For retailers whose paid search depends on a file nobody owns

Google Shopping Feeds: Why Your Best Products Quietly Stop Showing

A shopping feed is an export of your catalogue that decides which products can be advertised, at what price, and whether the shop down the road is shown as having stock. It is usually set up once during a website build and then left alone, which is why so many retailers are paying an agency to manage campaigns for a catalogue that is a third disapproved.

Feeds fail silently. A supplier renames a field, a price changes on the site but not in the file, an image is replaced, availability drifts, and the items disappear from results with no alert to anyone. This page covers what a feed actually needs, what Australian pricing and tax rules mean for it, how local inventory works, and how to keep the whole thing approved.

Book a Feed Review (03) 9003 0111

Realistic ROI

10 to 30%

Of items disapproved in a neglected feed

Common range when nobody has reviewed the account in a year, and the items lost are not random

Price and stock

The two mismatches that cause most suspensions

The advertised value has to match the landing page, including GST and any sale price

Daily at minimum

Refresh cadence for a trading catalogue

Local inventory needs more frequent updates than that if stores sell the same stock

20 to 40 attributes

In a feed that is doing real work

Beyond the required set, the optional ones are what allow sensible campaign segmentation

Four Things That Decide Whether a Feed Performs

Feed problems are rarely about the advertising platform. They are about the data going into it and how often it is refreshed.

Product identifiers that are actually correct

Barcodes and manufacturer part numbers are how a listing is matched to a product the platform already understands, which affects both eligibility and how competitive the listing looks. Australian retailers frequently carry a mix of branded goods with valid barcodes, private label items with none, and imported lines where a supplier has supplied the wrong number entirely. Each of those needs different handling, and guessing produces disapprovals.

Price that matches the landing page exactly

The advertised price has to be the price a customer sees, in AUD, inclusive of GST, and a sale price has to be genuine. Mismatches happen when the feed is built from the ERP while the site prices from a promotions engine, or when a rounding rule differs by a cent. This is the most common cause of item level disapproval and account level warnings, and it is entirely preventable by sourcing both from the same place.

Availability that keeps up with trading

Advertising an item that sold out an hour ago costs money twice, once for the click and once for the customer experience. Availability has to be driven from real stock, with the same buffers and exclusions you use everywhere else, and it has to refresh often enough that a fast selling line does not stay live for a day after the last unit went.

Labels that let you segment spend sensibly

Custom labels are the difference between bidding on a catalogue and bidding on a strategy. Margin band, stock cover, season, clearance status, supplier and whether an item is available in store are all things the marketing team wants to bid differently on, and none of them exist in the platform unless the feed puts them there.

What a Well Built Feed Pipeline Does

The feed is an output. These are the six jobs the pipeline behind it has to perform.

One catalogue

Source and consolidate

Titles and descriptions from the product information source, price from the system that owns pricing, stock from the system that owns stock, images from wherever they are hosted. The feed is assembled from the owning system for each attribute rather than scraped from the website, which is what keeps it consistent with what a customer will see.

Attribute quality

Enrich and normalise

Titles built to a rule rather than by hand, brand and category mapped, sizes and colours normalised into the values the channel expects, variants grouped so a size run appears as one product family. This is where a mediocre catalogue becomes a competitive one, and it is unglamorous work that pays for itself in impression share.

Pre flight checks

Validate before sending

Missing images, prices of zero, broken links, invalid barcodes, descriptions with retailer only notes in them, and items that should never be advertised at all. Catching these before the file is published is far cheaper than discovering them as disapprovals a week later.

Scheduled delta

Publish and refresh

A full catalogue file on a schedule plus frequent updates for the fields that move, which are price and availability. Sending everything constantly is wasteful and slow. Sending only changes keeps the channel current without pushing a hundred thousand unchanged rows every hour.

In store availability

Local inventory by store

A second feed carries quantity by store code so a shopper nearby can see that an item is in stock at a particular shop. Store codes have to match the business listing exactly, quantities need the same buffers you apply online, and stores that cannot support the promise should be excluded rather than optimistically included.

Disapproval alerts

Monitor and report

Item level disapprovals, sudden drops in the number of active items, and price or availability mismatches reported to a named person the day they occur. A feed with no monitoring is a slow leak: nobody notices until the revenue report does, and by then the cause is a month old.

Feed Problems and What Should Happen Instead

Task	Traditional	Feed Done Properly	Notes
Item sells out mid morning	Still advertised until tonight	Availability updated within the hour	Frequent availability refresh is cheaper than the wasted clicks and the disappointed visits.
Promotion starts at midnight	Feed price lags the site	Sale price and dates published together	Feed and site must price from the same source or the mismatch triggers disapproval on your best sellers.
Private label with no barcode	Disapproved as missing identifier	Flagged as having no identifier	There is a correct way to declare this. Inventing a number is worse than declaring its absence.
Supplier changes a product image	Broken link, item drops out	Validation catches it before publish	Image checks are the cheapest pre flight test in the pipeline and catch a surprising share of failures.
Clearance lines need harder bidding	Marketing guesses from a spreadsheet	Custom label carries stock cover and age	Segmenting by real inventory data is where feed work turns into margin rather than tidiness.
Shopper near a store	Only the online offer shown	In store availability shown by location	Requires accurate store level counts, so pair it with the same buffers used for collection orders.
Size and colour variants	Listed as unrelated products	Grouped into one product family	Grouping improves how the listing is presented and stops variants competing with each other.
Discontinued line still live	Advertised for weeks	Removed on status change	The feed should reflect lifecycle status from the owning system, not a manual delete list.

Where Feed Integrations Quietly Cost Money

The feed and the website price from different systems

If the site applies promotions in one engine and the feed is exported from another, the two will disagree the moment a discount runs. Pick one source of truth for the customer facing price, feed both the site and the file from it, and include GST the same way in each. Nearly every price mismatch suspension traces back to this one decision.

Sale pricing that does not hold up under the Australian Consumer Law

A was price has to be a price the item was genuinely sold at for a reasonable period, and a discount claim has to be accurate. Feeds make this visible and comparable at scale, so a sloppy promotions process becomes a public one. Keep the price history that supports the claim, and build the feed so the sale price and its dates come from the promotion record rather than being typed in.

Availability with no buffer, on a shared stock pool

If shops, marketplaces and the website draw from the same stock, advertising the last unit invites an oversell and a cancellation. Apply the same per location buffers to the feed as you do to the storefront, and be more conservative for local inventory, because a customer who drives to a shop for an advertised item and finds it gone is a worse outcome than a delayed delivery.

Store codes that drift from the business listing

Local inventory only works while the store identifiers in the feed match the verified locations exactly. Openings, closures, relocations and renamed centres all break the link, usually without an obvious error. Treat the store list as controlled data with an owner, and reconcile it against the feed on a schedule rather than after a quiet quarter.

Nobody is named as the owner of the feed

Feeds sit between marketing, ecommerce and whoever manages the ERP, which in practice means nobody. Disapprovals accumulate, the active item count drifts down, and the agency reports on the campaigns rather than the catalogue behind them. Name an owner, give them a weekly report that shows active items and disapproval reasons, and the problem stops being invisible.

Feeds built by scraping your own website

Generating a feed by crawling the storefront seems convenient and creates a fragile dependency on page structure, plus it inherits any display rounding or personalised pricing. Build from the systems that own the data through supported interfaces. If a platform genuinely has no other way out, that is a reason to schedule the export carefully and monitor it closely, not to pretend it is robust.

How Yes AI Approaches Feed Work

An audit before a build

We review the current feed against your live catalogue and report what is disapproved, what is missing identifiers, what is mispriced and what is advertised while out of stock. Often the first fix is a configuration change rather than a project, and we will tell you when that is the case.

One pipeline, several channels

The same enriched catalogue can serve shopping feeds, marketplaces and your own site search. Building the enrichment once and publishing to each channel in its own format is far cheaper than maintaining a separate export per destination, and it stops the channels disagreeing.

Hosted and monitored by us

Feed generation runs on a managed cloud automation layer we operate, with validation before publish and alerting when item counts move sharply or disapprovals spike. You are not depending on a scheduled task on somebody’s laptop.

Labels built for how you actually bid

We work with your marketing team to define custom labels that reflect margin, stock cover, season and store availability, so campaign structure can follow commercial reality rather than category names.

From Neglected Export to a Maintained Feed

Five steps. Most retailers see the disapproval count fall within the first fortnight.

Audit the current state

Active versus disapproved items, mismatch warnings, missing identifiers, stale availability and how the current file is produced. You get the findings whether or not we build anything.

Agree the source of truth per attribute

Which system owns the title, the price, the image, the stock figure and the lifecycle status. This is the decision that keeps the feed and the website in agreement once promotions start.

Build enrichment and validation

Title rules, category and attribute mapping, variant grouping, custom labels, and a pre flight check that blocks a publish when something is obviously wrong.

Publish, then add local inventory

The main catalogue feed goes live on a schedule with frequent price and availability updates. Once store counts are trustworthy, we add quantity by store so nearby shoppers see in store availability.

Monitor and tune

Weekly reporting on active items and disapproval reasons, alerting when counts move sharply, and periodic revision of labels and rules as the range and the trading calendar change.

FAQ

What is a product feed?

A product feed is a structured file or data connection that sends your catalogue to an advertising or shopping channel: identifiers, titles, descriptions, images, prices, availability and shipping information for every item you want to advertise. The channel uses it to decide what to show, to whom, and at what price. Because it is generated rather than typed, its quality is a direct reflection of the systems behind it, which is why feed work is really data integration work.

How often should a shopping feed be updated?

The full catalogue at least daily, and price and availability far more often than that, ideally within the hour for a fast moving range. The right cadence depends on how quickly your stock turns and whether shops and marketplaces share the same pool. Local inventory by store needs the tightest cadence of all, because the promise being made is that a customer can walk in and buy the item today.

Why do our products keep getting disapproved?

The usual causes are a price in the feed that does not match the landing page, missing or invalid product identifiers, images that do not meet requirements or have moved, availability that contradicts the site, missing shipping or tax configuration, and policy issues on restricted categories. Each has a specific fix, and the important part is having the disapproval reasons reported to a named person weekly so the pattern is visible rather than the total.

What are local inventory ads and do we need stores for them?

Local inventory listings show a nearby shopper that an item is in stock at one of your physical shops, using a feed that carries quantity by store code alongside the main catalogue. They only make sense if you have shops and if your store level counts are reliable, because the whole value is the promise that the item is there now. Retailers with weak store counts should fix counting and put buffers in place before turning this on, not after.

Can we run one feed pipeline for several channels?

Yes, and it is usually the right design. The expensive part is the enrichment: consistent titles, mapped categories, normalised attributes, grouped variants and useful labels. That work should be done once against a single catalogue, then published in the format each destination expects, whether that is a shopping channel, a marketplace or your own site. Maintaining a separate hand built export per channel is how the channels end up disagreeing about your prices.

Should the feed come from our website or our ERP?

From whichever system owns each attribute, which in practice means several sources assembled by an integration layer. Price usually comes from the system that sets pricing, stock from the system that owns stock, and rich content from wherever product information is maintained. Exporting purely from the website is tempting because everything is in one place, but it inherits display formatting and any personalised or cached pricing, and it breaks whenever the theme changes.

What does feed integration cost to run?

There is a build cost that depends on how many systems the catalogue is assembled from and how much enrichment is needed, then an ongoing cost for hosting, monitoring and rule changes as your range and calendar change. If your catalogue is small, stable and already clean, an off the shelf app from your ecommerce platform may be all you need, and we will say so. The case for a built pipeline is strongest when product data lives in several systems or when store level inventory is involved.

Find Out What Your Feed Is Really Doing

Book a call. We will audit your current feed against your live catalogue, show you what is disapproved and why, and give you a priced plan to fix it. The audit is yours either way.