What is point-in-time data?
Point-in-time data is financial data exactly as it looked on a given date, rather than as it looks today. If a company later restated a figure, a point-in-time query for the original date still returns the original number — the restatement is walled off behind its own date.
Most financial databases store one current value per figure. When a company files a correction, the old value is overwritten, and the history quietly becomes a description of the present rather than a record of the past. This is fine for looking up a number and fatal for testing a decision, because it means every historical query is answered with information the decision-maker did not have.
A point-in-time store keeps every vintage of every figure alongside the timestamp at which it became public. That timestamp is the load-bearing part: it is what lets a query say "as of 14 March 2019" and get an answer that does not depend on anything filed afterwards.
The practical test is simple. Ask a provider what a company reported for a quarter, as of a date shortly after that quarter closed, for a company that later restated. If the answer matches today's restated figure, the data is not point-in-time, whatever the marketing says.