The Bayesian Adjusted Social Network (BASON) Survey successfully solves the biggest issue in polling – it eliminates bias from respondents’ answers.
It answers the following questions: “Will people buy this product?”; “Who is my customer?”; “What value statement works best for which customer segment?”; “How do we reach our consumers”; “Who will win an election?”; “What’s the optimal price for my product/service?”
We analyze the social networks of customers (or voters) in order to draw inferences from these networks and predict outcomes with amazing precision. Our BASON survey is the only poll in the world that successfully predicted both Brexit and the victory of Donald Trump.
It is the best prediction tool available on the market, and it represents a significant improvement over regular surveys in both accuracy and information gained. Combining it with our machine learning software yields unprecedented precision in figuring out what the customers really want and why.
How does it work?
- Combines a wisdom of crowds (WoC) approach to polling adjusted for the network analysis of groupthink bias
- The idea is to incorporate wider influences, including peer groups that shape an individual’s choice on voting day. Which is why the method is perfectly suitable for social networks
- BASON (and WoC in general) does not require a representative sample to make a good prediction. It goes beyond representativeness, self-selection problems and random sampling, and focuses simply on trying to find out how people estimate their local conditions and sentiments
- Survey participants log in via Facebook or Twitter
Part 1 – Wisdom of crowds
We ask our participants three basic questions:
1. to express their voting preference (e.g. Clinton vs Trump)
2. how much do they think their preferred choice will get (in percentages) in their state/region
3. and how much do they think other people in their state/region will estimate that Clinton or Trump could get
Part 2 – Uncovering bubbles
- The key goal of the network analysis is to reduce individual-level bias (confirmation bias) of each survey respondent
- We look for clustering within groups based on expressed preferences (voting, products, prices, etc.)
- The enables us to recognize echo chambers, members of which are less likely to be good forecasters
Advantage over competition
- Proven accuracy in predicting election outcomes and consumer sentiment
- Predictions up to 4 times more accurate than other pollsters
2. Fast execution
- Usually 3 weeks of surveying, 2 weeks of prep work, 1 week for analysis (can be adjusted to what the client needs)
- Can be used continuously before and after a campaign to test accuracy
3. Large samples
- Generated through viral traction online
- Using an interactive game/app to attract users into the survey; answering desired questions as a part of the game
- Main advantage of the approach is that it can deliver accurate outcomes even with small samples (between 500 to 1000 people)
4. Network analysis
- The key value added for scientific research surveys
- Oraclum is the only company on the market that does a network analysis as part of its surveys