When you go to the telescope, follow the plan as closely as you can. If the telescope and instrumentation are unfamiliar to you, you may have trouble keeping to the schedule for the first night or two, so try to work with a previous observer on the equipment for a night or two of training before your observing run. If you fall behind, try to keep as many low-altitude observations of extinction stars as you can, and drop some circumzenithal ones if necessary.
Don't change your mind in the middle of the night! If you think you will need to adopt a different plan in the middle of the run, make sure you generated that alternate plan ahead of time. The time to be creative is during the planning stage, not while observing.
Remember Steinheil's principle: only similar data can be compared. That means: DON'T switch to a different filter set or detector in the middle of a run. DON'T use different focal-plane apertures for different stars or different nights; they change the instrument's spectral response as well as its zero-point. DON'T alter temperature, high-voltage, or discriminator settings.
There are very great advantages to getting homogeneous data, because then the determination of the instrumental parameters can be spread over several nights; for example, see [26], [10], [19], [15], and [21]. Likewise, the extinction can be determined much more accurately from observations spread over full nights than from partial nights. If you are only interested in a small region of sky that is not up all night, either use the rest of the night to get extinction and standard-star data, or try to combine your run with another program that uses the same equipment; then all the data from both programs can be reduced together.
If there were instrumental parameters you weren't sure of when you made up the plan, make sure you find out their values during the observing run. The printed schedule will provide blank spaces for you to write them in. Many useful instrumental tests can also be run during the daytime, or during cloudy nights.
Remember that the extinction changes more on long time scales than on short
ones; so arrange your observations to chop as much extinction
variation as possible out of the most important data.
If you need very good 
colors, observe those filters in direct
sequence.
If you need to determine an eclipsing binary's light curve in several filter
bands, you may do better chopping between program and comparison stars before
you change filters (see [28] for further discussion).
A very common problem in observing records is an incorrect clock setting. Even a one-second error is worth correcting. An hour off is a common and much more serious mistake. Sometimes tired observers enter the wrong kind of time (say, UT instead of LST) into the observing records. It's a good idea to note both times in your observing log --- you do keep a written log of your observations, don't you? (It's easy to annotate the printed output of the planning program for this purpose.) Even if the computer is supposed to keep track of these things, what happens if there is a power failure, or the computer crashes in the middle of the night?
It doesn't hurt to take note of moonrise and moonset; they can be used to double-check the recorded times for gross errors. Be sure to write down anything that seems suspicious in the data, or possible problems with the equipment. Don't imagine you will remember all the details of what went on when you get around to reducing the data, which might be weeks or months later. For example, which night was it the mirror got snowed on? You can expect a big zero-point shift (or worse!) after that.
If temperature and humidity readings are not recorded automatically, be sure you write them down at least once an hour. (Space for this information is provided on the planning schedules.) The temperature that is important is the temperature of the photometer itself, not the observing room! But sometimes you only have a thermometer available on the wall; then record that, as any information is better than none at all. Be sure to record the time when each reading was taken. And don't forget to check the dark level now and then; it is a useful check on the health of the instrument.
It is very easy to forget that ``nuisance'' parameters, like extinction and sky brightness, have effects that propagate directly into errors of the measurements you care about. But this means you must be just as careful in measuring sky as in measuring stars, and as careful with extinction and standard stars as with program stars. Accurate and careful centering of stars is very important, particularly with photometers using opaque (inclined) photocathodes. Machine-like regularity is the ideal, attainable with some (but not all) automatic centering systems.
Likewise, one must always measure the sky the same distance from each star, regardless of brightness. Otherwise, you include a different fraction of the star's light in the sky measurement. A common mistake is to offset the telescope until the glow from the star is no longer visible in the measuring aperture; this guarantees that bright stars are measured with a different system than faint ones. A fixed angular offset --- with due regard to the telescope's diffraction spikes -- should always be used. If possible, always measure the sky in the same place for each star, to make sure you include the same unseen background stars in every measurement.
Finally, because sky brightness fluctuates on short time scales, it's best to measure sky for each star. You can use much shorter integrations for sky than for bright stars; but don't let more than a few minutes go by without re-measuring sky. During lunar twilight, or when the Moon is near the horizon, when the sky is changing rapidly, measure sky both before and after each star.
And, if you are new to photometry, you won't believe how sensitive the results are to the slightest hint of a cloud anywhere in the sky. Even a tiny cloud anywhere in the sky means you probably have ``incipient clouds'' (variable patches of haze) all over the sky, because of the layered structure of the atmosphere. Even Steinheil, a century and a half ago, found that the slightest trace of cirrus made photometric measurements impossible --- that's how gross the effects are.
It's a good idea to look at the sky around sunset. If there are ``layers'' visible in the sky near the western horizon, or ``notches'' in the sides of the setting sun, you are in for trouble, and will need to put extra effort into extinction determinations. During the night, any indications of clouds should be noted in your log. Often, cirrus is not visible until the Moon rises --- or until the data are reduced!