https://www.youtube.com/watch?v=EHl2tU40ibs In this Election night coverage (link duly posted above) it is said that different outlets "count in different ways" and that different outlets have different counts. How is this so? (Timestamp:- 2:50:08 to 2:50:52)
There are the official numbers. Their advantage is that they are official, their disadvantage is that they take a long time to get assembled.
Then there are the unofficial numbers from various news agencies. They could try to beat the official tally by simply counting faster, and by taking more risk with less careful double-checking that nothing is scrambled in transmission.
Or they rely on incomplete results and calculate predictions on that basis. There have to be adjustments, or qualifiers in the reporting based on past results.
"With 2 out of 3 districts reporting, X has 1,001 votes and Y has 999 votes. But Y has traditionally been strong in the missing district, so X has little chance to win."
"District whatever is reporting a victory for X by 5%. Traditionally X has received 10% more in whatever district than statewide, so we predict that X will trail by 5% overall and that Y will win when all results are in."
If the pollsters are honest those adjustment factors are based on data, but there are judgement calls in that. How long does one go back? How do you deal with demographic shifts? And so on.
Interestingly, you see some of that in Corona reporting as well. The famous Johns Hopkins numbers tend to be higher than official numbers because they take numbers e.g. from press releases of the local health authorities rather than waiting for the official tally.
It's not clear what specific differences Katherine Miller was referring to, but in the United States there are generally two points where vote totals are reported: At each county -- which is the largest governmental entity smaller than the state, and where the votes are actually tallied by the Registrar of Voters or Board of Registrars -- and the state level, where the tallies are aggregated by the Secretary of State or similar office. (Also, counting doesn't begin until the polls close; you may get some results minutes after the poll closure because the local offices are processing mail-in ballots.)
This isn't an instantaneous process. As a news organization (or a citizen), you you can go to the Secretary of State's website and get their totals summed from the numbers submitted to them by the counties -- or you could go to each county's website (Alabama has 67) and see their reported tallies throughout the night and keep a running total. (With luck, each county has standard API that will let you query the data with ease; often you'll wind up having to use a script to scrape their results pages.)
Depending on how the state's infrastructure works and how frequently the news organization is retrieving the county-level numbers, there could easily be a temporary difference in the totals being reported.
(Election-night totals reported by the state are usually "semi-official final," and not final, since there will be a lag from absentee ballots mailed by election day, sealed ballots turned in at the polls and such.)