First of all, gather together everything you need for your experiment before you get started. Doing this means you wont be interrupted, so that your momentum will be maintained and you will get through the experiment as efficiently as possible.
Great Science Fair Projects
Design Your Experiment
Physics 101 - Basic Information
Elsewhere on the Web
Hundreds of Science Fair Projects for Students
Science Fair Central offers ideas for science fair projects and experiments
Follow the steps you designed earlier as closely as possible. If the results are very different to what you expected, you might need to modify your procedures on the fly. In that case, be sure to note down (in your notebook – where else?) any changes you make to your experiment, so that you can include them in your report. When you need to modify the experiment, you should consider why things are not going as expected – is your hypothesis wrong, or are there assumptions that need to be rethought? Does your new design still answer the question you are interested in? It's perfectly fine to move to a new question if something more interesting comes up in the moment.
As you go through the experiment, record notes on your materials used, your observations and any thoughts you have about the experiment – they might be useful when you come to writing your report. Obviously, you need to record all the results of your experiments – ideally in a table format, with dependent variable measurements cross-referenced with their corresponding independent variable settings.
You should take the mean of several identical trials for each set of independent variables. This is done to minimize the effects of the uncontrollable random events that affect most experiments. Virtually anything can happen once or twice, so single results do not form a very convincing argument for or against your hypothesis. The more trials, the better. Depending on your level, you might be required to perform more a complicated statistical analysis – in this case, it is especially important to record every result, so that you can take means and find standard deviations and errors as required. This sort of analysis will tell you how accurate or reliable your results are.