deal with unreliable data points

[kurofufu]

This is a Search Engine Marketing question. Suppose I bid 100K keywords in Google. 70K of them have no conversion and many of these 70K have a very low number of clicks(<50). The average conversion rate is 1%(i.e. one conversion per 100 clicks). For each keyword, I have a data point of its conversion rate, # of clicks and profit along with many other data elements for one year period. I would like to build a model to predict the profit per click for each keyword. I would like to know if I should use 100K data points or should cut off those data points with a low number of clicks. If latter, then what the cut off value should I use?