Survey nonresponse leaves some outcomes unknown rather than automatically negative or positive. Use this fictional register to calculate the returned-answer fraction, the respondent yes fraction and attainable full-register bounds, then save a reproducible note.
Content checked: .
Freeze the register before reading the percentage
Survey nonresponse creates a gap between what returned answers say and what remains unknown across an invited list. This exercise makes that gap visible without estimating anybody’s missing opinion. Work with ten fictional eligible IDs, a fixed cutoff C1 and one binary question. Every ID is unique and belongs to the closed register by stipulation.
The teaching question asks whether a scheduled service arrived on time. Yes and no are the only eventual outcomes allowed in this model. Unknown means no answer has been supplied by the cutoff. It does not mean neutral, satisfied, dissatisfied or unreachable. No real customers were contacted, and no actual service result is reported.
Every unknown is allowed to become either yes or no independently of the others. That assumption is essential for the endpoint examples below. If a real register contains eligibility disputes, duplicate people or constraints linking missing outcomes, resolve those issues before claiming these endpoints are feasible.
Write the target as these ten IDs, not all customers. A closed register can be useful for a small internal follow-up, but it is not automatically a representative sample of a wider audience. The saved output must preserve that boundary even when someone later shortens the headline.
Count known answers and unknowns separately
| ID | Status at C1 |
|---|---|
| EN1 | Yes |
| EN2 | Yes |
| EN3 | Yes |
| EN4 | Yes |
| EN5 | No |
| EN6 | No |
| EN7 | Unknown |
| EN8 | Unknown |
| EN9 | Unknown |
| EN10 | Unknown |
Copy the ten rows into a note before calculating. Count four yes answers, two no answers and four unknowns. Reconcile 4+2+4=10. There are six returned binary answers, not ten. Keep both the total register count and returned count beside the working.
Do not remove the unknown rows to make the source look complete. Do not relabel them no merely because no positive answer arrived. Silence provides no answer to the teaching question. It also provides no stated reason for the missing answer; a claim that people were busy or unhappy would be invented.
This is not a classification exercise with a complete answer key. The classification error ledger compares supplied labels with known reference labels. Here, the missing outcomes have no reference answer. They must remain unknown in the observation record, even while hypothetical completions are explored separately.
If the wording itself is still being designed, use the survey question review for that different job. This exercise starts after a response-status card has been supplied; editing a question cannot recover its missing answers.
Report two observed fractions with their denominators
The returned-answer fraction is 6/10=60%. Among those six returned answers, the yes fraction is 4/6=2/3, approximately 66.67%. Those percentages answer different questions. The first describes how much of the register has an answer. The second describes the answers actually supplied.
A statement that two thirds of all ten IDs said yes is not supported. Four IDs said yes, while the other six include two known no answers and four unknowns. A more accurate observation is: six of ten IDs have returned binary answers, and four of those six answers are yes.
AAPOR’s Standard Definitions page explains that a response rate alone cannot determine whether nonresponse error exists or how large it is. Our simple unweighted return fraction is a teaching quantity, not a claim to implement a standard AAPOR outcome rate for every survey design. Real eligibility and disposition rules require their own method.
Likewise, this binary question is not Net Promoter Score. There are no recommendation ratings or promoter and detractor groups. Giving a familiar metric’s name to a different calculation would obscure rather than simplify the record.
Construct the lowest and highest possible completions
For the minimum possible yes count, imagine all four unknown outcomes were no. The four observed yes answers remain, giving 4/10=40%. Label this a hypothetical completion, not newly collected information. The original unknown rows must not be overwritten.
For the maximum, imagine all four unknown outcomes were yes. That gives 4+4=8 yes outcomes, or 8/10=80%. Under the stipulated unconstrained binary model, the full-register yes fraction can therefore range from 40% to 80%. Its actual value has not been identified.
| Hypothetical completion | Total yes | Fraction of all ten |
|---|---|---|
| All four unknowns become no | 4 | 4/10=40% |
| All four unknowns become yes | 8 | 8/10=80% |
These two constructions establish that both endpoints are attainable under our assumptions. Other attainable yes counts are five, six and seven. A finite register does not allow every decimal percentage inside the numerical range: here the possible fractions move in ten-percentage-point steps.
More generally, write total register size as N, observed yes count as Y and unknown count as U. The lower bound is Y/N, and the upper bound is (Y+U)/N. Their difference is U/N. For this card, the width is 4/10=40 percentage points. These are elementary deductions from the supplied binary rules, not formulas attributed to the survey organizations.
This range is not a confidence interval. It gives no probability to either endpoint or to the middle. It does not estimate the missing outcomes, correct nonresponse bias or demonstrate representativeness. An unusually wide range is a reason to describe the missing information clearly, not a license to choose its midpoint as a result.
The separate group and pooled rate exercise combines fully observed groups. Here the central task is different: enumerate what unknown binary outcomes could still do to a fixed register’s total.
Revise the headline without inventing an explanation
| Proposed claim | Decision | Reason |
|---|---|---|
| Two thirds of all invited IDs answered yes | Revise | Two thirds describes four of six returned answers |
| The four missing answers count as no | Reject | No missing answer is supplied |
| The fixed-register possible yes share is 40–80% | Accept with stated assumptions | Both endpoints have feasible hypothetical completions |
A usable replacement reads: At cutoff C1, six of ten eligible IDs have returned answers; four are yes. Four outcomes remain unknown. Allowing each unknown to be either yes or no gives a possible full-register yes share of 40–80%, not an estimated population percentage.
The US Census Bureau’s response definitions explain that differing characteristics of responding and nonresponding units can affect estimates. The ACS also has its own eligibility and weighting methods. We do not import those weights into this unweighted teaching card or suggest our simple fraction reproduces its official response measurement.
Keep edge cases explicit. If N is zero, a full-register fraction has no denominator. If N is positive but no answers are returned, the returned fraction is zero while the respondent yes fraction is undefined. By contrast, zero yes answers among a positive returned count gives a genuine zero respondent yes fraction.
If a real card leaves eligibility unknown or allows something other than binary outcomes, stop applying this specific model. Ask the record owner what each status means. A broader interval is not a remedy for an undefined register, and a missing answer should not be silently turned into a category it never supplied.
Save a note that another person can reconstruct
On Windows, search Start for Notepad, open the application and create a blank note. Copy the complete register, the count reconciliation, both observed fractions and the two hypothetical endpoint completions. Add the revised headline and a separate unresolved-information note.
Card:V1,fictional,closed ten-ID eligible register,C1 Yes4;no2;unknown4;total10 Returned6/10=60%;yes among returned4/6=2/3 Full-register bounds4/10=40%to8/10=80% Witnesses:all unknown no;all unknown yes Actual missing outcomes remain unknown Not an estimate,confidence interval or bias correction
Press Ctrl+S, select Documents and save as survey-bounds-V1.txt. Close the editor. Open File Explorer, go to Documents and double-click that file. Read all ten IDs again, confirm that the unknowns remain unknown, and check that the two-thirds label still refers only to returned answers.
A suggested allocation is three minutes for the register, four for counts and observed fractions, five for bounds and claims, and three for saving and reopening. This is a proposed fifteen-minute workflow, not a timed experiment. An existing calculator and text editor are enough; no account, installation or contact with actual invitees is required.
Anar’s published Perfect/Effect reflection describes project delays from pursuing perfection. Applied here, the principle is to complete one reconstructable decision note rather than endlessly polish a survey dashboard. It is not a report that he ran this fictional survey or measured an improvement with a client.
The artificial intelligence guide can help define a separate assistant task, but an assistant cannot supply missing opinions. If it rewrites unknowns as likely answers, reject that addition. Your core output is complete without any model response.
For actual invitation delivery and follow-up, the survey automation guide covers another workflow. Here, record the owner of unresolved data and a review point. If genuine answers arrive later, create a new version and recalculate; never present a hypothetical endpoint as collected evidence.
Frequently asked questions
Does no returned answer mean no?
No. Unknown is a missing binary outcome in this card, not an observed no answer.
Is 40–80% a confidence interval?
No. It is the attainable completion range for this fixed ten-ID register under the stated binary assumptions, not a probability statement or population estimate.
Does two thirds describe all ten IDs?
No. It describes four yes answers among six returned answers. Four register outcomes remain unknown.
What if nobody has returned an answer?
With a positive register size, the return fraction is zero but the respondent yes fraction is undefined. Zero yes among a positive returned count is different.
