I catch the 23:33 a few times a month and it usually has no more than 60/70 people on board.
Was involved in a study a generation ago in Washington DC. It was a commuter bus service from a far suburb, four inward journeys arriving between 0730 and 0900, four outward departing between 1630 and 1800. The trip took an hour each way so it needed four vehicles.
After various suggestions an additional evening departure was added at 1915, which the first outward vehicle returned to cover. Loadings were noted and passengers were quizzed, but what surprised some (not me) was how loadings on the main services picked up as well, because a key issue had been people, often only occasionally, having to work late, and if they missed the last one at 1800 they would be stuck. So although they were rarely on it, it contributed to the overall service revenue.
Typical actual load on this last trip was about a dozen (the main ones were around 35-45), and this was thus felt not to cover its costs. A range of accountants came, looked at the narrow figures, and proposed it for withdrawal, just not listening to the experiences from its introduction. After a couple of years they won, it was cancelled, and the service overall lost about 40 daily passengers.