What is maxdiff analysis?

A guide to maxdiff analysis

Comparison of features

Prediction market interface and method

For respondents In a chatbot-like interface, ask people about their gender and age first. Display points at the top of the screen (round down to 2 decimal places) Intro text: “You have been allocated 1000 points that you can use to bet in this survey.” “We will ask you a series of “Yes or No” or “A or B” questions. You can select either of the two choices and bet your points on your answer.” “If your response is correct, you will keep your points and also win prize points.” “If your response is incorrect, you will lose the points that you have bet on this question.” “We will also show you how other participants responded before you. The odds (i.e. how many points you get if you win) will be based on their previous responses.” “Here is the first question:” First question: Question: “This question is about the soda brands that are sold at supermarkets in Australia” “Does Coke (including Coke Zero and special flavours of Coke, such as Coke Vanilla) have more than 50% of the shelf space in supermarkets across the whole country” “Here is what other people bet on this question:” Yes: 95% of bets, No: 5% of bets “The odds for “Yes” are 19:1. It means if you bet 10 points on “Yes” and you happen to be correct, you keep your 10 points and also win 190 points. If you bet on “Yes” and happen to be wrong, you lose your 10-point bet.” “The odds for “No” are 0.05:1. If you bet 10 point on “No”, you can win 0.5 points (and keep your 10-point bet).” “So, back to the question:” “Does Coke (including Coke Zero and special flavours of Coke, such as Coke Vanilla) have more than 50% of the shelf space across the whole country?” Interface: Show options at the bottom: “Yes” - “No” - “I don’t know” Once clicked say “You decided to bet on “Yes” “How many points will you bet?” Value from 0 to number of points they have now “Okay, you placed 4 points on “Yes”. If you are correct, you can win 76 extra points. If you are wrong, you will lose your 4 points. “You were wrong and lost the 4 points” - update the counter “Congratulation, you are right. You won 76 extra points.” Show
Second and third questions: Question: “Time for a new question” “This question is about the soda brands that are sold at supermarkets in Australia” “Is Coke Vanilla sold at the majority of supermarkets in Australia?” “Here is what other people bet on this question:” Yes: 30% of bets, No: 70% of bets “The odds for “Yes” are 2.33:1. It means you can win 2.33 if you place 1 point on Yes (and keep the 1 point you bet).” “The odds for “No” are 0.42:1. It means you can win 0.42 if you place 1 point on No (and keep the 1 point you bet).” Show interface “The previous questions related to the present and the past. You will now be asked some questions relating to the future. You will be able to place hypothetical bets.” Question about the future: Question text: “Please consider the following product:” Show picture “Will this product be appealing to young people in Australia?” “Here is what other people bet on this question:” Yes: 30% of bets, No: 70% of bets “The odds for “Yes” are 2.33:1. The odds for “No” are 0.42:1.” Past comments: “Someone who said “Yes”, explained their decision with the following comment:” “” “Other who said “No”, explained their decision with the following comment:” “” “What would you say?” Show options at the bottom: “Yes” - “No” - “I don’t know” Once clicked (unless I don’t know) “You decided to bet on “Yes” “How much would you bet of your points on it?” Value from 0 to number of points they have now Can you please provide a short comment to explain why you bet on “Yes”? “OK thank you” Three times: “Someone who said “Yes”, explained their decision with the following comment:” “” Do you think it is a valid reason for saying “Yes”? Yes - NoOther questions. FinishSet up for the users Calibrating questions In each question: Question text Correct option text Incorrect option text Library of questions. Future-oriented questions Add a library of questions For each question: Question type Question text Option 1 / Option for success Option 2 / Option for failure Template: all text to be replaceable Only chatbot interface Settings: How many calibrating questions to supply Order of calibrating questions: As specified / Random How many prediction questions to supply Order of calibrating questions: As specified / Random Collect comments: no, yes,collect only, yes and let respondents see others’ comments (to be convinced and rank them)Reports for the users Diagnostic for calibrating questions: For each question: x% of bets on yes/no % of better on yes/no. Average bet on yes/no Summary diagnostic for correct prediction For the prediction questions For each question: x% of bets on yes/no Bias-adjusted percentage % of better on yes/no. Average bet on yes/no Top 5 reasons for yes / no Table of all reasons and bets