The most common measures of central tendency are the (arithmetic) mean AM,

(geometric) mean GM, (harmonic) mean HM, median and mode. Their choice should depend on their characteristics and the data. The most commonly used is the AM. The reason is that all data items are equally represented in AM. However it is easily influenced by outliers. Of the means the least influenced by outliers is the HM. By virtue of its definition, the GM may be chosen within the context of a multiplicative rather than an additive model. Moreover where open classes are involved none of the means is determinable. In that case the median or mode is appropriate. Also the mode and median are not affected by extreme values. In nonparametric applications the median is usually chosen in preference to any other. In opinion surveys as for electioneering the mode is the natural choice.

It totally depends on "which type of data you have and what you want to extract" from them. All measures of central tendency (Mean, Median and Mode) have both its pros and cons. Only by observing the nature of data suitable measures can be applied.

There's a lot of discussion about that on this site.

See if this will work for you:

https://www.google.com/search?q=mean+median+mode+site%3Aresearchgate.net%2Fpost

https://www.google.com/search?q=central+tendency+site%3Aresearchgate.net%2Fpost

There is no proper answer to this question till this date.

This is a challenging problem for the researchers.