If the context I was searching this word in dealt with numeric data, then my term of choice would be "statistic", defined (by a Google search) as:
A fact or piece of data obtained from a study of a large quantity of numerical data.
And by Wikitionary as:
A quantity calculated from the data in a sample, which characterises an important aspect in the sample
It's different from "summary" or a "gist" because a summary is expected to have an brief overview of all things in the original data, but a statistic is very lossy (for example if you have the statistic median then you don't know what values lie before or after median, so it is not a summarization of data).
Thus a desirable property for this word is that it isn't expected to have complete (not in terms of depth, i.e. details but in terms of breadth) information about the data.
And a gist is supposed to capture the essence of something, but here there may be multiple statistics, each serving a different purpose (e.g. the median which tells you the middle value of the data if you sort it, or the mode which tells you the value which occurs with highest frequency)
And it's different from "property" or "attribute" because they also refer to meta-things about data, for example the data is "clean" may be a property, it is "preprocessed" may be another property.
So another desirable property of this word is that it is directly the consequence of information of the data (without regard to other things like form)
To give an example of non-numerical data, suppose that your data is text for a job advertisement. Then things like "target demographic", "salary", "location", "job title" may together form a summary of the advert, but individually they are ___________ of the data.