Studio Data Exports

Data Exports

Export on the Data tab downloads a whole table as CSV, Excel, SPSS, Stata, Parquet or SQLite. This page describes exactly what each file contains, what every export includes and leaves out, how multiple answers are written, the size limit, and the other ways to get data out of Studio.

---

Exporting a table

1. Open Data and pick the table in the rail (responses, a table a flow wrote, quota_counters, …).
2. Click Export ▾ at the bottom right of the grid.
3. Choose CSV, Excel, SPSS, Stata, Parquet or SQLite.

A toast says Preparing responses.csv…, then Exported responses.csv. The file is named after the table (responses.sav, responses.dta, clean_responses.xlsx, …).

The export always contains the whole table — not just the 100 rows the grid shows, and regardless of the grid's search, sort or quick filters.

If an export fails, the reason stays under the grid with a Retry button (e.g. Retry CSV) until you dismiss it. A file the format cannot hold is refused with a reason rather than a server error, for example "The table cannot be written as .dta: …" or "The table cannot be written as Parquet: …".

Every member of the organization can export; exports are not limited by plan. Each export is recorded in the project's activity log (Settings → Activity) as data.export, with the table, the format and the number of rows — never the data.

---

What every export contains

For the responses table, in this order:

Included	Notes
one column per variable, in the questionnaire's order	the answers as codes, each column named after its variable (not the question's Id), in the order the current Save asks them. A matrix gives one column per row (the chosen column's code); a Multiple choice in the wide layout one column per choice (1 chosen, 0 offered and not chosen, empty when not answered or when the option was hidden from the respondent). "Other (please specify)" is the Other code (-66 unless the question names another) in the question's column and the typed text in <variable>_other, right after it; "None of the above" is -77; "Not applicable" the declared not-applicable code (-1 from the Builder). See Codes for Other, None of the above and N/A
codebook-only variables	variables no question asks — hidden, computed or assigned ones, such as an experimental arm — right after the questions
other answer keys	__status (completed, screened_out or redirect for interviews that ended on a special page; empty otherwise), a column left under an old Id, a flag a custom script wrote
url_<name>	each URL parameter the link carried: a source tag, a panel id (url_prolific_pid), the invitation token (url_inv)
duration_s, started_at	seconds from the start of the interview to the last save, and when it started: the opening of the survey page on which the respondent began, in that browser; a reload followed by Resume, or Save locally and finish followed by Resume, keeps it; Start over, or answering after a reload without choosing Resume, starts the count again from that page
captcha	pass or unavailable, when the captcha is on
tab_switches, hidden_seconds, pastes	the behavioral counts (see Data Quality)
last_page	the last page with questions the respondent saw, so partial responses can be told apart by where they stopped and screened-out ones by the page that screened them out
id	the row number (the respondent's Response ID)
survey_id	the environment that collected the row
respondent_id	the random id of the interview
partial	true for unfinished interviews
created_at, updated_at	first and last write

• All rows: completed and partial interviews, from all environments (pilot and main together). Filter before you analyze: partial = false for submitted interviews, survey_id for one environment (it is the 12-character id in that environment's link), __status ≠ screened_out to drop screen-outs.
• Fieldwork columns. The metadata columns carry the names a flow gives them, so a downloaded file and a flow's data agree. A column appears when at least one row has the value (a survey without the captcha has no captcha column). If one of your variables has the same name as one of them (say, a variable duration_s), the answer wins in the rows that hold one, and the other rows keep the metadata value. Only last_page is a download's alone: a flow's data leaves it out.
• Personal data. The url_* columns include panel ids and invitation tokens, which tie a response to a person. Treat a Data export accordingly. A research bundle leaves out the parameters no flow reads, and never carries the invitation token.
• If a variable is named like a table column (id, partial, …), the variable keeps the name and the table column becomes _id, _partial, ….
• Responses collected by an older survey build are written in the same layout as new ones: a matrix stored as one object becomes one column per row with the column's code (a 0–10 scale stored as 1–11 is written 0–10), a wide Multiple choice stored as one list becomes 0/1 per choice, Other stored as __other__ becomes the Other code plus <variable>_other, __none__ the None code, na the declared not-applicable code (still na without one), and script flags stored under __flags__ become columns (speeder). The survey page's internal columns (__options__, __pages__, __errors__, __timers__) no longer appear. An export you make now can therefore differ from one you downloaded earlier for the same responses. See Older surveys and responses.
• Answers collected by earlier versions of Studio under a question's Id have been moved to the variable's column; the few that could not be moved safely stay in a column named after the Id (see The responses table).

Other tables — quota_counters, survey_meta, the tables your flows write — are exported as they are, column for column.

Multiple answers

Answer	CSV, Excel, SPSS, Stata, SQLite	Parquet
multiple choice (several codes)	codes joined with ; — 1;3	a real list — [1, 3]
unanswered	empty cell	missing
other structured answers	JSON text	JSON text

The 1;3 form is the same one the engine and a flow's Export file node write, so every file of a project spells multiple answers the same way. (A Multiple choice in the wide layout has no list to write: it is one 0/1 column per choice.)

---

Formats

CSV

• UTF-8, comma-separated, one header row.
• Timestamps include their time-zone offset.
• Formula protection. A text cell (or header) that starts with =, +, -, @, a tab or a carriage return is written with a leading apostrophe ('=SUM(…)), so a spreadsheet opens it as text instead of running it. Numbers stay numbers; a -5 that a respondent typed as text is also prefixed.

Good for: anything, anywhere.

Excel (.xlsx)

• One sheet, Sheet1, with a header row.
• Timestamps without time zone (the stored time, UTC).
• Text that looks like a formula is kept as text, never as a formula.

Good for: colleagues, quick pivots.

SPSS (.sav)

The file carries the codebook of your questionnaire:

• variable labels,
• value labels (including the labels you give missing codes in the Codebook, e.g. -9 Refused),
• declared missing values,
• the measurement level (nominal, ordinal, scale).

The labels come from the current Save of the questionnaire — not from the version that collected the data. If you renamed a variable or changed codes mid-field, older responses are labeled with today's codebook (see Data from more than one version). The codes Studio adds for Other, None of the above and N/A carry their labels ("Other", "None of the above", "Not applicable" as a declared missing code), and a <variable>_other column has its own variable label. Columns that are not questionnaire variables (id, survey_id, the url_* and other fieldwork columns, …) are written without labels. If the questionnaire cannot be read, the file is still produced, unlabeled.

Variable names are adjusted to SPSS rules — the same names in every export:

Rule	Example
characters other than letters, digits and _ become _	q1.a → q1_a
a name must start with a letter; otherwise v_ is prefixed	1st → v_1st, __status → v___status
SPSS keywords get v_ (ALL, AND, BY, EQ, GE, GT, LE, LT, NE, NOT, OR, TO, WITH)	by → v_by
at most 64 characters	longer names are cut
names that clash (SPSS ignores case) get _1, _2, …	Q1, q1 → Q1, q1_1

Timestamps are written without time zone. Good for: SPSS, with labels ready for tables.

Parquet

• Column types are kept: numbers, booleans, timestamps with time zone.
• Multiple answers stay lists.
• A Parquet column holds one type. When a column does not — text next to numbers (for example na next to codes, where the codebook declares no N/A code), or lists next to single values — the whole file is written the way the other formats write it: multiple answers as 1;3, and each column that mixes text with numbers as text. A column that mixes true/false with numbers — a script flag such as speeder, stored as true by an older survey build and as 1 by a current one — is written as 1/0.

Good for: Python and R pipelines, large tables, type-faithful archiving.

SQLite

• One .sqlite file with a single table named data.
• Timestamps as text (2026-06-04 14:41:07+0000).

Good for: one file with the whole table, queryable offline with any SQLite tool.

Stata (.dta)

• Labeled like the SPSS file: variable labels, value labels, declared missing values, from the current Save.
• Names follow the SPSS rules above, cut at 32 characters (Stata's limit); names that clash after the cut get _1, _2, ….
• Timestamps are written without time zone.
• A missing text answer is written as an empty string, which is how Stata stores a missing string anyway. Long open answers (2,045 characters and more) are written as Stata long strings, next to skipped ones.

Good for: Stata, with labels ready. A flow's Export file node with Format Stata (.dta) writes a labeled Stata file of the data at that point of the flow. (The Simulate mode of Builder → Test downloads simulated responses as CSV, Excel, SPSS, Stata or Parquet — useful for preparing your analysis before fieldwork, but it is not your data.)

From a flow: R and the codebook

The Data tab exports the table as it is stored. A flow's Export file node writes the data as it is at that point of the flow — after cleaning, recoding and weighting, with the variables the flow made. You type the file's name in its File name field, such as clean (the field keeps outputs/ and the ending fixed), and choose what it writes in Format:

Format	Files written for the name clean (under Files, outputs/<flow>/)
CSV (.csv), Excel (.xlsx), Parquet (.parquet)	the data and clean.dictionary.json
SPSS (.sav), Stata (.dta)	labeled SPSS or Stata data and clean.dictionary.json
R bundle (.R)	clean.csv, clean.dictionary.json and the script clean.R
Codebook only (.json)	the codebook alone, no data

A node added from the palette writes <flow>_data.csv. There is no .xls: a name typed with it is answered "Choose Excel (.xlsx): an .xls name would hold an .xlsx workbook, and Excel warns about that.", with a one-click fix. A single run of the flow keeps these files under Files; Run all keeps them only when the flow lists them among its outputs, as the example study's flows do, and the line under File name says which (see What Run all keeps).

clean.R reads clean.csv (UTF-8) and its dictionary with the jsonlite package into a data frame survey_data: the codebook's missing codes become NA, labeled codes become factors (a code without a label keeps a level of its own), each column's label attribute is the variable's label from the codebook (not the question's text), and multiple-choice columns (1;3) stay text. Put the three files in one folder and run Rscript clean.R, or source("clean.R") from R — the script finds its files beside itself. See Export the cleaned data for R. The example study's flow cleaning writes one: outputs/cleaning/clean_responses.R, clean_responses.csv and clean_responses.dictionary.json in Files, the cleaned responses with their weight — kept by a run of the flow and by Run all alike, since that flow lists them among its outputs.

An .xlsx a flow writes keeps text as text: an open answer such as =HYPERLINK(…) stays the string it is, never a formula Excel would run — as in the Data tab's Excel export.

From a flow: a report's tables in Excel

Where the data exports above give the answers, a report's tables give the results. Check Also save tables to Excel on a flow's Save report node (Also Excel in the Report view) and each run writes, beside the report, outputs/<flow>/<report>.xlsx:

• a Contents sheet with the report's title and one linked row per table (Sheet, Section, Table);
• one sheet per table of the report — frequencies, crosstabs, banners with their significance letters, group means (their post-hoc pairs on a sheet of their own), test and model tables — named by the table's caption, each with its statistics under it; numbers are numbers, text is text;
• no charts, and no statistics wired into a section as a line of their own.

Download it from Reports (Excel), from Files or from the run's card; a Run all keeps it too. See Tables in Excel.

From a flow: a tab book

A Tab book (Excel) node writes every question of the study crossed by a banner of segments into one workbook — the file a client asks for after fieldwork. With Banner gender, region and Questions left empty, a run writes the workbook named in its File name — <flow>_tabbook for a node added from the palette, so outputs/<flow>/<flow>_tabbook.xlsx under Files:

• a Contents sheet: one linked row per question (#, Question, Variable, Base, Sheet) and the questions it did not tabulate, with why (an open answer, a ranking, a question nobody answered);
• one sheet per question, named after its variable: Total, then a column per answer of each banner variable, lettered (Male (A), Female (B), …, Capital (D), …); the Base of each column; each answer's count and column percentage with its significance letters in the cell beside it; the mean and standard deviation of a number you named; footnotes;
• a Notes sheet on the weight, the test, the level, the correction, the minimum base, the missing codes left out and the date.

The codebook's missing codes are left out and counted, a weighted run shows both bases, numbers are numbers and text stays text (an answer that begins with = is never a formula). Download it from Reports (the Tab book entry), from Files or from the run's card; Run all keeps it too. See Tab book (Excel) and Tab books.

---

Size limit

A Data-tab export produces at most 100,000 rows. A larger table is refused with: "This table exceeds the synchronous export limit of 100,000 rows. No file was generated; use a database export for the complete dataset." No partial file is produced.

For larger tables, use a flow: Responses (or Project table) → Export file writes the file during a run of that flow; it appears in the run's outputs and under Files. Connectors and a research bundle that includes the responses are limited to 100,000 rows as well.

---

Other routes out

Route	What you get	Page
Flow Export file node	.csv, .xlsx, .sav, .dta or .parquet of the data at that point of the flow — cleaned, weighted, with url_* and timing columns — plus its data dictionary; or an R bundle (.R), or the codebook alone (.json) — see From a flow: R and the codebook	Analysis Flows, Node Reference
Flow Write table node	a new project table, which you can export from Data	Analysis Flows
Flow Save report node, Also save tables to Excel	every table of the report in one .xlsx beside it, a sheet per table with its statistics, and a Contents sheet — see From a flow: a report's tables in Excel	Reports
Flow Tab book (Excel) node	every question by a banner of segments in one .xlsx — a sheet per question with bases, counts, percentages and significance letters, a Contents and a Notes sheet — see From a flow: a tab book	Reports
Connectors (Plus; some targets Pro)	a table pushed to Google Sheets, Excel 365, Supabase, Airtable, Dropbox, HubSpot (Plus), or S3, GCS, Azure, BigQuery, Snowflake, your own database, SFTP, REDCap, Salesforce, HTTP (Pro)	Connectors
Research bundle	data, questionnaire, code, codebook and a provenance file in one zip — the export for a co-author or a paper. Its data files keep the fieldwork columns but only the url_* parameters a flow reads, and never the invitation token	Reproducibility
API	the same exports for a script, with an API key	API and API Keys

See also

• Responses and the Data Tab
• Codebook and Variables
• Cleaning and Weighting Data
• Reproducibility: Code, Bundles and Citation

---

← Responses and the Data Tab · Studio contents · Data Quality →