Automating Category Migration for Cleaner Ecommerce Taxonomies

When an ecommerce catalogue moves from an older structure to a new platform, category names are rarely the only issue. Legacy departments may contain inconsistent spelling, duplicate products, outdated attributes and navigation paths that no longer suit modern search behaviour. A carefully designed taxonomy migration preserves useful product knowledge while making the catalogue easier to browse, filter and maintain.

Using automation tools to map old categories to new taxonomies can reduce manual data entry and speed up a platform transition. For Australian retailers moving away from NCR Retail Online towards solutions such as Magento or WooCommerce through NCR Counterpoint partners, the process also needs to account for local pricing, GST, delivery options, stock locations and customer expectations.

Why Taxonomy Migration Matters

A taxonomy is the organising system behind an online store. It determines where products appear, which filters shoppers can use and how search engines understand relationships between categories. An old structure such as “Ladies,” “Mens” and “Kids” may need to become a more detailed hierarchy based on customer intent, product type, age group, size and use case.

Poor category mapping creates practical problems. A product can disappear from navigation, appear under several unrelated departments or inherit the wrong filters. A customer in Melbourne looking for outdoor furniture may find indoor accessories mixed into the results, while a shopper in Perth may be unable to filter products by weather-resistant materials or delivery suitability.

The change also affects reporting and merchandising. If historical categories are discarded without a crosswalk, sales comparisons become difficult. A well-managed migration retains old category identifiers where possible, links them to new terms and records every transformation for future audits.

Audit the Existing Catalogue First

Automation works best when the source data is understood before rules are applied. Export category names, product titles, descriptions, brands, SKUs, attributes, URLs and stock information. Look for spelling variations, empty fields, duplicate categories and products assigned to overly broad departments.

A useful audit separates categories into clear, ambiguous and obsolete groups. “Dining Chairs” may map directly to “Furniture > Dining > Chairs,” while “Homewares” could contain cookware, table linen, storage and decorative items. Obsolete categories should not automatically be deleted because their URLs may still attract visitors or support historical reporting.

Review local requirements during this stage. Australian stores commonly need GST-inclusive pricing, suburb and postcode delivery logic, click-and-collect locations and distinctions between metro and regional shipping. A product taxonomy that ignores these operational details may look tidy while still creating friction at checkout and in fulfilment.

Build Rules for Category Matching

Start with a controlled vocabulary containing approved category names, parent-child relationships and definitions. Add synonyms and spelling variants, such as “runners” and “sneakers,” or “barbecue” and “BBQ.” The correct choice depends on the store’s audience, product range and search data rather than a generic dictionary.

Rule-based automation can then assign categories using keywords, existing codes, product attributes and brand information. A mapping rule might send products containing “saucepan,” “frypan” or “stockpot” to “Kitchen > Cookware,” while a rule for “rechargeable lithium battery” could trigger a fulfilment or shipping review as well as a product classification.

Use confidence scores to separate easy decisions from uncertain ones. Exact legacy category matches may be approved automatically, while products with conflicting signals should enter a review queue. Human reviewers are particularly important for multi-purpose goods, bundles, gift sets and products whose descriptions contain marketing language rather than clear specifications.

For examples of catalogue groupings and merchandising language, retailers can examine everyday dining products as a reference point. The goal is not to copy another store’s structure, but to observe how a broad theme can be divided into useful shopping paths.

Select Tools That Fit the Data

A spreadsheet remains effective for a small catalogue, especially when it includes columns for legacy category, new category, confidence level, reviewer, redirect status and notes. Larger stores may need a product information management system, an ecommerce import tool or a custom script that applies mapping rules through an API.

Automation platforms should support bulk edits, version history and rollback. They should also preserve stable identifiers so a product does not become detached from its images, stock records or order history. Test the connection between the taxonomy tool and the destination platform before processing the complete catalogue.

Artificial intelligence can help classify descriptions and suggest parent categories, but it should operate within approved boundaries. A model may identify that a product is related to “home,” yet still choose an unsuitable department. Give it the controlled vocabulary, examples of accepted mappings and clear instructions for uncertain cases.

Test Navigation, Search and Business Data

A successful migration should be tested on a staging site before the new categories go live. Check category pages, filters, internal search, breadcrumbs, product recommendations and mobile navigation. Test common journeys such as finding a product, filtering by size, adding it to the cart and selecting delivery to Sydney, Adelaide or a regional postcode.

Review redirects for old category and product URLs. A removed category should lead to the closest relevant destination rather than a generic homepage. Monitor broken links, crawl errors and changes in organic traffic after launch. Search Console data, onsite search reports and customer service enquiries can reveal gaps that a spreadsheet missed.

Business data needs equal attention. Confirm that mapped categories still feed inventory reports, purchase orders, promotions and accounting exports correctly. Validate GST treatment, sale pricing, stock by store and click-and-collect availability. For retailers serving customers in Queensland, New South Wales and Victoria, different warehouse or store rules may affect what a category page promises.

B2B catalogues require additional checks around customer groups, negotiated pricing and account permissions. Guidance on B2B account management can help teams consider how taxonomy changes interact with business buyers and restricted product visibility.

Govern the New Taxonomy

Taxonomy work should continue after launch. Products change, suppliers introduce new ranges and customers use language that may not match internal terminology. Assign ownership to a merchandising or ecommerce team, and schedule regular reviews of zero-result searches, high-exit category pages and frequently filtered attributes.

Useful operating practices include:

• Keep a versioned mapping file with dates, owners and approval notes.
• Create a review queue for low-confidence or conflicting assignments.
• Track redirects, search performance and category-level conversion rates.
• Document rules for synonyms, seasonal ranges, bundles and discontinued products.
• Reconcile taxonomy changes with inventory, reporting and fulfilment systems.

A clear governance process prevents category sprawl. New departments should be added only when they improve product discovery or support a meaningful business process. Seasonal landing pages, promotional collections and temporary campaigns can often sit outside the permanent taxonomy, avoiding unnecessary structural changes.

The most reliable migrations combine automation with deliberate review. Rules handle repetitive classification, scripts process large files and machine learning surfaces likely matches, while experienced staff resolve exceptions and protect commercial logic.

Old categories are valuable source data, but they should not dictate the future structure of an online shop. Preserve identifiers, redirects and reporting relationships while building a taxonomy around how Australian customers search, compare and buy. The key point to remember is that automation should accelerate category decisions without removing the human judgement that makes the new catalogue useful.