conducted using a maximum sample size of 1,000 people, which produces a margin of error of plus-or-minus 3 percent.
Depending on the length of the script, a poll can be a very cost-effective way to gauge where a population stands on specific issues. Results can be tracked over time to determine if attitudes of a population are moving one way or another.
The output of a poll is typically toplines and some crosstabs that allow conclusions to be drawn about groups of the population.
Modeling is a thermostat.
In contrast, modeling is a thermostat, monitoring the temperature of a population. Modeling also prescribes actions that can be taken to help move that population.
Modeling projects can be conducted with a sample size anywhere between 2,000 and 10,000 people, which shrinks the margin of error to below plus-or-minus 2 percent. This reduction in the margin of error produces more statistically reliable results that can be used to build analytical models on an entire population.
The output of a model is typically a score for each individual in the universe for every issue modeled, meaning that conclusions can be drawn about specific individuals in the population.
Models can be tracked over time to determine if specific individuals are moving one way or another. Because of the more in-depth nature of modeling, the modeling process can take longer and cost more than the polling process.
Both polling and modeling have a place in data-driven campaigns. Polls are excellent for message testing, checking name ID, and getting a quick snapshot look at where the population stands, but polls cannot prescribe any action, don’t identify exactly which individuals need communication, and don’t yield the depth of insight that a model does.
Models are great for detailed insights about individual people, and they can prescribe a course of action that enables a campaign to reach its goals, but they are not great at message testing, aren’t useful if there is no name ID, and might be overkill if all you need is a quick snapshot of the population.