Election polling data is like weather forecasts: everyone knows they're probably wrong, but we listen to them anyway.
Actually that analogy is off -- weather forecasts are sometimes accurate. At various points along the way pollsters have had Ted Cruz over Donald Trump, and vice versa, and Bernie Sanders over Hilary Clinton, and vice versa... compare polling results to actual results and and the difference is often great.
And That's why fellow Inc. columnist Dana Severson and his colleagues at Promoter decided to try a different approach. Promoter measures brand sentiment for companies using Net Promoter Score and figured out a way to do the same for presidential candidates, launching the site Net Presidential Score.
(If you aren't familiar, NPS is tool that can be used to gauge the loyalty of a firm's customer relationships; it's an alternative to traditional customer satisfaction research that claims to correlate with revenue growth.)
Their goal was to use their technology to disrupt a broken political polling system and give the average voter a voice without media biases.
How did they do that?
Predicting the election using NPS® ... sounds crazy, right? We actually think it's the most sane and reasonable option that exists today.
But, before we get to how we're doing it, let's discuss why.
About two weeks ago, six of us at Promoter were having a discussion about politics. More specifically, about the current race for POTUS. We learned that, amazingly, each of us had a different choice of candidate. We each had our own valid argument as to why we felt the way we did, and none of the reasoning had to do with party affiliation.
Our discussion just happened to occur on the afternoon of the second Super Tuesday. So, naturally, we found ourselves looking at the polling numbers to get an early indication of whose candidate was ahead.
Looking over the variety of polls that day, we came away with the sense that they were terribly inaccurate, likely biased and way out of touch with what seemed to be the national sentiment.
In other words, we lost faith in the data.
After spending many hours digging into how today's polling data is tabulated, we came away with several reasons why current polling is failing. Here are just a handful of examples:
- Many polls are still done over the phone, finding random participants by using an actual phone book.
- That approach obviously favors an older generation who still maintain a home phone with a publicly listed number.
- What's more, these are largely done with automated recordings and can last over 10 minutes.Needless to say, all of this has resulted in smaller subsets and skewed results.
- Voters are given cash-based incentives. To make up for the lower number of participants, pollsters have started to provide incentive to voters for completing the questionnaire. As we've mentioned previously, when you pay someone to complete a survey, you change their motivation. It becomes less about accuracy and more about efficiency.
- Some polls include up to 15 questions. As we've noted numerous times, each question you ask someone beyond two reduces response rate by 50%. This is how you end up with "national" polls with only a few hundred responses.
- Some polls are thinly disguised propaganda. Many polls ask leading questions, followed by a candidate question that most closely aligns with the suspected answer to the first question.
- Polls are binary, giving you only the option to say yes or no. Party A or party B. Candidate 1 or candidate 2. The truth is, most people have a middle, especially this year.
- And lastly, there's the corruption. Not all of them, but all it takes is one untrustworthy egg to ruin the confidence in the rest of them. Such is the case with mega pollster, Frank Luntz, who has openly admitted to taking money to help shape Marco Rubio's career.
Overall, we realized that current polling methods are tired and don't reflect the new political shakeup that we're seeing in this election cycle.
With that thought in mind, we decided to fix it.
Our company, Promoter.io, has been using the NPS methodology to help companies measure customer/brand sentiment with incredible accuracy for nearly two years... so why couldn't the same be done with Presidential candidates?
Could we take a proven indicator of brand sentiment and attach it to humans? After all, just like any product, each of these candidates have features (policies and beliefs) that voters have opinions on. And, just like Apple can use their NPS score to measure their favorability as a KPI, we believe candidates, as a brand, can do the same.
So, after approximately 120 hours of development, we're proud to introduce you to Net Presidential Score®, which we define as the most accurate and unbiased Presidential scoring system ever created.
Here's How it Works
We're using the methodology behind NPS (Net Promoter System) to measure the human sentiment of each candidate individually. We've randomized the order and removed party affiliations.
If you're familiar with NPS, you know the basis of the survey is deceptively simple, which is partially the reason for its success. We're taking each candidate individually, not matched up against each other and/or their party affiliation and asking two questions for each:
- How likely are you to recommend Donald Trump as a presidential candidate to a friend or colleague?
- What is the most important reason for your score?
For the first question, participants are given a scale between 0 (not at all likely) to 10 (extremely likely). Depending on their score for each candidate, they'll be defined as either a detractor (0-6), a passive (7-8) or a promoter (9-10).
Each participant is asked to consider each candidate individually and consider all factors when making a decision. Just because someone is unlikely to vote for any particular candidate doesn't mean that they'll automatically receive a 0.
Each candidate will then be given a Net Presidential Score (between -100 to 100), which is based on the percentage of their promoters minus the percentage of their detractors, removing the influence of the "fence sitters" who have identified themselves as passive supporters or unknowns.
What's interesting is that each of those profiles (promoters, passives and detractors) have intrinsic characteristics which are indicators of future momentum or decline. For example, we know based on historic product measurement that 20% of promoters are those that are actively supporting/endorsing. As the percentage increases, we can make assumptions around an increase in ground-based momentum.
What makes this system super-interesting is the second question. This is open-ended feedback and provides context and reasoning for each score provided. Each candidate has a feedback tab, where all discussion around scoring occurs and can be viewed by anyone.
And lastly, we're authenticating all submissions through Facebook which allows us to not only control polling fraud, but breakdown scoring on any number of geo and demographic factors in the coming weeks.
As far as we are aware, this is the first time that NPS has been used to measure election results (at this scale) and we're excited that you'll be among the first to use it. We believe this is the right indicator at the right time that most accurately reflects the sentiment of the nation.
We believe that @NetPrezScore is the indicator that most accurately reflects the sentiment of the nation.
That said, we're not pollsters and we're not pretending to be. There are a lot of factors go into an election aside from pure general population sentiment. While we're 100% confident in our technology and the power of NPS, outcomes can still vary due to unpredictable factors.
Come take the Net Presidential Score survey (see what we did there with the initials?) for a spin, share your thoughts on the candidates and be a part of this NPS-first.