Fieldwork Design and Data Collection Reliability

Start with the geographical question

Good fieldwork begins with a question, not simply a list of things that could be measured. The question determines which data are useful, where to collect them and how to record them.

For example, an enquiry into urban regeneration could use pedestrian counts to investigate how much people use an area. These counts are primary data: information collected directly for the investigation. They can reveal differences in footfall, but cannot, on their own, explain why people visit or prove that regeneration has succeeded in every respect.

Before collecting data, decide exactly what each observation will represent. A pedestrian count might record people passing a fixed point during a set period. This is different from counting the number of people present in a whole shopping area. Clear definitions make the results easier to compare.

Design a sheet that makes consistent recording easy

A data collection sheet should be designed before the visit. It needs clear headings, units where relevant, enough space for the planned sample and a way to distinguish sites and repeated observations.

For a pedestrian-count enquiry, the sheet could use the following layout. This is a proposed design, not a record of an actual survey.

Give each observation its own row. If the enquiry measures river depth instead, the sheet would need fields such as site, distance across the channel and depth, with units in the headings. A questionnaire needs clear question numbers and response categories rather than measurement units for every answer.

Trial the sheet before using it for the full investigation. A short pilot can reveal that a heading is ambiguous, a response category is missing or there is too little room for tallies. Fixing these problems beforehand reduces inconsistent recording and missing data.

Use a clear, repeatable procedure

A procedure explains how data will be collected. It should be detailed enough for another group to follow the same method.

For the pedestrian enquiry, choose safe, accessible counting points and define a fixed line that pedestrians must cross. Use a timer to count for five minutes at each site. Agree which direction each observer will count and whether the measure is pedestrian crossings or unique individuals. If measuring crossings, someone returning across the line creates another crossing; that does not make the result a count of unique people.

Record tallies as people pass, then enter the totals immediately. Record the start and finish times and any interruption. Comparing sites at the same time helps reduce the influence of changing activity during the day. If this is impossible, record the timing difference and consider how it could affect the comparison.

Shared procedures matter when groups pool their results. Combining counts increases the available data only usefully if the groups have measured the same thing in comparable ways.

Choose a sample that represents the enquiry

A sample is a selection from a larger population. In geography, that population could be people, river locations, streets or observation times. Sampling is usually necessary because collecting every possible observation would take too long.

Random sampling selects observations by chance. For example, number all the eligible survey locations and randomly select some of those numbers. This reduces deliberate selection bias, although a small random sample can still miss important variations.

Systematic sampling uses regular intervals. Taking river-depth readings every 50 cm across a channel gives an organised spread of observations and is straightforward to repeat. However, regular intervals can miss features between the sample points.

Stratified sampling divides the population into relevant groups and samples within them. A questionnaire might include different age groups so that one group is not overlooked. If the aim is to represent the whole population, the numbers sampled from each group should reflect the investigation's needs and the population's composition. Equal numbers from each group may be useful for comparing groups, but do not necessarily represent their proportions in the population.

Sample size is the number of observations, measurements or responses collected. Repeated counts at one site and counts at several different sites both increase sample size, but serve different purposes: repeats help assess consistency, while additional sites improve spatial coverage.

A larger, well-chosen sample usually gives a stronger basis for conclusions because a single unusual observation has less influence. Nevertheless, many observations from just one street or one time of day may still give an unrepresentative picture. Balance coverage of relevant places, times or groups against the time available to collect data carefully.

Separate accuracy from reliability

Accuracy concerns closeness to the true value. In a busy pedestrian survey, missing people makes the recorded count too low. Counting the same crossing twice makes it too high. Clear counting boundaries, separate responsibilities and immediate recording help reduce these errors.

For measurements, choose suitable equipment, check it before use and follow a consistent reading technique. Recording the wrong unit or copying a value into the wrong row can also make the final dataset inaccurate.

Reliability concerns consistency when observations are repeated under comparable conditions. Repeat the method and check whether the results are similar. Large differences should prompt investigation: was the method applied differently, or did conditions genuinely change?

Geographical conditions often vary. A shopping street may be busier at lunchtime than early morning, so different counts at those times are not automatically evidence of poor counting. Record conditions and distinguish genuine change from inconsistent collection.

Repeatedly making the same error can produce reliable-looking but inaccurate results. Repetition therefore complements careful measurement; it does not replace it.

Use controls to strengthen comparisons

A control group or control site provides a comparison that does not experience the condition being investigated. In a regeneration enquiry, an area without regeneration could be compared with the regenerated area.

Choose a comparison area that is otherwise as similar as practicable, and apply the same counting procedure to both. Differences in shopping facilities, transport access or street size could also affect footfall, so a comparison alone cannot prove that regeneration caused a difference.

A control group is distinct from controlled conditions. Keeping the counting duration, timing and method consistent makes observations more comparable; it does not itself create a control group. Not every geographical enquiry needs a control group, but where a comparison is useful, its suitability and limitations should be explained.

Judge the quality of the collection

Evaluate fieldwork by tracing each weakness to its consequence. Missing people in a crowd affects count accuracy. Sampling only during one quiet period limits how well the data represent the day. Different counting durations weaken comparisons between sites.

Improvements should address the actual problem: clearer counting responsibilities, observations at additional times, or a standard duration across all sites. The aim is not simply to collect more data, but to collect relevant, accurate and sufficiently representative data using a procedure others can understand and repeat.

Design and procedure

• Start with the enquiry question; collect data that can answer it.
• Sheet: clear headings, site identifiers, dates/times, units, repeat numbers and space for conditions.
• Pilot the sheet and method before full collection.
• Standardise what is measured, equipment, duration and recording rules.

Sampling

• Random: selection by chance; reduces deliberate selection bias.
• Systematic: regular intervals; straightforward but can miss variation between points.
• Stratified: sample within relevant groups; ensures those groups are included.
• Larger samples help only if selection and collection are sound. Consider spatial, temporal and population coverage.

Data quality

• Accuracy: closeness to the true value; improve counting, measurement and recording.
• Reliability: consistency of repeats under comparable conditions.
• Similar repeated results do not guarantee accuracy.
• Genuine changes in conditions can explain different results.

Comparisons and evaluation

• A control group/site provides a comparison without the condition investigated.
• Controlled conditions keep the collection method comparable; they are not a control group.
• Other differences between sites can affect results, so comparisons do not automatically establish cause.
• Evaluate using: specific weakness → effect on results → targeted improvement.