Collect your data
This is the part everyone pictures when they think “research” — actually running your study and gathering the information. Your plan from Step 3 becomes real here. Do it carefully and consistently, and the rest of your project rests on solid ground.
Collecting data means following your method to gather the actual information that will answer your question — running the experiment, sending out the survey, doing the interviews, or pulling the numbers from a dataset. The whole game here is consistency and honesty: gather it the same way every time, write it down carefully, and record what really happened — not what you hoped would happen.
Before you collect a single number
Five minutes of setup now saves you from unusable data later. Make sure all of this is ready:
- Your recording sheet is built. A spreadsheet or table with the columns already labeled — decided before you start, not during.
- Your tools and materials are ready and the same ones you'll use the whole way through.
- Your procedure is written out so you do it identically each time.
- Consent is sorted if people are involved — and you've cleared it with a teacher or mentor.
- You've done a pilot — one trial run to catch problems before they cost you real data.
The golden rules of good data collection
Whatever your method, these five habits are what separate trustworthy data from a mess. Tap each for why it matters.
Same method, same conditions, same measurements — every single time. If you weigh things in the morning on day one, do it every day. Consistency is the only reason your data points can be fairly compared.
Write every measurement down the moment you take it. Never trust your memory — “I'll log it later” is how data gets lost or misremembered.
One clear home for all your data, with labeled columns and units. A tidy spreadsheet now means you can actually analyze it in Step 5 — instead of decoding mystery numbers.
Surprising, messy, or “wrong-looking” results are still real results — keep them. Deleting data you don't like isn't science, and the odd result is sometimes the most interesting finding.
Keep your original recordings untouched. When you start cleaning or calculating, do it on a copy — so you can always get back to exactly what you observed.
Decide how you'll record it — first
The single best move you can make: build your data table before you collect anything. Decide what you'll write down for every observation, and give each its own column.
A plant-growth study might log this for every measurement:
| Date | Plant ID | Light | Height (cm) | Notes |
|---|---|---|---|---|
| May 3 | A1 | Red | 4.2 | — |
| May 3 | B1 | Blue | 3.9 | leaf drooping |
| May 4 | A1 | Red | 4.8 | — |
Always include a Notes column — that's where you capture anything unusual (a hot day, a knocked-over pot) that might explain a weird result later.
Is your data collection trustworthy?
Run your setup through these six checks. The more it passes, the more you can believe your own results.
How much data is enough?
There's no magic number, but the logic is simple: the more you gather, the less a single fluke can fool you.
- More is better — up to what's realistic. Collect as much as your time and access reasonably allow.
- Repeat, don't rely on one shot. Run multiple trials or gather many responses so a single odd result can't drive your conclusion.
- For surveys, aim for as many responses as you can get — a few dozen is a common rough floor for spotting a pattern; more is stronger.
- Small samples can still be fine for interviews or a first exploration — just be honest that they show a hint, not proof.
Stay fair and honest
The easiest person to fool is yourself. A few guardrails keep your hopes from leaking into your data.
Guard against bias
Measure every group the exact same way, and don't let what you're hoping for change how you read a result. Where you can, “blind” yourself — e.g., label samples so you don't know which is which while measuring. Keep outliers (note them, don't delete them), and if people are involved, protect their consent, privacy, and comfort throughout.
Traps beginners fall into
- Changing your method partway through — now the early and late data don't match.
- Waiting to write things down and forgetting the details.
- Throwing out data you don't like instead of reporting it.
- Too few data points to tell a real pattern from random luck.
- Messy, unlabeled records you can't make sense of a week later.
- Letting your hopes steer what you notice or record.
Pro tip · back it up & date it
Save a copy of your data somewhere separate the day you collect it — a lost spreadsheet can mean a lost project. And log the little details next to each measurement (date, time, conditions); that “metadata” is often what explains a strange result later.