Likert results: show the whole distribution.

Store the categories in their intended sequence: Strongly disagree, Disagree, Neither agree nor disagree, Agree, Strongly agree. Alphabetical sorting breaks that sequence. Keep the question text, collection period, population invited, and any routing rules beside the export.

A single ordered item is not automatically a validated multi-item Likert scale. Combining several items into one index requires an explicit scoring design and appropriate validation. Do not average unrelated questions merely because each uses five response categories.

Suppose 55 people reach the question. Fifty give valid agreement answers, three select Not applicable, and two skip it. The valid-response distribution uses a base of 50, while the coverage note records all 55 people who reached the question. Neutral remains a valid answer; it is not a missing-value code.

Use one record per respondent for this single-choice item and check duplicate submissions under a documented policy. If the survey is anonymous and no reliable deduplication key exists, state that limitation. An analysis cannot prove unique participation from response text alone.

Imagine the fifty valid responses are five Strongly disagree, five Disagree, ten Neutral, fifteen Agree, and fifteen Strongly agree. The percentages are 10%, 10%, 20%, 30%, and 30%. They sum to 100% before display rounding. The favorable share is thirty out of fifty, or 60%, if favorable is defined as the top two categories.

A horizontal ordered bar chart or a clearly labeled distribution makes the split visible. Show the counts as well as percentages. Small percentages can look precise even when they describe only one or two people, and rounded labels may sum to 99% or 101% without a calculation error.

Coding the categories from one through five gives a weighted score of 180 divided by 50, or 3.6. That mean assumes the distance from Strongly disagree to Disagree is comparable to the distance from Agree to Strongly agree. State the coding and assumption, and keep the distribution beside the score.

For reverse-worded items, confirm the intended direction before reversing scores. Preserve the original answers and create a separate scored field. A reversed code on one question can create an apparently dramatic change without any change in respondents' views.

Report each group's valid count, missing count, and distribution. A department with three answers should not receive the same visual certainty as a department with three hundred. Choose and document a privacy suppression policy before viewing sensitive subgroup results; there is no universal safe minimum for every organization.

Changes in question wording, order, survey channel, or respondent mix can affect comparisons. Pew's question-design guidance is useful context for these risks. A voluntary survey of customers describes respondents unless the design supports broader inference; a high favorable share alone does not establish representativeness.

Use comments to explain possible reasons for a pattern, but avoid treating a vivid quotation as the most common view without counting the theme. Remove identifying details before sharing comments or uploading them elsewhere. Theme counts can overlap if one comment covers several issues.

The Excel survey method covers the wider preparation workflow. For a report, retain the ordered response columns and a short coding note, then start a project. Review whether the findings preserve the valid base, show missingness, and distinguish observed patterns from claims the survey design cannot support.