# Typeform Partial Responses: Which Data Should You Analyze?

Analyze Typeform partial responses using dated exports, question-level inclusion rules, and an example separating status from missing answers.

- Canonical URL: https://www.harmate.com/en/blog/typeform-partial-responses-analysis
- Author: Harmate Team
- Published: 2026-10-25
- Updated: 2026-10-25T22:48:34.586300+00:00
- Language: en

## Content

# Typeform Partial Responses: Which Data Should You Analyze?

A Typeform response marked “Partial” is neither a completed questionnaire nor a row you must automatically discard. It may contain a useful answer to an earlier question while leaving the rest of the journey unknown. To analyze it, freeze an export, retain the observed status, and decide inclusion **for each question**.

The practical output is a short inclusion table: which answers exist, which rows belong in each analysis, and what is known about missing values? This prevents a count of started forms from becoming a count of opinions about a question near the end.

## What the exported status tells you

Typeform supports collecting answers before final submission through the partial-response features available on an account. Its [partial-response documentation](https://help.typeform.com/hc/en-us/articles/21102221958676-Collect-partial-responses) describes a Partial Submit Point, which can collect preceding answers without the respondent completing the form. “Response type” distinguishes Partial and Completed responses and is also included in exported data.

This is the recorded state, not a permanent classification. Someone may return, change answers, and complete the form; answers and status can then change. Keeping only a row count without an extraction date makes the next comparison difficult to interpret.

Availability and collection options depend on the plan and configuration. Check the current help and the settings actually used for your form. Do not assume every answer was captured at the same point, or that partial collection was enabled from the start of the survey.

## Freeze the data before interpreting empty cells

Keep the downloaded file unchanged. In a working copy, retain the available response identifier, exported status, extraction date, and exact wording of the questions being studied. Record the partial-collection mechanism and any filters applied to the export.

If you export again tomorrow, preserve two separate snapshots. Compare a response across them only with a reliable identifier checked against the files. Identical text does not prove two rows belong to the same person; edited text does not prove they belong to different people. Without a reliable key, do not silently combine the snapshots.

The [guide to preparing survey responses before analysis](/en/blog/prepare-survey-responses-before-analysis-a-practical-method-and-example) covers this traceability. The Typeform-specific concern is that a later status must not silently replace the state used for an earlier finding.

## Three rows do not mean three answers to every question

**This example is entirely fictional.** A consultancy asks for workshop feedback. Q1 is “What situation was useful to you during the workshop?” Later, Q2 asks “What would you change for the next workshop?” The analyst uses an export taken on November 18 at 9 a.m.

| Fictional row | Observed status at 9 a.m. | Q1 | Q2 | What can be used |
|---|---|---|---|---|
| R-A | Completed | “The exercise with two different roles.” | “More time for the final discussion.” | Two available answers, each with its question. |
| R-B | Partial | “The moment we compared our approaches.” | Empty | Q1 is available. This export alone does not prove Q2 was seen. |
| R-C | Completed | “I did not find an example close to my work.” | Empty | Q1 is available. Check branching and whether Q2 was optional before describing the missing answer. |

If the agreed rule is to explore every available text answer to Q1, its corpus contains three answers, including one from a partial form. Q2 has one text answer. The two empty cells remain visible in the report with their limitations; they do not become two votes for “no changes.”

Suppose R-B becomes Completed the following day and gains an answer to Q2. The second snapshot describes a different state of collection. It can support an updated analysis, but its two Q2 answers must not be presented as the information available the previous morning.

The example also shows why a negative comment should not be excluded just because it is inconvenient. R-C answers the question by reporting a lack of a relevant example. Its disagreement belongs in the interpretation of Q1, alongside the positive experiences, with the same source traceability.

## Write an inclusion rule for each question

The right rule follows the analytical purpose. Exploring situations described in Q1 may include text from a partial response. Comparing two answers within a response requires both necessary values and a verifiable link between them. A form-level status does not replace either check.

Use a short decision record before examining the findings:

| Field | Decision to record |
|---|---|
| Question | Wording, instruction, relevant placement, and branching. |
| Snapshot | Date, time, filters, and form version. |
| Included rows | For example, available Q1 text answers, retaining Partial and Completed status separately. |
| Rows outside this analysis | Missing value or insufficient context; keep a reason when established. |
| Reporting base | Number of admitted answers to this question, not all exported rows. |
| Limitation | What cannot be established about other respondents or later steps. |

Distinguish “not presented,” “presented without a recorded answer,” and “reason for missingness unknown” only where supporting information exists. The questionnaire, branching rules, and available records can help; an empty cell alone cannot decide. An explicitly selected “no changes” option is an answer and must remain separate from a blank.

If you need a paired comparison, report how many rows contain both answers and what was excluded. Do not fill an absent score with zero just to keep a row. If you analyze each question separately, label each base separately so readers do not assume the same respondents supplied all the findings.

## Keep service counts separate from people

Typeform’s help explains that partial and completed response counting can depend on the browser session. A partial and final response in the same session can count once; submissions in separate sessions can count twice. That service-usage count is not a guaranteed count of unique people.

Likewise, Partial does not explain why someone stopped. It establishes neither fatigue nor lack of interest nor a technical problem. Keep collection monitoring distinct from analytical inclusion; the [survey response-rate guide](/en/blog/survey-response-rate-why-this-number-can-mislead-you) explains how to state the relevant bases.

Once the corpus is defined, open comments can reveal an unexpected situation or useful disagreement. Preserve their wording and question context. Do not erase what partial responses contribute by reading only completed forms, or present the partial responses as representative of everyone missing. The [open-response analysis method](/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions) covers the next stage.

With a compatible prepared file, you can [manually import responses into Harmate](/en/product/response-import) as CSV, JSON, or XLSX and review the column mapping. The cutoff, observed statuses, and inclusion rules remain part of your analytical method. Importing does not decide which partial answers are valid or synchronize later changes in Typeform.