How to Categorize Other Survey Answers in a Spreadsheet

To categorize open-ended survey responses, keep the original text, define a small set of categories, apply the same rules to every answer, and leave unclear answers unresolved. Then count the orders that answered your chosen question. This turns a list of comments into a report you can check and repeat.

For a Shopify discovery survey, the immediate question is often: what are buyers putting in Other? They may name a creator, a podcast, ChatGPT, or a source your answer list missed. The workflow below uses OrderSurvey's Responses export and a spreadsheet. It is a manual analysis process, not an automatic classification feature.

Start with one question and a defined reporting period

Choose one survey and one question, such as “Where did you first hear about us?” Write down the reporting dates, intended audience and exact question wording. Keep discovery answers separate from answers about what persuaded someone to buy today.

In OrderSurvey, open Responses, select the survey and date filter, then choose Export CSV. If your reporting window differs from the available date filters, export a wider period and filter the Date column in your spreadsheet. Export dates use UTC; choose a reporting time zone and apply it consistently.

The Responses CSV contains Date, Order ID, Question, Answer, Other Text, Location, Tags and Customer Tag Status. It has one row per answered question. It does not include order revenue or an automatic channel classification.

Keep an untouched copy of the export. In your working copy:

1. Filter to the discovery question. If two questions have identical wording, resolve which rows belong to the intended question before counting them.
2. Read Answer alongside Other Text. If you used a separate conditional text question, its answer appears on another row. Use the same order ID and response date within the selected survey to find that detail; flag ambiguous matches for review.
3. Keep order IDs as text. Resolve repeated IDs before counting answering orders. Do not combine all blank IDs into one order; report those rows separately if you cannot identify them reliably.
4. Add your analysis columns beside the original data. Do not overwrite what a customer wrote.

Watch the product walkthrough and inspect its illustrative export to see why one order can produce several question rows.

Write a short category guide before counting

A category guide defines what each label includes and what it excludes. Start by reading a small batch of answers, then write rules another person could apply without knowing your marketing strategy.

For a discovery report, useful working columns are:

These are columns you add yourself. They are separate from the export's Tags and Customer Tag Status fields. Editing a spreadsheet category does not apply a tag to a Shopify customer.

Use a conservative rule: classify what the answer says, not what you hope it means. “Instagram” identifies a platform; it does not tell you whether the buyer saw an ad, a creator's post or a friend's message. “A reel” does not even identify the platform.

If an answer names several sources without saying which came first, use a Multiple sources; order unclear category for this single-source report. Keep the detail. Do not silently pick the channel with the largest advertising budget.

Worked example: twelve Other answers

Every answer below is invented to demonstrate the method. These are not OrderSurvey customer results.

This yields three AI-assistant answers, two word-of-mouth answers, four other identifiable or partly identifiable source answers, and three answers without usable discovery detail. Retain those last three in the report; removing them would make the identified sources look more common.

For larger batches, have a second person review the ambiguous answers and a sample of the clear ones. Discuss disagreements, clarify the rule, and apply the revised rule to earlier rows that it affects. Do not let the same phrase mean one thing this week and something else next week.

Build the summary without changing the denominator

After resolving duplicates, make a table with one row per answering order for the discovery question. Keep a separate record of excluded or unresolved rows and why they were excluded. This is a count of answering orders, not unique people or all store customers.

In Google Sheets, select that table, choose Insert → Pivot table, add Category under Rows, and count Order ID under Values. Use COUNTA only after your input table has one row per nonblank order ID. Google's pivot-table instructions explain the source-range and field controls. When adding a new batch, check that the pivot's range includes it.

Keep these two questions distinct:

The second example assumes there were 48 distinct answering orders in the period. If other answer options also contain AI mentions, include those before reporting a total AI-mentioned share. Neither calculation establishes the share of all orders acquired by AI, and neither measures incremental sales.

Include the reporting dates, audience, question wording, rule version, unknown count and denominator beside the summary. If you need order values, that is a separate join; follow the survey-to-channel-report guide.

Use the findings to improve one decision

Repeated Other answers can reveal a missing option, a creator worth asking about, or wording customers find unclear. Choose one next action: review a named source, add an optional follow-up, or clarify an answer choice. A few mentions are a reason to investigate, not proof that a channel deserves a larger budget.

If you add a new option, record the date and start a new comparison period. Asking customers to recall a source in their own words is different from showing it in a list. Pew Research Center's question-design guidance explains why wording, answer choices and question context matter when measuring change over time.

Use the survey change-log guide to save the old configuration, document the new one and compare periods without confusing a questionnaire change with a marketing result.

The answer-options guide helps choose the next list. If a new option reveals a conditional follow-up, use the branch-testing checklist before publishing it.

Choose the plan for the reporting job

OrderSurvey Free includes 100 responses per rolling 30 days, with conditional follow-ups and first/repeat-order targeting. Unlimited is $24 USD/month and adds unlimited responses, CSV export and webhook alerts. The manual spreadsheet categorization described here is work you do with the exported file; OrderSurvey does not automatically perform it for you.

Install OrderSurvey to start collecting answers, or book personal setup help to choose a useful question and verify your Thank you and Order status blocks. Keep the first survey short enough that you can read and act on what comes back.

Frequently asked questions