Half the group was asked if they thought a person was above 165cm tall, which he was.
The other group was asked if they thought the same person was below 190cm, which he also was.
Then after, both groups were asked to guess his height. The second group guessed significantly higher than the first. Since both groups were told a number first, their response got anchored/biased towards that value.
If you present people with several options, they'll be biased towards selecting options in the middle of the range. This is particularly relevant to SaaS startups and pricing plans; but in the article the author was using it to steer the decision maker towards the option he had already determined to be optimal.
The optimal cutoff would probably have been $20 million rather than the $10 million that was chosen, but that change would have been too much for a lot of people. If I had only presented options from $1M up to $20M, then the group probably would have chosen $5M no matter how much I argued for $20M. I had to put the $50M number in people's heads so $10M seemed less of a jump.
Ya that's kind of what it seems like you did... throwing a big number out on purpose so that $9M jump would be "rational." The math in my head says you could weight figures on estimated acceptance probability. By throwing out a low probability figure first before a relatively higher probability acceptance figure - versus just presenting them with a reasonably high probability figure - give a better result. It is almost like expected value.
Half the group was asked if they thought a person was above 165cm tall, which he was.
The other group was asked if they thought the same person was below 190cm, which he also was.
Then after, both groups were asked to guess his height. The second group guessed significantly higher than the first. Since both groups were told a number first, their response got anchored/biased towards that value.