The Feed Fix That Turned a Variant Problem Into Revenue

Business Overview

Artem Straps sells watch straps direct to consumers. The catalogue is variant heavy by nature. Every strap exists in several widths, materials and colours, and every buyer is searching for compatibility with one specific watch they already own.

In a business like this the catalogue is the marketing. If Google cannot read the product data, no bid strategy will save the account.

The Challenge

Shopping and Performance Max were serving the wrong variant against the wrong search. Someone searching for a 20mm strap for a specific watch model would land on a product page and still have to work out whether it fitted. Every one of those clicks was paid for and most of them left.

The account looked like a bidding problem. It was a data problem.

Objectives

• Make the product feed match the way buyers actually search
• Get the correct variant in front of the correct query
• Raise conversion rate without raising budget
• Scale Shopping and Performance Max on product data rather than bid pressure

Initial Metrics

Audit Findings

• Titles were written for the website rather than for search. A title such as Classic Leather Strap AS 204 tells Google nothing about width, material or compatibility, which are the only three things the buyer is searching on.
• Size, material, colour and GTIN were incomplete across a large share of SKUs. Missing attributes stop Shopping from matching a query even when the product is a perfect answer to it.
• Variants were submitted as separate items with no item group ID, so the account was competing against itself on the same physical product.
• One Performance Max campaign held the whole catalogue. Best sellers and slow movers shared a budget, so the reported return was an average of a strong product and a dead one.
• No custom labels existed, which means there was no way to bid differently by margin, stock position or seasonality.

Strategy

In a variant catalogue the feed is the campaign. Rewrite the product data so every variant answers a query on its own, then segment the campaigns using the labels the improved feed provides.

This sequence matters. Segmenting campaigns before fixing the feed just splits the same bad data into more buckets.

Feed Optimization

• Titles restructured to the pattern buyers actually search, leading with width, then material, colour, style and brand. A title became 20mm Brown Horween Leather Watch Strap Artem
• Size, material and colour attributes completed on every SKU in the catalogue
• GTINs added where available and item group ID applied so variants clustered correctly instead of competing
• Custom labels added for margin tier, best seller status, stock level and seasonality
• All title and label changes delivered through a supplemental feed so the store’s own product data stayed clean and editable by the client

Campaign Structure

• Shopping split into three tiers using the new custom labels: Best Sellers, Core Range and Long Tail
• Performance Max split into a branded asset group and a separate non-brand catalogue campaign, so brand demand stopped inflating the non-brand return
• Low stock SKUs excluded automatically, so budget stopped chasing products that could not ship on time

Bidding Strategy

Best Sellers ran on a lower target return in order to buy volume, because those products convert reliably and carry the account. Long Tail ran on a higher target to protect margin, because those products need a better return to be worth the shelf space. Branded activity was separated entirely so it could be measured and priced on its own.

Tracking Improvements

Enhanced conversions with value were implemented so the account optimised toward revenue rather than order count. Variant level revenue reporting was added, which is what made it possible to measure the impact of title changes at the SKU tier rather than guessing at account level.

Creative and Landing Page Changes

Product pages were rebuilt to answer compatibility in the first screen. A watch model fit list, a width selector above the fold and a clear fit guarantee. Fixing the page is part of media buying. Buying the click is only half the job, and it is the cheaper half.

Optimization Process

Titles and labels were reviewed monthly against search term data. Any query pattern appearing in the data that was not represented in a title was added in the next supplemental feed push. Over time the feed started to mirror the market’s own language rather than the brand’s internal naming.

Results

Key Learnings

Final Business Impact

The catalogue became the growth engine. Same products, same prices, same budget. The difference was that Google could finally read what was being sold, and the buyer could finally see that it fitted their watch.