Get the most out of Phraza.

Upload the richest file you have.

Phraza writes from facts, so every useful column counts.

Two columns you can't do without:

• SKU or another unique ID. It is the key that brings every text back to the right product when you import the export into your webshop, PIM or feed add-on.
• Product title. Even a raw supplier title tells Phraza a lot about the product and is a key source of facts.

Three columns that make a big difference, and you often have them already in your supplier data:

• HS code. The customs tariff number is almost a key to the Google Product Taxonomy. It tells Phraza what type of product it is, before it has read a single word.
• Image URL. Phraza reads the photo and finds details the text never mentions: color, shape, pattern, closure, material finish.
• GTIN / EAN. Keep it in your file. Google Shopping and marketplaces use it to identify your products, and Phraza carries it through to your export.

Everything else you have helps too: brand, specifications, sizes, existing descriptions. Messy supplier data is fine. More facts in, better texts out.

Build your core data first.

Analysis & enrichment is mandatory, and for good reason. It maps every product to the right Google Product Taxonomy code, fills that category's optimized attribute set from your text and images, and resolves conflicts when text and image disagree.

The result is your core data: one clean set of facts per product. Every content type, in every language and channel, is written from it.

Polish the facts, not the texts.

Before you generate, take a quick look at your core data. If something needs changing, fix it at the source with the AI Assistant, across as many products as you need.

• "Set the material to merino wool for every product in the Autumn collection."
• "Show me all products without a color."
• "Change the brand name from NORDWOOL to Nordwool everywhere."

Generate in the right order.

Content builds on content. Start with titles and descriptions. They are the richest texts and become useful input for the rest:

• Highlights distil the description into scannable bullet points.
• Meta-title and meta-description build on the title and description for search.
• Image title and image description connect the photo to the product's name and key facts.

Test on a small batch first.

Pick 10 to 20 typical products and run one content type at a time. Try a few tones, writing styles and title structures until the result reads exactly the way you want. Then save your choices as your default and run the rest.

Do this for every content type you plan to use. Each one has its own job, and the right settings for a title are not always the right settings for a meta-description.

Perfect one language, then add the rest.

Get your main market right first. When you are happy, run the same content types in the next language. Each language is written from the same core data, not translated from the first one.

Work in collections.

Group products into collections, for example a new season, one supplier or one category, and run enrichment, content and export on just that group. Imports and runs can create collections for you.

Use the campaign prefix for seasons.

For Black Friday, summer sale or a launch, add a free-text label like "Black Friday Deal:" in front of every meta-title when you export. Write it in the language you export, and clear it when the campaign ends.

Review smart, then export.

Every product carries a status and a quality score. Start your review with the products that need it most, and use the AI Assistant to find wording across the catalog, such as claims your market doesn't allow. Then export in the layout you uploaded, and it goes straight back into your shop.