Trial
A trial is one complete run of an experimental procedure; repeating trials reveals variability and makes the average more reliable.
Definition
A trial is one try of your experiment from start to finish, and scientists repeat trials because a single try might be a fluke. Each trial is one execution of the procedure under one set of conditions; multiple trials at each level of the independent variable are averaged to reduce random error, and the spread between them shows how repeatable the measurement is. Trials must be genuinely independent, meaning the apparatus is reset and re-measured rather than the same setup re-read, or the repeats will understate the real variability. The mean of many trials estimates the true value while random errors partially cancel, and the range or standard deviation quantifies precision.
Example
Releasing a cart down a ramp and measuring how far it rolls is one trial; three trials give $82\ \text{cm}$, $85\ \text{cm}$, and $84\ \text{cm}$. Timing $5\ \text{g}$ of salt dissolving at $30^\circ\text{C}$, three trials of $25$, $27$, and $26\ \text{s}$ give a mean of $26\ \text{s}$ and a range of $2\ \text{s}$. Five cart speeds of $2.38$, $2.42$, $2.35$, $2.44$, and $2.41\ \text{m/s}$ average $2.40\ \text{m/s}$ with a range of $0.09\ \text{m/s}$, and since the standard error of the mean falls roughly as $1/\sqrt{n}$, twenty trials would about halve that uncertainty.
Key Insight
If your trials are close together you can trust the result; if they are far apart, something is changing that you did not notice. Three trials is a common classroom minimum, but it is a floor, and noisy measurements need more. More trials shrink random error but do nothing for systematic error: a miscalibrated timer gives the same wrong answer every trial, and averaging just makes you more confident in it.