It's OK to cite a figure around 30% if we want to indicate the "effect size". A figure of 75% is only correct to indicate the percentage of population affected by this bias.
However care should be taken because there are moderating variables that can easily negate the effect, and thus the effect should not be cited as general rule.
Full discussion
Asch's conformity experiment is very well regarded, and has thousands of citations. However, Asch made many such studies and their results were also replicated hundreds of times, with different results.
Wikipedia has a more nuanced view on the matter than you report:
In the confederate condition also, the majority of participants’ responses remained correct (63.2 per cent), but a sizable minority of responses conformed to the confederate (incorrect) answer (36.8 per cent). The responses revealed strong individual differences: Only 5 percent of participants were always swayed by the crowd. 25 percent of the sample consistently defied majority opinion, with the rest conforming on some trials. An examination of all critical trials in the experimental group revealed that one-third of all responses were incorrect. These incorrect responses often matched the incorrect response of the majority group (i.e., confederates). Overall, 75% of participants gave at least one incorrect answer out of the 12 critical trials.
This actually contains the 37% figure mentioned on youtube. It seems to be merely the case of two different measurements: 75% got the wrong answer at least once over multiple experiments, but on average 37% of participants were swayed on each experiment.
The article linked by the Wikipedia in question is Asch, S.E. (1951). Effects of group pressure on the modification and distortion of judgments. In H. Guetzkow (Ed.), Groups, leadership and men(pp. 177–190). Pittsburgh, PA:Carnegie Press. It does not really contain the figures mentioned, but they are contained in other literature found below.
I can't comment on the specific 29% figure because the link you gave originally is dead, I can't find the study or book available anywhere (Pubmed, Google Scholar, OUP or Prentice Hall) and we need more context.
However a nice meta study shows that many subsequent effects found a statistically significant swaying effect of around 30%. The big caveat is that there are tons of moderator variables (majority size, gender composition of the majority, cultural norms in the country of experiment, etc.).
