You do not need a rule that says every product variant must receive advertising spend. You need a clear reason for including or excluding each type of offer. A size that matches a customer's search, a colour with strong demand, and a costly made-to-order configuration are different decisions.

For a small store, the useful question is not simply “all variants or best sellers?” It is “which variants can we present accurately, fulfil reliably, and advertise within our economics?” This guide separates product data from campaign selection, then shows how to build a test without making your catalogue misleading.

Start with three separate decisions

What exists in your catalogue? These are the products and variants a customer can actually order. The information should describe the real offer, including its price, stock status, and distinguishing attributes.

What do you submit to Merchant Center? This is your product data for the relevant marketing methods. Submitting an offer is not a promise that it will receive impressions, clicks, or a particular share of budget.

What do you include in a campaign? This determines the inventory eligible within that campaign. A product can remain in your catalogue while being excluded from a particular advertising test. Removing an offer from Merchant Center has a different scope from excluding it in one campaign.

Keep these decisions distinct in your notes and reporting. Otherwise, “we stopped advertising the large sizes” can turn out to mean that someone removed useful product information from every destination.

When advertising a broad set of variants makes sense

A broad selection can be useful when customers search for particular attributes and you can fulfil those choices. A shopper looking for a green jacket in a specific size has a different requirement from someone who wants any jacket in the collection.

The case is stronger when the variants have similar fulfilment costs, clear landing pages, accurate images, and reliable stock. You can then evaluate performance without every variation introducing a different operational problem.

It can also be reasonable to include variants with little history. A new colour is not automatically a poor candidate because it has never sold through Shopping. Lack of historical spend is not the same as evidence of poor demand.

Broad inclusion does not mean equal spend. Avoid dividing your daily budget by the number of variants and treating the result as an expected allowance per item. Actual delivery and cost will not necessarily follow that calculation.

When a narrower variant selection is worth testing

Selection becomes more important when variants have materially different economics or buying experiences. Think about a standard item with a premium material upgrade, a much larger version that needs freight delivery, or a personalised product with many possible configurations.

In those cases, the cheapest variant is not necessarily the best advertising candidate. It may leave little contribution after fulfilment, attract a different customer, or create expectations that the more expensive configuration cannot satisfy.

Consider testing a narrower set when you can identify a specific reason:

• Some configurations have weak contribution after fulfilment and expected returns.
• The available stock is too limited or replenishment is too uncertain for the planned promotion.
• A variant's images or landing-page selection do not represent the advertised offer clearly.
• The product has a very large number of custom combinations, and a smaller set better represents what customers commonly buy.
• You have a defined test for a particular customer need, material, price range, or use case.

Google provides specific advice for advertising customised products, including considering readily available and best-selling configurations when the possible combinations are too numerous. That guidance concerns customised products. It is not a blanket instruction to remove ordinary sizes and colours from every store.

Example 1: a clothing store with sizes and colours

Imagine a jacket in four colours and six sizes. All combinations are regular products, but their stock and return patterns differ. The store initially considers advertising only medium and large in black because those variants sold most last season.

That choice may simplify the test, but it also excludes customers looking for other sizes and colours. Before applying it, ask whether the past sales mix reflects demand or simply past availability. A size that was sold out for most of the season cannot be compared fairly with one that was always available.

A more informative approach is to review the model as a whole, then inspect variants. Keep track of stock history, clicks, orders, and expected contribution. If you restrict certain variants, record the reason and a review date. Do not turn a temporary stock problem into a permanent assumption about demand.

Also avoid describing a model as fully available when only one unusual size remains. That situation may affect the usefulness of your landing page even when the remaining offer is technically purchasable. Review what a shopper actually sees after choosing the advertised variant.

Example 2: furniture with very different configurations

Now consider a shelving unit in several widths and finishes. The smallest configuration costs £120. A larger version costs £320 and has a higher delivery charge. Some finishes are stocked; others are made to order.

Advertising only the £120 version can attract shoppers who will never consider the £320 version. Advertising every configuration without checking the economics can also be expensive. The right test needs to reflect the differences in price, delivery, availability, and intent.

You could start with a small set of representative configurations that have a clear commercial case. Each advert should lead to the configuration it describes. The small version's price should not be used to imply that the larger version is available for the same amount.

Compare the resulting orders with your original hypothesis. Did the low-priced configuration create profitable baskets, or mostly interest in an offer that customers abandoned? Did a more expensive variant attract fewer clicks but leave more contribution per order? The answer should guide the next selection.

Do not merge variants to make the catalogue look smaller

Campaign simplification should not come at the expense of accurate product identity. Google's item group ID specification explains how separate variant offers are linked through a shared group identifier while keeping their own IDs.

A parent product is useful for understanding the family. The individual offer is useful for identifying the exact choice. Mixing those levels can make stock, price, and performance harder to interpret.

Keep identifiers stable when the same offer continues to exist. Record which IDs belong to which model and which attributes distinguish the variants. If you need fewer products in a campaign, use campaign inventory controls rather than rewriting several real variants as one ambiguous offer.

Bundles and multipacks raise a different question: what exactly is being sold together? They should not be treated as ordinary colour or size choices. Our guide to multipack versus is_bundle explains that distinction.

Selecting variants in Standard Shopping and Performance Max

In Standard Shopping, product groups organise campaign inventory. In Performance Max with a product catalogue, listing groups specify the included product selection within an asset group. Google documents these separately in its guides to Shopping product groups and Performance Max listing groups.

Selection and bidding are not interchangeable. In particular, do not assume that a manual bidding workflow from Standard Shopping works the same way in Performance Max. Define the commercial objective first, then check the controls available for your campaign type.

Before judging the effect of an exclusion, check whether the same offer remains eligible in another campaign. Also review any campaign features that can send traffic to additional pages. “Excluded from this listing group” is not a complete audit of every way the business might advertise that product.

Where useful, keep a stable internal classification for test candidates, established products, and offers held back for an identified reason. The classification should explain your process. It should not pretend to be a quality score supplied by Google.

Never use false stock status as an advertising switch

If a variant is available but you do not want to advertise it, control its advertising eligibility. Do not label it out of stock to avoid spending money. That changes a factual statement about the offer rather than simply changing your campaign selection.

Likewise, a sold-out variant should not remain available in the data just because the parent model has stock in another size. Google's availability requirements distinguish whether the actual offer can be ordered. Keep the selected variant, its price, and its purchase conditions consistent across the customer journey.

Run a simple manual check before a test. Open an included variant on mobile, confirm its selection, inspect its image and price, and continue towards checkout. Repeat for a few variants at the edges of the range, such as the largest size or the premium finish.

How to judge the test without overreacting to small numbers

Write down the reason for your selection before changing it. For example: “These stocked finishes have similar fulfilment costs and enough contribution to justify the target acquisition cost.” This is a testable assumption. “Google prefers fewer variants” is too broad to be useful without evidence from your account.

Set a review window that reflects the buying cycle and conversion delay. Decide in advance what amount of spend you can afford to risk while learning. There is no universal click count or number of days that makes a variant's result reliable.

Review both the variant and its product family. One purchase may be too little information to justify promoting a size permanently. A family-level view can show whether the overall model is viable, while variant-level details identify stock or offer problems.

Then distinguish the product advertised from the product purchased. With the appropriate implementation, conversions with cart data can help explain baskets and cross-selling. Without that detail, do not assume every attributed order contained the exact variant that received the click.

Compare contribution as well as revenue. A higher-priced variant may produce a larger order value while also costing more to ship and return. A lower-priced variant may attract more purchases but leave too little money after advertising. Neither wins on price or conversion count alone.

A practical checklist for your next variant review

1. Separate the real catalogue, Merchant Center submission, and campaign selection.
2. List the attributes that make each offer different to a shopper.
3. Check contribution, stock reliability, delivery conditions, and landing-page accuracy.
4. Decide whether broad inclusion or a narrower selection answers the current business question.
5. Record exclusion reasons and review dates.
6. Verify eligibility across campaigns before interpreting a change.
7. Evaluate purchases, basket contribution, and product-family performance after allowing for conversion delay.

The goal is a selection you can explain and improve. Keep useful variants when they serve a real customer need. Restrict offers when there is a commercial or operational reason to do so, and test the assumption rather than treating it as permanent.

If the harder question is whether your prices are competitive enough to advertise at all, read Google Shopping when you are not the cheapest. That is a separate decision from how many variants to include.