Send Epoq the products
worth showing.

Epoq (epoq internet services GmbH) is a German AI-powered personalization provider for online shops, based in Karlsruhe. Rather than a single tool, Epoq is a suite of four connected modules that all draw on the same underlying knowledge base: epoq Search (onsite search and filtering), epoq Advise (guided-selling product advisors), epoq Inspire (product recommendations and cross-selling), and epoq Connect (personalized newsletters and retention emails). That shared knowledge base is built by combining your product catalog with customer behavior signals such as clicks and purchases, then processed with machine learning to power search results, filters, recommendations, and email content across the shop.

Every Epoq module is only as good as the product data behind it. Epoq's own documentation states plainly that each service can only work with information present in the catalog, and recommends providing as many product attributes as possible across all services. In practice this means: onsite search results and filters are limited to the fields in your feed, guided-selling advisors can only ask questions your data can answer, cross-sell recommendations depend on how granular your category structure is, and personalized emails can only show the images and prices your feed supplies. A feed that is complete, correctly formatted, and kept current directly determines how well Epoq can search, advise, recommend, and re-engage customers on your behalf.

Epoq's documentation is specific about how each module depends on catalog attributes:

1. epoq Search: filters are limited strictly to fields available in the catalog, and Epoq warns that filter names must be spelled correctly and filter values consistently populated, since mismatches break the filter.
2. epoq Advise: guided-selling flows need the specific attributes that correspond to each advisor question (for example, an 'area of use' attribute for a step that asks where a product will be used).
3. epoq Inspire: cross-selling and recommendation quality depend on category attributes being broken down in detail rather than left flat or generic.
4. epoq Connect: personalized emails require every field that will actually be displayed in the message, including image URLs sized appropriately for the email layout and both original and reduced price fields for discount messaging.

Epoq accepts product catalogs as either XML or CSV, delivered via HTTP(S), Amazon S3, or FTP(S), optionally secured with standard authentication.

The XML format is based on RSS 2.0 extended with the Google Merchant product-data namespace (g:) plus an Epoq-specific namespace (e:). A minimal document declares both namespaces on the <rss> root, carries shop-level <title>, <link>, and <description> tags inside <channel>, and lists each product as an <item> with tags such as <g:id>, <g:price>, <g:image_link>, <g:product_type>, and Epoq extensions like <e:variant_of> and <e:ean>.

The CSV format must be plain text, encoded consistently in UTF-8 or ISO-8859-15, comma-delimited, with values wrapped in double quotes and any literal quotes inside a value escaped as two double quotes. A header row defines the column names, and every data row must contain exactly the same number of columns as the header.

Epoq's attribute documentation distinguishes required fields, recommended fields, and a longer list of optional fields. The core set, with both the XML tag and the equivalent CSV column, is:

Beyond this core set, Epoq's documentation also lists optional fields for condition, color, size, a large image URL, language key, delivery time in days, related accessories, and a 'new article' flag — the more of these are populated, the more precise filtering, advice, and recommendations can be.

Feedance lets you generate and host an Epoq-ready feed without maintaining custom export scripts:

1. In your Feedance dashboard, open Data Channels and click Create New Channel, selecting Product Feed as the data type.
2. Choose whether to output XML (RSS 2.0 with the Google/Epoq namespaces) or CSV, matching what was agreed with Epoq during onboarding.
3. Map your source product data to the required and recommended Epoq fields — product ID, name, product URL, price, image, description, category, brand, variant-of, and stock quantity — using Feedance's field mapping tools, including the '>' category hierarchy and separate rows for each color/size variant.
4. Feedance hosts the resulting file on an HTTP(S) URL (or delivers it to an FTP/S3 destination) that you hand to Epoq, together with any access credentials, so their crawler can retrieve it on schedule.
5. Set up a refresh schedule in Feedance so the hosted file, and any DELTA update file, stays in sync with your product data automatically.

Epoq's own FAQ states that it retrieves the product catalog at least once a day, and more often depending on the catalog's content and how important up-to-the-minute accuracy is for that shop. For catalogs that are large or change frequently, Epoq supports DELTA files as an alternative to re-sending the full catalog every time: a DELTA file mirrors the structure of the full catalog but includes only changed rows, plus a deleted column set to 1 for products that should be removed. Epoq recommends placing DELTA files in their own subfolder (named 'DELTA') and using timestamps in the filenames so they can be picked up and processed in the correct order.

These are constraints Epoq documents explicitly, not error codes — but they are the issues most likely to silently break search, filtering, or recommendations:

1. Misspelled or inconsistent filter names/values — epoq Search filters are generated directly from catalog field names and values, so a typo or inconsistent capitalization creates a broken or duplicate filter.
2. One product ID covering every size/color — if all variants share a single ID, Epoq cannot tell which variant a visitor is viewing, which breaks size- and color-based filtering and personalization. Each color (and its sizes) should be its own catalog entry linked back with variantOf.
3. Mismatched CSV column counts — every data row must have exactly as many columns as the header row; ragged rows will be rejected or misread.
4. Mixed or undeclared character encoding — the CSV encoding must be applied consistently across the whole file (Epoq supports UTF-8 or ISO-8859-15, not a mix of both).
5. Flat, non-hierarchical categories — categories submitted without the '>' hierarchy separator limit how well epoq Inspire can build cross-sell and recommendation logic.

General best practices, based on recommendations in Epoq's own documentation and blog:

1. Enrich attributes beyond the basics. Epoq's blog on feed preparation specifically calls out extracting descriptive attributes (style, material, finish, and properties like 'vegan') from free-text descriptions, since these become filterable and searchable facets.
2. Provide both specific and parent color values. If your catalog uses detailed shades like 'sky blue', also supply a broader parent color (e.g. 'blue') so customers filtering by general color still find the product.
3. Standardize size labels. Use a consistent scheme (S/M/L) rather than mixing letter and numeric sizing in the same field, so size filters and guided-selling steps work reliably.
4. Send more historical data where possible. Epoq notes that providing more historical catalog and behavior data improves the initial quality of recommendations and search relevance.
5. Populate categories in depth for Inspire. Detailed, multi-level category paths give the recommendation engine more signal for cross-selling than flat or generic categories.

Does Epoq require XML or CSV?
Either is accepted. XML follows RSS 2.0 extended with Google's product-data namespace plus Epoq's own namespace; CSV must be plain text with a header row, comma delimiters, quoted values, and consistent UTF-8 or ISO-8859-15 encoding.

How does Epoq handle product variants like size and color?
Each color variant should be submitted as its own catalog entry, linked to its parent product with the variantOf (e:variant_of) field, so that size and color filtering and personalization work correctly.

How often does Epoq fetch the feed?
At least once a day by default, more frequently if agreed during setup, with DELTA files available for shops that need faster, lighter-weight updates than resending the full catalog.

Can I host the feed anywhere?
Yes — Epoq supports HTTP(S), Amazon S3, and FTP(S) hosting, with optional standard authentication on the file.

What happens if I only send the required fields?
The feed will work, but Epoq's documentation is explicit that filtering, guided selling, cross-sell recommendations, and personalized email content all depend directly on which additional attributes (category depth, brand, EAN, size, color, stock) are present — richer data means richer functionality.

Epoq (epoq internet services GmbH) is a German provider of AI-powered personalization for online shops, based in Karlsruhe. It offers a suite of connected modules: epoq Search for on-site search and filtering, epoq Advise for guided-selling product advisors, epoq Inspire for product recommendations and cross-selling, and epoq Connect for personalized newsletters and retention emails.

Epoq is not a marketplace or ad channel. Its modules run on the merchant's own shop and newsletters, using a shared knowledge base that combines the product catalog with customer behavior such as clicks and purchases.

The catalog reaches Epoq as a product feed in XML or CSV format, delivered via HTTP(S), Amazon S3 or FTP(S). The XML format is based on RSS 2.0 with Google's product-data namespace plus Epoq's own extensions. Epoq retrieves the catalog at least once a day, and more often where needed.

Because every Epoq module can only work with the information in the catalog, feed quality is decisive. Search filters depend on consistent attribute names and values, recommendations depend on accurate categories and availability, and advisors need rich product attributes. A complete, regularly updated feed improves search results, filters and recommendations across the shop.