Menu → Sources → New import → File import → CSV
In short: a CSV import brings existing feedback (former survey tool, CRM export, history) into Feedier in one go. For regular, automated sends, use CSV via SFTP instead.
What to know before you start
Item | Rule |
Rows | ⚠️ 10,000 rows max per file (use SFTP otherwise) |
Size | ⚠️ 50 MB max per file |
Encoding | ⚠️ UTF-8 (the most common CSV encoding) |
Separator | ⚠️ Comma by default, semicolon accepted. Never mix the two in the same import flow. |
Headers | ⚠️ Required on the first row, 180 characters max (spaces included) |
Template | A sample file can be downloaded at step 2 of the import |
Who can import a file?
Role | Access |
Admin, Editor | Full access |
Viewer, Restricted Viewer | No access |
Prepare the file
Each question type feeds the KPIs differently (satisfaction ratio, NPS, sentiment), so each one expects a specific format. A wrong format skews your KPIs. The calculation details are in KPIs and scores.
Data type | Expected format | Example |
NPS | Whole number from 0 to 10 |
|
Satisfaction rating (slider) | Number within the destination question's scale (1 to 5, 1 to 10…) |
|
Rating table | One column per criterion, each with a whole number within the scale. Feedier averages them for the satisfaction ratio. |
|
Emoji / smiley | The emoji's rank, as a whole number (digits, not icons) |
|
Single choice | The option's exact label |
|
Multiple choice | The options separated by |
|
Text (verbatim) | The text, in quotes if it contains commas or quotation marks |
|
Date | ISO format, consistent across the whole file |
|
Attribute | Any other column (email, ticket ID, country, branch…). Multiple values separated by |
|
Three pitfalls to avoid:
NPS and satisfaction both use 0 to 10, but are calculated differently: NPS classifies respondents as promoters, passives and detractors; satisfaction converts the rating into a percentage (7/10 = 70%). An NPS column mapped to a satisfaction question (or the other way around) produces wrong KPIs.
Reversed scale (1 = very satisfied): reverse the values before importing, otherwise your best customers will show up as the most dissatisfied.
Ratings as text ("Satisfied", "Not at all satisfied", "Not applicable"): replace them with numbers, or leave the cell empty.
Name the attribute columns
Rule | Example |
Lowercase, underscores, no special characters |
|
Reuse the exact name of an existing attribute | Otherwise a new attribute is created |
Attribute dates in ISO format |
|
Avoid duplicates
If the same record can appear in several files (a ticket "in progress" then "resolved", for example), add a unique key: a column present on every row that never changes, such as ticket_id or case_number. Feedier will then import each record only once.
Sample file
date | nps | satisfaction_welcome | channel | store | comment |
2026-08-24 | 8 | 4 | In store|Website | Chicago | Very friendly staff |
Import the file
Go to Sources and click New import.
Choose File import, then CSV.
Drop your file into Upload a file.
Check the import name and the date column, suggested automatically by AI. Click Re-detect if needed.
Choose the source:
Option | Effect |
Automatically create a source | Feedier creates a source whose items match the file's columns |
Existing source (manual mapping) | You match each column to an existing item (question or attribute) in the feedback model |
Match each column to a question or an attribute, then run the import.
Without a date column, each feedback takes the import date. Always choose a date when importing history: otherwise, everything will appear on the same day in your reports.
What AI does for you
When | What AI does |
When you upload the file | Suggests an import name and detects the date column |
After the import | Analyzes each verbatim: sentiment and topics from your taxonomy, even for several years of history |
Personal data | Names, places and numbers found in the text are masked and stored in |
Each analyzed verbatim is deducted from your AI analysis budget: for a large history, approval may be requested before the analysis.
Pre-import checklist
Check | OK? |
The file is UTF-8, under 50 MB and 10,000 rows | ☐ |
Headers are under 180 characters | ☐ |
NPS columns contain whole numbers from 0 to 10 | ☐ |
Ratings follow the question's scale, in the right direction | ☐ |
No rating is written as text | ☐ |
Verbatims containing commas or quotation marks are properly enclosed | ☐ |
Dates use a single format | ☐ |
Attribute names match existing ones exactly | ☐ |
The unique key (if you use one) is present on every row | ☐ |
FAQ
Why isn't my date recognized?
The format isn't consistent across the file, or it's ambiguous (02/09/2026 could be February 9 or September 2). Convert your dates to the ISO format YYYY-MM-DD.
What happens to "Not applicable" values in a rating column?
A rating column only accepts numbers. Clear these cells before importing.
Why did the import create a new attribute?
The column name doesn't exactly match an existing attribute. Use the exact name for future imports, and contact your Feedier project manager to merge the data already imported. See Attributes and referentials.
How do I import several values for the same attribute?
Separate them with | in the same cell, for example M2|RER C|T3b.
My file is over 50 MB or 10,000 rows. What should I do?
Split it into several files and import them into the same source.
Last reviewed: November 11, 2026, version 4.23.5.
