I was watching an online video on statistics and in the video we were making inferences on a real estate companies data, to predict our future customers. The instructor had all the previous data of the company present with him, but he said that because we want to make an inference about our future customer, we need to use sample formulas instead of the population ones. So is it that whenever we are trying to predict something, we always consider the data to be a sample, and never population? If yes, is there a time when we make use of the population formulas?
In the specific case you describe, the real estate companie data is probably a sample of all transactions that occurs. So the data at hand is in fact not exhaustive.
In other cases, it is however true that data analysts tend to apply formula designed to samples also in some the case when we have access to the full universe. (to account for other sources of uncertainty than sampling) This may be changing "soon".
For instance, regarding the computation of standard errors, this exact question is the theme of a (just-published) Econometrica article by some of the most prominent researchers of the field:
Abadie, A., Athey, S., Imbens, G. W., & Wooldridge, J. M. (2020). Sampling‐Based versus Design‐Based Uncertainty in Regression Analysis. Econometrica, 88(1), 265-296. https://onlinelibrary.wiley.com/doi/full/10.3982/ECTA12675