
PRODUCT FEED OPTIMISATION
Your product feed decides whether you appear in Shopping ads, free listings and AI answers. s360 is a marketing and AI agency, and our feed team optimises your feed for you as a managed service. We test every major change against a control group, so you know what each one was worth. Your team doesn’t have to learn a new tool, and nothing ends up in your e-commerce backlog.
product feeds optimised for brands and retailers across Europe
additional clicks a year for JYSK
more conversions and 21% more clicks for RevolutionRace on the same budget
Global Search Awards 2025 and European Paid Media Awards 2026
HOW WE CAN HELP
Our feed team helps brands and retailers get their products shown in Shopping ads, marketplaces and AI answers by improving the product data behind them, and we measure what each change was worth.
WHY IT MATTERS

ONE FEED, EVERY SURFACE
Paid, organic and AI search now run on the same product data. For the brands and retailers we work with, product ads take around 80% of paid search spend. Yet most retailers still run a feed that was built for Shopping ads years ago, on surfaces that didn’t exist when it was built. Your feed now supplies:
Google Shopping ads, through Performance Max and Standard Shopping campaigns
Google’s free listings on the Shopping tab and in Search, with no ad spend
AI Overviews and AI Mode, which ground shopping answers in Merchant Center data
ChatGPT product ads, which read structured product feeds
Marketplaces like Amazon and Zalando, which onboard your catalogue from feed exports
Meta catalogue ads on Facebook and Instagram
And when Gemini, ChatGPT or Claude recommends a product in an answer, there’s no auction to enter. Whether yours is the product named depends on how well your data describes what the shopper asked for.
Incomplete data can keep your product out of the results altogether. Take a shopper asking for a “waterproof women’s hiking jacket under €300 that packs down small“. Your jacket costs €280, but product_highlight and question_and_answer are empty, so your data answers two of the four parts of the question. A competitor’s jacket costs €290, and its feed says “fully waterproof with taped seams” and “Packable? Yes“. It answers all four, so it’s eligible and yours isn’t. The competitor wins despite the higher price.
Your product feed now decides whether you appear in Shopping ads, free listings and AI answers.
THE CHALLENGE
Google keeps changing what a complete feed looks like. It keeps adding attributes, such as the conversational attributes like question_and_answer that answer the longer, more detailed questions people ask in AI Mode. A feed that was complete a year ago often isn’t today. On top of that, the work itself doesn’t stop:
Thousands of SKUs, every attribute, in every market
What works differs by retailer and by category, so there’s no single recipe
New products arrive daily, and existing ones change
New attributes and new destinations keep arriving
The data coming out of the PIM is thin and inconsistent
There’s an organisational reason too. Marketing is measured on the results, but the feed is usually generated from a PIM that marketing doesn’t control, by a team with its own roadmap. Feed improvements end up competing with checkout releases, and they usually lose. The work sits undone even when everyone agrees it matters.
And most feed changes are never measured. In a global survey of PPC specialists, respondents named missing product data and tracking the impact of feed changes as their two biggest feed challenges. Some of those changes make things worse. When we added the product type to the start of the title for two advertisers, clicks rose 21% for one selling paint and fell 9% for one selling furniture.
OUR APPROACH
Our feed specialists run the work in three phases, and the third keeps going for as long as you work with us.
THE WORK
Attributes that are missing
Most feeds leave attributes like material, product_highlight, question_and_answer and related_product empty. We generate them, including the conversational attributes AI search uses.
Attributes that were never usable
Some values exist but can’t match a search. A colour field that says “AUCO” holds a manufacturer code, and nobody searches for AUCO. After optimisation it says “Brown“, localised per market. A title like “FIREBIRD WVN TP - Casual” becomes “adidas Originals Firebird Woven Track Pants – Men’s – Brown“, so brand, product type, gender and colour can each match a query.
Images
As an add-on, we remove and replace backgrounds, generate lifestyle images from packshots, and fix alignment and shadows.

Colour
AUCO
Brown
Title
FIREBIRD WVN TP - Casual
adidas Originals Firebird Woven Track Pants – Men’s – Brown
PROOF, NOT GUESSWORK
We measure effects with FeedX, an open-source framework for feed experiments built by Google’s ads data science team. Your catalogue is split in two: one half gets the change, the other doesn’t, and both run at the same time in the same account. Halfway through, the groups switch, so differences in how the two halves normally perform average out. We leave out a short period after the start and after the switch, so effects don’t leak across.
You get the effect of each change with a confidence interval. A before-and-after chart can’t give you that, because it mixes the change with seasonality, price moves, competitor activity and budget shifts.

Catalogue
First half of the test
After the switch
Half A
Gets the change
Control group
Half B
Control group
Gets the change
Short periods after the start and after the switch are left out, so effects don’t leak across. The result is the measured effect with a confidence interval.
OUR TECHNOLOGY
The work runs on Quantum, the workbench we built in-house for optimising feeds with thousands of products across many markets. It gives our specialists:
Full control over how each attribute is generated, per category and per market
Context for the AI through worked examples, and freedom to use any LLM
Optimised attributes pushed into Merchant Center and other channels without touching your master feed
Nordic Outdoor DK
12,480 products
Synced 06:00
Feed items
Audit
GMC evaluation
Custom audit
Term library
English
Colours
36
black
white
navy
olive
grey
red
+30
Genders
3
male
female
unisex
Age groups
2
kids
adult
Brands
20
Fjellkraft
Nordvik
Trailborn
Havbris
+16
Materials
20
Recycled nylon
Merino wool
Down
Fleece
+16
Title & description audit
1,204
Titles missing colour
3,880
Short descriptions
Attribute coverage
64%
98%
Colour, material and gender filled for 4,120 products
FEED TOOLS AND QUANTUM
Feed management tools map what you already have. As an agency, we use Quantum to create what’s missing and test it, and our specialists decide what to change next. You can keep your feed tool and add our work on top.
FEED MANAGEMENT TOOL
What a feed management tool does
Maps fields from your source feed to each output feed, per market
Applies rules and hardcoded overrides that hold until someone changes them
Schedules updates and keeps every destination in sync
Flags feed health: errors, disapprovals and missing required fields
QUANTUM BY S360
What Quantum adds
Generates the attributes that don’t exist in your source at all
Creates variants you can test and revert, without touching the master feed
Measures what each change was worth, against a real control group
Comes with an s360 specialist who decides what to optimise and test next
CLIENT RESULTS
Our agency has optimised more than 300 product feeds for brands and retailers across Europe.


JYSK
JYSK: 20 million additional clicks
JYSK is one of Europe’s largest retail chains, with more than 3,300 stores in over 50 countries. We deployed Quantum as a layer between JYSK’s master feed and its advertising channels, which gave the marketing team a sandbox to enrich product data, test it and scale only what worked. Based on tests across 29 markets, the winning changes are estimated to add 20 million clicks a year. The work won Best Global PPC Campaign at the Global Search Awards 2025 and was published as an AI Excellence case on Think with Google.


REVOLUTIONRACE
RevolutionRace: 21% more clicks and 33% more conversions
RevolutionRace is one of Sweden’s fastest-growing outdoor clothing brands, with SEK 2 billion in turnover and more than 40 markets. Most feeds stack technical attributes in the title and trust the algorithm. We replaced the specs with benefit-led titles in the words shoppers use, and tested them in four markets against a control group. Clicks rose 21% and conversions 33%, without more budget. Based on the results, RevolutionRace appointed s360 to run the feed optimisation rollout across 20+ markets. The work won Paid Search Campaign of the Year at the European Paid Media Awards 2026.
Andreas Castillo Kristensen, Head of Digital Media, JYSK
WHAT YOU GET

We do the work
You work with an agency team rather than buying another software licence. Our feed specialists operate Quantum and make the changes, and they work alongside s360’s paid search, paid social and SEO teams, so feed changes and campaign changes are planned together. There’s no new tool for your team and nothing for your developers to build.
A feed that doesn’t go stale
New products are optimised as they enter the feed, and existing products are re-optimised when they change.
Proof that it worked
Significant changes run as a split test before rollout. The read-out shows the measured effect with a confidence interval.
A feed ready for AI search
Conversational attributes, structured product detail and Q&A in place, so your products can answer the questions shoppers ask.
GETTING STARTED
The only things we need from you are access and approval of the first batch. We handle everything else.

Week 0
Kick-off, and you grant access
You grant access
Week 1
Audit, hypothesis mapping and first read-out
Weeks 2–3
First batch optimised and revised, then approved by you
You approve the first batch
Week 4
Changes live in the feed and the experiment started
Weeks 5–10
Experiment runs, followed by a read-out of the results
SCOPE
Image optimisation can be added to either scope. Both are priced per market, and additional markets get a discount.

OPTIMISATION
Feed Optimisation
Every text attribute optimised and kept current, including core, additional and conversational attributes. New products are optimised automatically and existing products refreshed when they change. It doesn’t include split testing.
RECOMMENDED
Feed Optimisation + Test & Learn
The same work, plus proof of what each change was worth. Every major change runs as a FeedX split test, effects are reported with confidence intervals, and an s360 specialist owns your testing roadmap.
QUESTIONS & ANSWERS
What does the free feed audit include?
+
Do we need to change our master feed or PIM?
+
Do we have to replace our feed management tool?
+
Does our team need to learn Quantum?
+
How do you know a change worked?
+

FREE FEED AUDIT
We send it within three days of getting access, and there’s no commitment. It answers three questions:
How complete is your coverage of core, additional, image and conversational attributes?
What’s the quality and depth of the text and images in the attributes you have?
What could optimised attributes look like for your products?