What data do you collect?
You are talking about a source crowd, and therefore (I assume) aggregating data for that crowd. Since they talk about the products they buy, I suspect that you are going to use the attributes and prices of the product.
Some possible approaches. If users enter non-numerical data (e.g. colors), simply record the most common entries or the mode (most frequently entered).
If they enter numerical data, cancel outliers. those. the lowest and highest results, and on average the rest (you can do it at prices, say, this is the approach that electronic exchanges use to solve the closing price of many transactions).
Depending on your application, you may want to have a historical bias regarding recent entries.
But it all depends on your application and on the amount of storage and crackling data that you are ready to do.
source share