A small business can receive plenty of survey answers and still not know what to change. One reason is a question that combines several experiences. If customers are asked whether collection was quick and convenient, a positive answer might refer to the speed, the opening hours or both. A negative answer leaves the same uncertainty.
This guide follows an original fictional parcel-collection business reviewing its collection arrangements. It shows how to define the decision, separate questions and report responses without treating them as the voice of every customer. The example, wording and numbers are invented for explanation. No survey was sent, no customer data was collected and no commercial result is claimed.
Define the decision before choosing the questions
In the fictional business, owner Nila is considering whether to trial a different collection window. Her decision is narrower than finding out whether customers like the company. She needs to distinguish whether customers can reach the counter during its hours from what happens after they arrive. A single overall satisfaction score would not separate those problems.
Statistics Canada's guidance on commissioning surveys starts with defining information needs and their intended use. It also suggests checking whether the information already exists. Applied here, Nila first reviews what her own records can establish before asking customers to remember the same details. This is an application to an invented business, not an endorsement of a particular survey template.
Write the proposed decision in one sentence and list the information needed for it. If a question's answer would not help distinguish the available choices, consider leaving that question out. Keep questions that reveal a relevant unresolved issue, even if they are less flattering than an overall rating. The aim is a decision that can be examined, not a form that mainly produces reassuring numbers.
Identify whose experience you are asking about
A recent collector can describe a collection experience. Someone who could not attend may have a different problem, and a person who has never used the service cannot answer the same factual questions. Define the group and the reference period before writing wording that assumes everyone completed the same journey.
Business Queensland's customer-research guidance encourages focused questions, testing and attention to the customers selected. For Nila's fictional exercise, that means recording whether the intended group is people with a recent collection to make, rather than only people standing happily at the counter. The selection affects what the results can describe.
Suppose Nila initially considers asking only customers available on a quiet morning. That would be convenient for her, but it may miss people for whom collection hours are difficult. Before building a conclusion, write down who had an opportunity to respond and who may be absent. Do not rename a convenient group all customers just because the questionnaire is short and easy to complete.
Split a question that asks about two things
Pew Research Center's question-writing guidance explains why questions covering more than one concept can be difficult to answer and interpret. In our exercise, How quick and convenient was your collection? combines the time spent waiting after arrival with the ability to attend. Separating them gives the reader a clearer task and the owner a more interpretable answer.
One proposed question could ask how easy or difficult it was to attend during the available collection hours. A separate question could ask about time spent waiting after arrival, limited to people who actually collected. These are proposed items requiring review and testing, not validated measures. They should use a clearly identified recent collection rather than relying on an undefined usual experience.
Check what different answers would mean. A person might find the hours difficult but receive the parcel promptly once there. Another might attend easily but wait a long time. Keeping these experiences separate prevents a middle rating from hiding two very different situations. The owner can then investigate the relevant part of the service instead of guessing what a combined score meant.
Use a reference period people can understand
The Australian Bureau of Statistics' archived questionnaire-design guidance discusses recall problems and the value of a specific, reasonable reference period. The page is archived methodological guidance, not a new 2026 publication. For a customer exercise, specify which encounter the question concerns instead of leaving recent or normally to each person's interpretation.
Nila's draft might refer to the customer's latest collection during a stated period. If more than one collection is possible, the instruction should make clear which one to use. Otherwise one person may describe the fastest visit and another an unusually difficult visit. Those are different ways of selecting an event, even if both respondents answer honestly.
Avoid asking for precision people cannot supply. A customer may remember that waiting felt long without knowing the exact minutes. If the business needs measured waiting times, ask what records or an appropriate observation could establish them. A survey report should retain the distinction between a person's recollection, their evaluation and a time measured by another method.
Check the answer choices as carefully as the question
A neutral-looking question can still offer an unhelpful set of responses. In an invented draft, ranges such as zero to five minutes and five to ten minutes leave five minutes in both categories. Decide whether exact durations are appropriate at all; if using ranges, make their boundaries clear and non-overlapping. This is an original illustration of a design problem, not real waiting-time data.
Pew also discusses the importance of reasonable response choices and how wording and order can affect answers. Do not scramble an ordered scale merely because randomisation sounds scientific. For Nila, the practical review is whether the scale's meaning is clear and whether it allows someone whose situation differs from the assumed experience to respond appropriately or skip the item.
Keep a non-applicable situation separate from an unfavourable rating. Someone who did not collect cannot rate the waiting time of a completed visit. If the form lets such people skip that question, check that the skip works. A blank answer should not automatically become zero minutes, neutral satisfaction or proof that nothing went wrong.
Leave room for an answer you did not anticipate
An optional open question can reveal what the fixed choices missed. For the fictional exercise, Nila could ask what, if anything, made the latest collection difficult. That wording does not require the respondent to invent a problem, and it leaves room for a detail the owner did not think to list. The proposed item still needs testing with the rest of the form.
Open comments require care when reviewing them. In one invented response, a customer mentions both a confusing message and an awkward collection time. Nila can record those as two themes while retaining that they came from one respondent. Counting two themes as two different customers would change the meaning of the result.
Keep reported experience and interpretation distinct. If a comment says the message arrived late, that is an account to investigate alongside the relevant records; it does not independently establish the cause. A useful review identifies what the customer described, what the business can confirm and what remains unresolved. Do not embellish a short comment with motives, diagnoses or a dramatic story.
Test the draft before treating it as finished
Before asking a larger group, check whether people understand the question and can choose an answer that fits. Ask a tester what event they thought about and what each option meant to them. Testing here is a proposed step in the fictional exercise, not a report that the questionnaire has already been validated or successfully used.
Statistics Canada recommends reviewing exact wording, keeping necessary questions and testing all language versions. In Nila's case, a translation should preserve the reference period, scale and distinction between attending and waiting. A fluent sentence is not enough if it asks a different question. Keep each revision with the form version so later readers know which wording produced which responses.
Test the route through the questionnaire as well as individual sentences. Does a person who did not collect avoid the waiting-time item? Can someone leave an optional comment blank? If the form changes after a problem is found, repeat the affected check. An attractive preview does not establish that the actual submitted responses will match the intended questions.
Record non-response instead of filling the silence
The ABS archived guidance on errors in statistical data distinguishes sampling and non-sampling problems, including partial and total non-response. It explains that people who do not respond may differ from those who do. Simply increasing the number approached does not automatically remove such bias. This limits what a small voluntary customer exercise can establish.
Use an entirely invented count: 40 eligible customers are invited, 12 respond, and seven of those 12 answer the collection-hours question. Five of those seven say attending was difficult. A report should keep the three denominators visible: invited, responding and answering that item. Saying five out of seven item respondents reported difficulty is different from saying most customers find the hours difficult.
Do not silently fill the other five respondents' missing answers with a favourable or neutral value. Nor should the 28 non-respondents become satisfied customers. Their views are unknown in this exercise. The purpose of retaining the missingness is not to dismiss the seven answers; it is to stop a useful signal being presented as a result the collection did not produce.
Keep the report small and traceable
Reconcile the counts before writing the headline. In the worked example, the seven item answers belong within the 12 respondents, who belong within the 40 invitees. Keep a short processing note if an accidental duplicate is removed or an answer cannot be interpreted, rather than changing the totals silently. A second reader can compare the reported counts with the retained records without needing every private comment reproduced in the public summary. This proposed check catches bookkeeping mistakes; it does not repair a biased selection method or establish that the responses reflect everyone who did not answer.
For Nila's fictional report, retain the decision, intended group, collection period, questionnaire version, how people were approached and the actual item counts. Summarise comments with their limits and identify unanswered questions. A reader should be able to see where a conclusion comes from without searching through an unlabelled sheet of numbers.
Include a practical explanation of how responses will be used and who will have access, reflecting the actual arrangements. Do not promise anonymity if the collection or follow-up can identify the person. Avoid requesting unnecessary personal details merely because the form has an available field. This is cautious planning advice, not a statement that one privacy notice satisfies every jurisdiction's requirements.
The report need not claim a margin of error for this fictional voluntary exercise. A precise-looking percentage would not turn its selection method into a representative sample. Describe the evidence at the level it supports, and seek suitable methodological help if the business needs defensible population estimates or comparisons with a much higher level of precision.
Use the findings to choose a bounded next step
The hypothetical responses may justify investigating collection hours, but they do not tell Nila the cost, staffing implications or actual effect of a change. Separate the survey finding from a proposed response. A limited trial would need its own plan and observations before the owner could say that a new window solved the problem.
Compare subsequent evidence with the question originally asked. If a later survey changes the wording, group or method, retain that change when describing any apparent improvement. If fewer complaints appear, do not automatically attribute the difference to the new hours without considering who responded and what else changed. A tidy before-and-after chart is not a substitute for examining the comparison.
More business planning examples are available in our Business section. A useful customer survey starts with an answerable question and ends with a conclusion proportionate to the responses. Sources consulted 6 November 2026: selected government and researcher guidance, including clearly identified archived ABS material. All Nila, parcel-business and numerical examples are hypothetical; no survey validation, representative estimate, revenue improvement or guaranteed commercial outcome is claimed.