How to Design and Write Up a Lab Experiment
2026-07-01 · 9 min read
A step-by-step approach to planning a fair test and writing a lab report that actually communicates what was found and why it matters.
Start with a question you can actually test
A good experiment begins with a question narrow enough to test directly, not a broad topic. "How does temperature affect reaction rate" is a topic; "how does increasing water temperature from 20 to 60 degrees Celsius affect the time taken for a reaction to reach a visible endpoint" is a testable question. Narrowing the question early makes every later decision - what to measure, what to hold constant, how many trials to run - much easier to make well.
Identifying variables correctly
Every well-designed experiment has an independent variable, the one thing you deliberately change; a dependent variable, the one thing you measure as a result; and a set of control variables, everything else that could affect the outcome and must be kept constant so it cannot confuse the result. A common mistake is listing control variables vaguely, such as "keep conditions the same," rather than specifically, such as "same volume of solution, same concentration, same container, measured with the same stopwatch." Specific control variables are also what let someone else repeat your experiment properly.
Writing a testable hypothesis
A hypothesis should state a predicted relationship between the independent and dependent variable, ideally with a reason drawn from existing scientific understanding, and it should be written so that the experiment could actually show it to be wrong. "Temperature affects reaction rate" is too vague to be falsified cleanly. "Increasing temperature will increase reaction rate because particles move faster and collide more frequently and with more energy" is specific enough that the results can clearly support or contradict it.
Designing a fair test
A fair test changes only the independent variable between trials, while everything else is held as constant as practically possible. Consider what could vary by accident - room temperature drifting during the day, a reagent from a new batch, a different person taking the measurements at different times - and either control for it directly or record it so it can be considered when interpreting the results. Repetition matters here too: a single trial at each value of the independent variable cannot distinguish a real effect from random variation, so multiple repeats at each condition, with an average taken, are usually necessary.
Choosing what and how to measure
Decide in advance the exact procedure for taking a measurement, including the instrument, its precision, and the specific moment or condition under which the reading is taken. Vague measurement procedures are a very common source of poor results - "measure the temperature" is much less reliable than "measure the temperature of the solution using a thermometer inserted to the same depth, thirty seconds after mixing, read to the nearest 0.5 degrees." Precision here is not pedantry; it is what makes the data trustworthy and repeatable.
Recording and presenting results
Raw results belong in a clearly labelled table, with units in the column headings rather than repeated in every cell. Processed results, such as averages or rates calculated from raw data, should generally sit in a separate table or be clearly distinguished from raw readings. A graph should be used when showing a trend or relationship is the point, with the independent variable on the horizontal axis and the dependent variable on the vertical axis, and it should have a title, labelled axes with units, and an appropriate scale.
Writing the analysis and conclusion
The analysis should describe the actual pattern the data shows, not simply restate the hypothesis. Reference specific data points or the overall trend, quantify the relationship where possible, and explain the observed pattern in terms of the underlying science. The conclusion should state clearly whether the hypothesis was supported, partly supported, or not supported, and should never overstate what the data can show - a single small experiment supports a specific, limited relationship, not a sweeping general law.
Evaluating limitations honestly
- Identify specific sources of error, such as a difficulty maintaining constant temperature or human reaction time in starting a stopwatch, rather than vague statements like "human error." - Distinguish random errors, which cause scatter and can be reduced with repeats and averaging, from systematic errors, which shift every reading in the same direction and are not fixed by repeating the experiment. - Suggest a specific, realistic improvement for each limitation identified, rather than a generic statement that more trials would help. - Note any factors that might limit how far the conclusion can be generalised beyond the exact conditions tested.