MaxDiff is a proprietary software message testing program (Sawtooth software) that works by exposing respondents to multiple combinations of messages tested in groups of 4. For each group, the respondent picks a #1 “favorite” and a #4 “least favorite”. The multiple exposure iteration process of choice produces two Statistics:

  • UTILITY SCORE which is the raw score, rank ordered, of each individual message’s power.  The combination of Utility Scores for all messages tested equals 100%and,
  • REACH which quantifies the % of the electorate (sample) positively impacted by each message with a cumulative tally.  For example, message A will have a 55% “reach” (meaning it works to persuade 55% of the electorate), message A+B will have a 70% “reach”; A+B+C will have a 90% reach and A+B+C+D will be 95%.  Therefore, you learn not just what is the most popular message that works for most voters, but what combination of 2 or 3 messages works for nearly ALL voters.

Output on “reach” to get close to 100% rarely exceeds 4 messages… often 3 messages carry most of the “reach” load.

MaxDiff is superior to traditional polling message testing that is really just one dimensional… a rank ordering of how well each message scores on some one-dimensional measure (1 to 5 scale or Very strong, Swht strong, Weak index). Too often, there is little differentiation between individual message scores and/or respondents simply become fatigued at answering a long battery of similar questions.

MaxDiff forces a choice between messages and does so with enough multiple iterations as to create literally thousands of data points… a much more sophisticated methodology.

As with standard poll testing of messages, MaxDiff scores can be cross-tabulated by subgroups… Independents, soft Republicans, swing voters, senior women, etc.