Product Titles and Attribute Enrichment, Tested

Titles earn their place because the matching system reads them, not because shoppers do. Google accepts 150 characters in the title attribute and most surfaces render roughly the first 70, so the decision that matters is what occupies the front of the string: the brand, or the noun a stranger would type. Across the apparel and sports nutrition catalogs I have tested, correcting product_type and the size and gender attributes moved impression share more reliably than rewriting titles did, because those fields are what the system falls back on when a title is ambiguous. How a test is actually run: Cohorts are split by item_group_id so variants of one product never land on both sides of a test. Reads are taken over three weeks against impression share, click-through rate and the share of queries you care about, not against revenue in week one, because seasonality will hand you whichever answer you were hoping for. Items below roughly five hundred impressions a week are not tested individually; they are grouped by category and read as a cohort, since a single item at that volume cannot produce a result you should act on. Where a test comes back inconclusive, it is reported as inconclusive. What gets enriched, and in what order: google_product_category and your own product_type path first, because they are cheap to fix and they change matching. Then the attributes that gate particular surfaces: gender, age_group, size, size_type, colour, material and pattern for apparel; flavour, form, serving count and dietary claims for supplements and sports nutrition. Titles after that, structured to a rule per category rather than one template across the catalog, since brand-first is right for a brand people search by name and wrong for almost everything else. Descriptions last, and only where the product page copy says something the marketing blurb does not. On Chinese-language catalogs: For Singapore, Malaysia and Taiwan, a Chinese title belongs on a Chinese landing page in its own feed. Translated titles pointing at English pages are a policy problem and a conversion problem at the same time. I write and review these myself rather than routing them to a translation vendor, because product nouns, unit conventions and size systems are exactly where machine translation breaks without telling you. What I refuse to write: Promotional text in titles, which is against policy and gets the item pulled. Keyword strings that repeat the category three times. Capitalised shouting. Attribute values invented to fill a column: an empty field is recoverable, a wrong one poisons matching and takes months to trace back. Where your catalog data does not support an attribute, the answer is to get the data from your supplier, not to guess it in a feed rule. Not included: Feed tool implementation, catalog syndication to marketplaces, and campaign management. Product photography and image compliance. Copywriting for the product detail page beyond what the feed carries. Bulk rewriting of a catalog with no measurement attached, which is available elsewhere at a tenth of this price. Who this is not for: Catalogs where the source data is missing so much that enrichment is really a data collection project. I will tell you that after reading the export, and the honest next step is a supplier data brief, which I can write but which is not this engagement. Teams whose feed is changed by three people with no record of who changed what, because a test needs a stable baseline and you do not have one yet. How the work lands: A written decision record per category covering the title rule, the attribute set and the reasoning, the enriched values as a file your feed tool or platform can consume, and the test design with reads scheduled and owners named.