Import or delete products from a file

To import products from a file, open the product list and select Import → Products. To delete the products a file lists, select Import → Products to delete instead — see Delete products. Both walk one file through a few steps, end on a Review step that says what the job will do, and then run in the background.

Use these for a one-off file. For a file that arrives again and again — from a supplier, or every night — set up a data source, which keeps its mapping and runs on a schedule; a supplier can upload the files to the heavendata FTP server. To run a file through a data source’s existing mapping instead, use Upload and import on the data source — see running files you have to hand.

You want to	Select	Steps	The button that starts it
Create products, or change products you already have	Import → Products	Upload · Map columns · Import options · Review	Import products
Delete products	Import → Products to delete	Upload · Map columns · Review	Delete products

Import is a button in the product list’s header, between Export and New product. If you cannot see it, you lack the permission to import products — ask an administrator for it. That one permission covers both menu items, deleting included.

Every step has the same footer: Cancel at the left, and the step’s own button at the right — Next on every step before Review — with Back beside it from the second step on. Nothing is imported or deleted until you select the button on Review.

Before you start: the identifier attribute

An import finds each product by its identifier attribute — the attribute that holds a unique code for every product, in most catalogs the SKU. Every row of your file needs a value for it, or for the article number when you map that instead — see Map columns.

If your account has no identifier attribute yet, the Upload step says so instead of offering the upload:

Choose the identifier attribute before you import products. An import finds the product to update for each row by this attribute’s value, usually the SKU.

The deletion steps say the same in their own words: Choose the identifier attribute before you delete products from a file. The file names each product to delete by this attribute’s value, usually the SKU.

Select Assign attribute roles beside the message and choose which of your attributes is the identifier. Attribute roles explains the role.

Import products

1 · Upload

Drop your file on the upload area, or select Upload file and choose it. One file per import; a second file replaces the first, and the mapping starts again from the new file.

Accepted	Not accepted
.csv, .xlsx, .xml	Anything else, including .json — the file is refused with Only CSV, XLSX, or XML files can be uploaded.

The file’s format is detected for you. While it is read, the step says Reading the file…. Once it is, File format shows what was detected — for a CSV file, for example, the character encoding, the field separator and which row holds the column names. Correct anything that is wrong and select Apply to read the file again.

Check the preview before you go on. Under Preview, the step shows about the first 200 rows of the file as they were read, before anything is imported. If columns are shifted or a value runs across two columns, the file format settings are wrong — fix them here. If a row cannot be read at all, the step shows Errors instead of the preview, with that row’s Row and Line numbers, the Columns found and Columns expected, and the Row as read. Reading stops at that row, so fix it and upload the file again to find any further ones.

If the file cannot be read, the step says Couldn’t read the file and why, and keeps your file format settings so you can change them and try again.

To start over with another file, remove this one with the × beside its name.

2 · Map columns

Each row of this step is one column of your file, under Columns of the file. Choose the product attribute it fills under Fields to import into.

The step opens already filled in. Most columns get a suggested attribute, and where its values need it, a conversion; a column nothing fits is left for you to map. Each suggestion carries a tag that says where it came from:

Tag	Where the suggestion comes from
Earlier import	The last import your account started with a column of this name — see what is remembered.
Name match	The column’s name matches an attribute’s code or name.
Value match	The column’s values fit one attribute — every value is a GTIN, for example, or every value is one of an attribute’s allowed values. It does not mean the values already exist in your catalog.
AI suggestion	A language model, for the columns the other three leave open. It sees the column names, up to 10 sample values of each, and your attributes — none of your stored products.

The icon at the start of each tag shows how sure the suggestion is — a green check for high, a yellow exclamation mark for medium, a red triangle for low — as a line above the rows repeats. Check the low rows first. A row you mapped or changed yourself has no tag.

If the values of a column do not fit the attribute it is mapped to, the row says so:

The import misreads or rejects the value in 3 of the 10 rows checked for this attribute. Select Edit pipeline to change the conversion, or map the column to another attribute.

The warning does not stop the import. Change the conversion with the button at the left of the row, or map the column elsewhere. Every node a conversion can use is listed in Field processing pipeline nodes.

Suggestions take a moment on a large file. The step shows Getting mapping suggestions… while it waits, and Next waits too. If the language model failed or could not answer in time, a note above the rows says how many columns it could not suggest for — map those yourself, or select Auto map to ask again.

Control	What it does
Auto map	Suggests again, for the rows that are empty. It never changes a row you set yourself.
Clear all mappings	Empties every row. If that would empty a row you set or changed, it asks first.
+ beside a row	Maps the same column to one more attribute.
Show only unmapped values	Hides the rows that are already mapped.

A column you do not map is not imported, so leave a column unmapped to leave that attribute unchanged on your products. Each attribute can be filled from one column only; choosing one that another row already fills is refused, and the row says which attribute is taken.

Translations have one target per language: a translatable attribute Name is offered as Name and as Name (en), Name (de) and so on, for each language of your account.

Next stays disabled until a column is mapped to the identifier attribute — or, if your account has one, to the article number attribute, the one with the base code role, by which the import then finds each product. A row found by its article number addresses the product as a whole, not one of its product variants. Until then the step names the identifier attribute — for an identifier attribute called SKU: Map a column to “SKU”, the identifier attribute, before you continue.

Data format, under the rows, sets how numbers, lists and dates in the file are read:

Option	What it sets	Default
Date format	How dates written as text in the file are read — see Date format	Date and time with offset
Decimal separator, Thousands separator	How numbers in the file are written: the decimal point, and what groups thousands	. and ,
Currency decimal separator, Currency thousands separator	The same, for prices	. and ,
Value separator	What separates several values in one cell	|
Timezone	Not used on an import	none

When the step opens, both decimal separators are set to the one your file uses, and the thousands separators to the other character — a file that writes 12,50 opens with , and ..

Date format

Date format says how the dates the file writes as text are read. Choose the one your file uses:

Choice	Reads dates written like
Date with offset	2011-12-03+01:00
Date and time with offset — the default	2011-12-03T10:15:30+01:00
RFC 1123 date and time	Tue, 3 Jun 2008 11:05:30 GMT
Custom format	the pattern you enter in Pattern

A date with an offset is converted to UTC; one without keeps the time as written.

A Pattern is a .NET custom date and time format string: it spells a date out with letters, and with the ., /, -, : or space between them that the file writes. The letters most files need are these; the app lists them under the box too, all but H:

Letters	Stand for	Example
d, dd	day, dd with a leading zero	3, 03
M, MM	month, MM with a leading zero	4, 04
yy, yyyy	year, in two or four digits	26, 2026
H, HH	hour, 0 to 23, HH with a leading zero	9, 09
hh	hour, 1 to 12, with a leading zero — use it with tt	09
mm	minutes, with a leading zero	05
ss	seconds, with a leading zero	07
tt	AM or PM	PM
zzz	offset from UTC	+01:00

So dd.MM.yyyy reads 03.04.2026, MM/dd/yyyy reads 04/03/2026, yyyy-MM-dd HH:mm reads 2026-04-03 09:05, MM/dd/yyyy hh:mm tt reads 04/03/2026 09:05 PM, and yyyy-MM-dd HH:mm:sszzz reads 2026-04-03 09:05:07+01:00. Every other letter is in Microsoft’s custom date and time format strings, which the app links too; names of months and days are read in English.

A custom format needs a pattern: until you enter one, Next stays disabled and the step says Enter a pattern for the date format first.

In the file	What happens
A date that does not match the format you chose	It is not read: the product is imported without that value, and the import’s log names the value as not recognized as a date.
A date written another way, while the Date format is still the default	It is still read where it can be, month before day — so choose a format whenever your file writes the day first.
An Excel cell formatted as a date	It is read as the date it is, whatever the Date format says.

What is remembered for the next file

When the import starts, your account remembers each mapped column — its name, its attribute and its conversion. The next file with a column of the same name opens with it mapped, tagged Earlier import. Only an import that starts is remembered.

The Data format and the Import options are not remembered; set them for each file.

3 · Import options

The options decide whether the import may create products, and what it publishes — for new products, and for the existing products the file changes.

Create new products is off when the step opens. Off, the import only changes products you already have: a row whose identifier matches no product is skipped — counted as skipped, and logged as information rather than as a warning. Turn it on if your file contains new products.

Option	What it does
Create new products	Lets a row whose identifier matches no product create one. A row that lacks a value its product type needs to split it into product variants is skipped.
Default product type	The product type a new product is created with when its row names none. Required while Create new products is on — Next stays disabled until you choose one.
Publish new products	New products are published, so channels get them.
Create drafts without publishing	New products aren’t published — channels get them only once you publish them. Selected by default.

Default product type, Publish new products and Create drafts without publishing can be changed only while Create new products is on.

Under Existing products, choose what channels get when a row updates a product that already exists:

Option	What it does
Apply changes to published products	Published products get the values from the file, so channels get them. Products never published stay unpublished.
Publish modified products	Each product variant the import finds by a row’s identifier attribute value is published with all its changes, including one never published before. A row without an identifier attribute value publishes nothing.
Save changes without publishing	Published products change, but channels see the new values only once you publish the changes. Selected by default.

So to import existing products without publishing anything, keep the default, Save changes without publishing; the changed products show their unpublished changes until someone publishes them.

These are the publish settings a data source has, under the same names. Here you choose one of the three options for existing products, where a data source lets you tick Apply changes to published products and Publish modified products together:

Here	On a data source
Publish new products	Publish new products ticked
Create drafts without publishing	Publish new products not ticked
Apply changes to published products	Apply changes to published products ticked
Publish modified products	Publish modified products ticked
Save changes without publishing	Neither of those two ticked

4 · Review

Review says what the import will do before it starts, in sentences: which file it reads — and, for an Excel file, which sheet — what happens to a row that matches no product, what is published, every mapping under Mappings, and last a warning:

Existing values in the attributes you map are replaced, and an empty cell clears the value. This can’t be undone.

If something is not what you want, go Back and change it. Select Import products to start the import.

After you start the import

The job then takes the place of the steps, and the import runs in the background: you can leave the page or close your browser, and it carries on. While it is queued or running you can have an email sent when it completes: select Notify me to be emailed at your own address, or enter another address first.

When it has ended, the page shows how it ended — Completed, with Check the warnings below. when the run logged warnings, or Failed with Couldn’t finish “Import default”. Select View details to see why. — and what became of your file’s records, as one sentence:

default (uploaded as products.csv): imported — 57 records created, 1,180 updated, and 3 skipped

An import from the product list always names its file default, with your file’s name after it. A count of none is left out. If some records could not be imported, the sentence counts them after the others, and if every record was refused, it says the file couldn’t be imported. Every form the sentence takes, and what each means, is in reading what a run did.

Download followed by your file’s name — Download products.csv — under the sentence gives you back the file the import read, View details opens the run with its log file, which names each skipped record and the publish options the import ran with, as chosen for this import, and Continue to product list takes you back. To find the run again later, see background jobs. The run is listed as Import default — for every import and deletion from this menu.

Delete products

Deleting is permanent. There is no trash and no undo. Keep an export of the products before you delete them if you might need them back.

Your file needs one column: the identifier of each product or product variant to delete. Any other columns are ignored.

1. On the product list, select Import → Products to delete.
2. Upload your file. Format detection, File format and the preview work as they do for an import, except that while the file is read the step says Upload a file to continue. rather than Reading the file…, and it shows no Errors list: check the preview before you go on.
3. On Map columns, map the column that holds the identifiers to the identifier attribute: select Select product beside it and choose the one entry offered, the identifier attribute, shown with its key — data. followed by its code, for example data.product_sku. A column headed with the code itself is mapped for you. Next stays disabled until the column is mapped.
4. On Review, read what will be deleted. The step names your file and the column it reads, and ends with a warning:

   Deleted products and product variants are unpublished first, so channels stop getting them. This can’t be undone.

5. Select Delete products. The deletion starts and its job takes the place of the steps, as for an import; its sentence counts the records deleted — default (uploaded as delete.csv): imported — 12 records deleted.

The row’s identifier matches	What is deleted
A product	The product with all its product variants
One product variant	That product variant, with any product variants below it. The rest of the product stays
Nothing	Nothing — the row is skipped and the log says so
A product or product variant that a published product or record still requires	Nothing — the row is skipped with a warning in the log, and the other rows go on

References to a deleted product from other products are removed.

If the import does not start

If you select Import products or Delete products and the job cannot start, Review says so first — Couldn’t start the import. or Couldn’t start deleting the products. — and why, and you can select the button again. If the start was refused, nothing was changed. If the answer was lost instead — the connection dropped, say — the job may have started all the same: pressing the button again then shows that job. If you had changed anything since, the job runs with the earlier settings, and above it the page says:

Couldn’t apply your changes — the import had already started with the earlier settings. To import with your changes, start the import again.

A deletion says the deletion had already started and To delete with your changes, start the deletion again. To start again, open Import → Products — or Products to delete — from the product list.

If you leave the page

Once you have chosen a file, every way out of the steps asks first — Cancel, the back button in the header, the breadcrumb, the sidebar and your browser’s back button:

Discard your import settings?

Select Discard settings to leave and lose the file, the mapping and the options, or Keep editing to stay. Reloading the page or closing the tab gets your browser’s own warning instead. The deletion steps ask Discard your deletion settings? the same way. Before a file is chosen there is nothing to lose, and nothing asks.

The mappings of earlier imports are remembered whatever you choose — see what is remembered for the next file. Once the import has started, nothing asks.