Just came to Rootschat to ask the very same question, now I've upgraded my FindMyPast subscription.
For previous censuses (and the 1939 register), FindMyPast (and Ancestry in a slightly less committed fashion) have included a full reference, that's nice and easy to copy and paste into one's preferred family tree software.
E.g.
1901: RG13/2579; Fol. 13; p. 18
1911: RG14PN16407 RG78PN1017 RD359 SD2 ED6 SN245
1939: Ref: RG101/1345C/028/3 Letter Code: CMRC
but on the first 1921 entry that I've looked at, I've got the fields previously mentioned by manukarik, plus
Archive series: RG 15
Piece number: 23338
Schedule number: 165
Schedule type code: E
District reference: RD 512 RS 3 ED 8
Replacing the %2F characters in FindMyPast's URL with the / that it stands for, I get what's presumably FindMyPast's own reference to this schedule:
GBC/1921/RG15/23338/0331/01
RG15/23338 is self evident, but that 0331 doesn't appear on the transcript, so presumably not part of the TNA reference, so would only work on FindMyPast to drill down to a particular schedule.
I would guess that the a reference RG15/23338/165 would be sufficient to point to a particular schedule. Or would I also need to include the district references too?
How are other rootschatters referring to 1921 census schedules?