Data begins with a recorded observation
Data is information recorded so it can be examined, compared or used to answer a question. It might be a number in a spreadsheet, a checked box on a form, a photograph, or a written account of what someone noticed. The recording matters: a passing impression can prompt a question, but it becomes usable data only when someone captures enough detail to revisit it.
Imagine a shop owner trying to understand why customers leave without buying. Notes such as “asked about delivery” and “could not find a size” are observations. They may reveal issues worth investigating even though they are not numbers. If staff record these notes consistently, they can look for patterns. If they jot down only memorable encounters, the notes may still be useful, but they cannot reliably show how common each problem is.
No form of data provides a complete copy of reality. The observer chooses what to notice, the recording method sets limits, and some events leave no record. Before interpreting any dataset, ask what was captured, what was missed and why it was collected.
Measurements attach values to defined things
A measurement assigns a value according to a rule. A delivery time, for example, could mean the minutes between placing an order and receiving it, or the days between dispatch and arrival. Both are reasonable measures, but they answer different questions. A number without its definition, unit and starting and ending points is easy to misread.
Measurements also depend on the tools and circumstances used to collect them. A person timing a queue with a watch may start when a customer joins the line; a ticketing system may start when the customer takes a number. Neither record is automatically wrong. They simply need to be understood before they are combined or compared.
Precision is not the same as accuracy. A system might report a duration to the second while using a start time that does not match the question you care about. When reviewing measurements, check whether the rule was applied consistently and whether the resulting value represents the idea it is meant to describe.
Categories turn differences into groups
Some data records kinds rather than amounts. A survey might group responses as “online,” “in store” or “by phone.” These categories make it possible to count and compare responses, but they also simplify them. Someone who researched online and purchased in store may not fit neatly into a question that allows only one choice.
Categories can be created before collection, as with a multiple-choice question, or applied afterward, as when a researcher sorts open-ended comments into themes. Either approach requires decisions about boundaries. Does “delivery problem” include a confusing delivery estimate, or only a late parcel? If different people would classify the same case differently, the category needs a clearer definition or a way to record uncertainty.
A category is not always a natural fact about a person or event. It may be a convenient label for a particular task. Keep the original responses when practical, document how labels were assigned and consider whether important distinctions disappear when groups are combined.
Missing and unrecorded information matters
A blank field has several possible meanings. The question may not have been asked, the person may have declined to answer, the answer may have been unknown, or a recording step may have failed. Treating every blank as “no” changes the meaning of the data. Treating every blank as identical can hide where the collection process needs attention.
The same caution applies to what never enters a dataset. A feedback form captures people who choose to respond, not necessarily everyone who used a service. Staff notes may reflect the times when staff were free to write. These records can still support useful questions, but they may not represent the full group. A pattern in the recorded cases is a reason to investigate, not automatic proof that the pattern holds everywhere.
Match the data to the decision
The useful question is not simply “Do we have data?” but “Do we have data suited to this purpose?” If a team wants to shorten a queue, timed waits can help show when delays occur, while customer comments may explain why a wait feels frustrating. Neither replaces the other. Combining them can produce a clearer picture, provided each record was gathered and defined in a way the team understands.
Before drawing a conclusion, write down the question you are trying to answer. Then check the source of each observation, the rules behind any measurements, the meanings of the categories and the reasons values might be missing. Also consider whether collecting or sharing a detail is necessary, especially when it concerns an identifiable person.
Good data literacy does not require turning everything into a number. It means recognizing what a record represents, what it leaves out and how far a conclusion can reasonably go. Observations, measurements and categories are all useful when their limits remain visible.
