Skip to main content

Automate your file imports via SFTP

Drop your CSV files on an SFTP server on a regular basis so Feedier imports them automatically, without duplicates.

F
Written by François Forest

Menu → Sources → New import → Sync → CSV via SFTP

In short: your systems drop a CSV file on an SFTP server, Feedier picks it up, imports it, then moves it to an archive folder. This is the solution for a regular flow of feedback (tickets, surveys from another tool, CRM exports, transcripts).

Which method for a regular flow?

Your situation

Recommended method

Your data is in Google Cloud

BigQuery: push your data into a table and Feedier picks it up on its own. No file maintenance.

Your systems produce files

CSV via SFTP (this article)

You're importing history just once

Step 1: prepare the SFTP server

Option

What you do

1. Your own SFTP server

Confirm the host, port and username. Create a dedicated user with a home directory. Allow Feedier's IP addresses through your firewall. Feedier provides its public SSH key.

2. An SFTP provided by Feedier

Send the public IP address your files will be sent from (so it can be allowed) and the public SSH key of the sending machine. Feedier sends back the server address and your credentials.

Access is by SSH key only, never by password.

Create one folder per flow, for example one folder per survey.

Step 2: create the SFTP connection

  1. Go to Settings → Connections and click New connection.

  2. Choose SFTP (private key authentication) and fill in the server information.

Step 3: create the import

  1. In Sources, click New import, choose Sync, then CSV via SFTP.

  2. Fill in the fields:

Field

What it's for

SFTP connection

The connection created in step 2

Folder to watch

The folder where files arrive. If left empty, Feedier uses the connection's path.

File name pattern

To pick up only certain files, for example tickets_*.csv

Columns to import

The exact headers of your files, used in the mapping step

Date column

The column that gives each feedback's date

  1. Choose the source, map the columns, then save.

After the import, each file is moved to the archive subfolder of the watched folder. A file is therefore never imported twice.

File rules

Rule

Why

Required and stable headers, lowercase, with underscores, no spaces or special characters (zip_code, ticket_id, status, created_at)

A changed header is no longer recognized by the saved mapping: the import fails, or puts the column in the wrong place.

A single separator for the whole flow (comma by default, semicolon possible)

Mixing commas and semicolons makes the read fail.

UTF-8, values containing a separator or line break in quotes, inner quotes doubled ("")

The file is read correctly, verbatims included.

Dates in a single ISO-type format, per column

If you use a local format, flag it during the first import.

Only new records in each file

Resending the whole database (tickets, transcripts) every time slows down the import and creates duplicates.

One file per period and per flow, with a date in the name (in/survey_a/data_2026_11_11.csv)

Tracking and recovery after an incident are easier. If several files coexist, use names that don't overlap.

The column formats of the CSV import

NPS from 0 to 10, numeric ratings, multiple values separated by |.

Example header and rows:

ticket_id,email,status,created_at,comment,rating12345,[email protected],in_progress,2026-08-08,"A customer reported an intermittent issue",412345,[email protected],closed,2026-08-10,"Issue resolved after applying a fix",5

Avoid duplicates

If the same record can come back in several files (a ticket "in progress" then "resolved"), ask your Feedier project manager to enable a uniqueness rule on a stable identifier present on every row, for example ticket_id. The new record then replaces the old one.

File strategy

What you need

Full snapshot (the whole period every time)

Uniqueness rule required

Incremental (new records only)

A stable unique key

With the example above and uniqueness on ticket_id, the second row replaces the feedback for the same ticket.

Reprocess non-compliant files

If your systems produce a file that differs from the expected format (headers, separator, values), the Feedier team can set up automatic preprocessing.

  1. Send your Feedier project manager two samples: the original exported file and the expected file, ready to import.

  2. Validate the final format with them, so the attributes match how you use them.

  3. After that, keep sending your original format: it is transformed automatically.

Advanced method: a flow driven by an automation

For flows with preprocessing, or set up by the Feedier team, the import can also be driven by an automation that calls a webhook on a schedule.

The in and out folders

The server has two directories: drop your files into in; once processed, they are moved to out. The structure of out must mirror that of in:

in

out

in/survey_A

out/survey_A

in/survey_B

out/survey_B

in/survey_C

out/survey_C

The steps

  1. Choose the method (BigQuery or SFTP, option 1 or 2) and share the access details.

  2. Drop a representative file into in/<survey_name>.

  3. Run a first manual import of this file from Sources (File import → CSV): the column mapping is saved and will be reused.

  4. Copy the UUID of this import, shown in the Imports tab, and send it to your Feedier project manager.

  5. Confirm the schedule (daily, weekly or monthly), the separator, the file names, the unique key and whether preprocessing is needed.

  6. The automation is created in Automations: a recurring trigger, then a Webhook call to https://connect.feedier.com/webhook-test/import/feedback/file.

  7. Monitor the first runs and contact support if anything looks wrong.

Schedule: leave some margin after the file drop. If your system drops the file on Monday at 9 a.m., schedule the import for Monday at 11 a.m.

Example of the payload sent to the webhook:

{  "api_key": "YOUR_PRIVATE_API_KEY",  "source_type": "SFTP",  "uuid": "ab1b0543-d59a-4fa3-8792-d895f5c4ab70",  "sftp_configuration": {    "host": "54.246.123.185",    "port": "22",    "username": "CLIENT_NAME",    "ssh_key": "default",    "file_path": "feedier_connect/CLIENT_NAME/in/survey_A"  },  "preprocessing": "false",  "delimiter": ";",  "unicity": "false",  "unicity_attribute_name": "attribute_name"}

Parameter

Role

Values

api_key

Your private API key

source_type

Where the files come from (internal or external SFTP)

SFTP

uuid

The UUID of a previous import with exactly the same file structure; its mapping is reused

ab1b0543-…

sftp_configuration.host / .port

The server and its port. For a Feedier SFTP, the ones provided by the Feedier team

54.246.123.185 / 22

sftp_configuration.username

The SFTP user

feedier_connect

sftp_configuration.ssh_key

The private key. Leave default, unless you use your own server: then enter your key

default or the key content

sftp_configuration.file_path

The survey's in folder

…/in/survey_A

preprocessing

Does the file need to be reprocessed? false if it has exactly the structure of the UUID's file

true / false

preprocessing_workflow

The processing applied before the import, if needed

Provided by the Feedier team

delimiter

The file's separator

, or ;

unicity

Apply the uniqueness rule (replace the old record with the new one)

true / false

unicity_attribute_name

The key attribute, which must already exist in Feedier

ticket_id

Who does what

Item

To provide or confirm

Owner

Method

BigQuery (with the dataset.table) or SFTP (option 1 or 2)

You

Access

Host, port, user and firewall rules (your SFTP), or IP and public SSH key (Feedier SFTP)

You

Folders

in/<survey_name> subfolders and a sample file

You

First import

Manual import and UUID sent

You

Preprocessing

Original and expected files, then setup

Feedier team

Uniqueness

The key attribute

You

Schedule

Frequency, file names and location

You, then configured by Feedier

What AI does with this data

Each verbatim in the imported files is analyzed as soon as it arrives: sentiment, topics from your taxonomy, and masking of personal data (nlp_… attributes). Your tickets or transcripts are classified into the same topics as your surveys and reviews. Each analyzed verbatim is deducted from your AI analysis budget.

FAQ

My separator is the semicolon. Is that a problem?

No, as long as it's the same in every file of the flow.

Can I manage several surveys from a single SFTP server?

Yes: one folder per survey, one import per folder, and consistent file names.

I renamed a column. What should I do?

Update the Columns to import list and the import's mapping, or let your Feedier project manager know if preprocessing is in place.

Can I automate without a first manual import?

For the advanced method, no: the first import saves the right column mapping (through its UUID) and prevents unpleasant surprises.

My IP addresses need to be allowed. Which ones?

Ask support for the up-to-date list.


Last reviewed: November 11, 2026, version 4.23.5.

Did this answer your question?