Managing Price, Inventory, Variants, and Promotions in AI Shopping

Executive Summary

In AI-powered shopping experiences, the accuracy of commercial product information is just as important as product discovery. Even when a product matches the shopper’s needs, the purchasing journey can break down if its price is outdated, the relevant variant is unavailable, or the promotion terms are represented incorrectly.

Brands need to manage four core data layers together:

Validating regular and sale prices by channel, country, and currency
Keeping availability current for every purchasable variant
Grouping variants under the correct parent product with unique identifiers
Structuring promotions with accurate dates, benefits, eligibility rules, and product targeting

At a more advanced level, a daily full feed may not be sufficient on its own. Fast-changing fields such as price and availability should be supported by intraday API updates, change logs, automated checks, and checkout validation as part of a hybrid data architecture.

Why Is Commercial Data Critical in AI Shopping?

In a traditional advertising journey, shoppers usually visit the website after seeing a product and verify its current price, availability, and promotion terms on the product page.

In AI shopping, this information can influence the decision much earlier in the journey.

A shopper might ask an AI assistant:

Which coffee machines under TRY 5,000 are currently in stock?
Which models are available in black and can be delivered tomorrow?
Is size 42 of this shoe currently on sale?
Do I qualify for free shipping if I buy two items?
Which colors and sizes are included in the promotion?

Product titles and descriptions alone cannot answer these questions. Price, inventory, variant, and promotion data must be connected, current, and verifiable.

These four areas should not be managed as separate files or isolated operations. Together, they form a single data system that represents the product’s current commercial status.

1. Price Management: One Price Field Is Not Enough

The most basic requirement is to send the product’s regular price to the feed in the correct currency.

In real ecommerce operations, however, pricing is more complex and may include:

Regular selling price
Sale price
Promotional price
Member price
Country- or region-specific price
Variant-level price
Unit price
Tax-inclusive or tax-exclusive price
Prices valid only within a defined date range

These values must be managed without creating conflicts for the same product.

Regular Price vs. Sale Price

The regular price represents the product’s standard selling price outside a promotion. The sale price represents a lower price that is valid for a specified period.

When a sale price is used, its start and end dates should also be defined. This prevents an expired discount from continuing to appear in AI shopping experiences.

In OpenAI’s product feed structure, regular price, sale price, and the sale period can be managed through separate fields. Prices must also be submitted with the relevant currency code.

Consistency Across the Feed, Product Page, and Checkout

The shopper should see the same price at three key points:

Product feed
Product detail page
Checkout

For example, if a product appears as TRY 1,799 in the feed but increases to TRY 1,999 at checkout, the issue goes beyond a technical mismatch. It can also damage the shopper’s trust in the brand.

Variant-Level Pricing

Not every variant of a product will necessarily have the same price.

A leather sofa may cost more than its fabric version, while a phone with 1 TB of storage may cost more than the 512 GB model. In these cases, submitting only a parent-level price can create misleading results.

Each purchasable variant should have its own price, and shoppers should see that same price when they are directed to the selected option.

Multi-Country Price Management

Brands selling across multiple countries also need to manage:

Currency
Tax structure
Regional pricing
Local discounts
Country-specific product eligibility
Regional inventory
Shipping costs

Applying an exchange rate is not always enough. Local pricing strategies, rounding rules, and country-specific promotions should also be taken into account.

2. Inventory Management: More Than “In Stock” or “Out of Stock”

Inventory is one of the fastest-changing data areas in an AI shopping experience.

Within minutes, a product may become:

In stock
Out of stock
Available for pre-order
Temporarily unavailable
On backorder
Available only in selected stores

Availability should therefore not be treated as a simple yes-or-no field. For pre-order and backorder items, brands should also communicate when the product is expected to become available.

At What Level Should Availability Be Managed?

Availability should be maintained at the purchasable variant level rather than only at the parent-product level.

For example, a pair of sneakers may have the following availability:

Black, size 41: in stock
Black, size 42: out of stock
White, size 42: in stock

Marking the parent product simply as “in stock” could provide incorrect information to a shopper looking specifically for the black model in size 42.

Using Safety Stock

Physical inventory does not always need to match the sellable inventory exposed to external channels.

For example, a warehouse may hold five units while the brand makes only three available for sale to reduce the risk of overselling across simultaneous orders.

This approach can be useful for:

Products that sell out quickly
Products sold through multiple channels
Brands sharing inventory between marketplaces and their own website
Systems where inventory updates are delayed
Situations where the final unit may be damaged or reserved

The availability sent to the feed should therefore be calculated using sellable inventory rules rather than copied directly from the physical warehouse quantity.

Should Products Be Removed or Marked as Out of Stock?

Immediately removing a temporarily unavailable product from the feed is not always the right approach.

If the product will return, keeping its record and updating its availability helps preserve product identity and continuity.

For permanently discontinued products, however, it may be better to remove the item from the catalog or exclude it from the relevant channel instead of leaving it marked as out of stock indefinitely.

3. Variant Management: Parent Products and Purchasable Options

Variant management is one of the most commonly misunderstood areas of AI shopping readiness.

A parent product represents a product family. A variant is the specific option that a shopper can actually purchase.

Variants may differ by:

Color
Size
Material
Capacity
Storage
Dimensions
Pack quantity
Model or technical configuration

Each variant should have a unique and stable product ID. Variants belonging to the same product family should be connected through a shared group identifier.

The Parent Product Should Not Be a Separate Purchasable Record

If the parent product cannot be purchased directly, it should not be submitted to the feed as an independent variant.

For example, the parent record for a T-shirt could be TSHIRT-100, with the following variants:

TSHIRT-100-BLK-S
TSHIRT-100-BLK-M
TSHIRT-100-WHT-S

Here, the parent identifier groups the related options, while the variant identifiers represent the individual purchasable products.

Variant-Specific Fields

If the following information differs between variants, it should be submitted separately for each one:

Price
Availability
Title
Product URL
Image
Description
GTIN or MPN
Color, size, and material
Delivery eligibility

If selecting a black variant displays an image of the white product, or an unavailable size still appears purchasable, the variant structure has not been configured correctly.

Why Should Variant IDs Remain Stable?

Frequently changing variant IDs can cause:

Fragmented product history
Loss of performance data
Reviews being matched to the wrong product
Broken promotion links
Existing products being interpreted as new records

The product identity should remain unchanged when the title, price, or availability changes. A new ID should be created only when a genuinely new purchasable product is introduced.

4. Promotion Management: More Than a Sale Price

Not every promotion is simply a discounted price.

Promotions may offer:

Percentage discounts
Fixed-amount discounts
Free shipping
Discounts above a minimum order value
Coupon codes
Discounts on a second item
Multi-buy offers
Offers limited to selected variants

Promotion data should therefore not be managed simply by adding promotional language to the product title.

What Should a Promotion Record Include?

A structured promotion record should include at least:

A unique promotion ID
Promotion title
Start and end times
Promotion status
The benefit offered
Target products
Target variants
Redemption conditions
Coupon requirements
Promotion URL

Promotions can represent different benefits, including fixed-amount discounts, percentage discounts, or free shipping. They may also apply to specific products or only to selected variants.

Promotion Eligibility Should Be Checked at Variant Level

A promotion may not apply to every variant within a product family.

For example, it may be limited to:

The red color option
The 512 GB model
Selected sizes
Previous-season variants
Products from specific sellers

If a promotion is attached only to the parent product and then applied to every variant, the shopper may discover at checkout that the selected option is not eligible.

Promotion targeting should therefore support variant IDs as well as product IDs.

Expired Promotions

Promotions should not only activate automatically at their start time. They should also be deactivated automatically when they expire.

To prevent expired offers from continuing to appear:

Start and end times should use a standard date and time format
The applicable time zone should be defined clearly
Promotion status should update automatically
The product feed and promotion system should be checked together
Eligibility should be validated again at checkout

When a promotion ends, the product’s regular price should automatically become active again.

How Should Price, Inventory, Variants, and Promotions Be Managed Together?

These data areas may originate from different systems:

Pricing from an ERP or pricing engine
Inventory from a warehouse management system
Variant structures from a PIM or ecommerce platform
Promotion data from a campaign engine
Checkout results from the ecommerce platform

The key requirement is to connect these sources through shared and stable product identifiers.

For every purchasable variant, a single record should be able to answer:

Which exact variant is this?
What is its current price?
Does it have a sale price?
Which currency is it sold in?
Is it currently available for purchase?
Which promotions apply to it?
Is the shopper being directed to the correct product page?
Are the same terms valid at checkout?

Checklist for Brands

Price

Are regular and sale prices stored in separate fields?
Is the currency correct?
Do prices match across the feed, product page, and checkout?
Are prices managed at variant level?
Are promotion start and end dates defined?
Are country-specific price differences supported?

Inventory

Is availability managed separately for each variant?
Are pre-order and backorder statuses defined?
Are sellable inventory rules applied?
Are inventory changes distributed quickly enough?
Are permanently discontinued products removed from the catalog?

Variants

Does every purchasable option have a unique ID?
Do variants within the same product family share a group ID?
Are price, availability, images, and URLs variant-specific?
Are GTIN and MPN values linked to the correct variants?
Do variant IDs remain stable over time?

Promotions

Does every promotion have a unique ID?
Are start and end times defined?
Is the benefit type stated clearly?
Are the correct products and variants targeted?
Can coupon and minimum-order requirements be managed?
Are expired promotions deactivated automatically?
Is promotion eligibility validated at checkout?

How Can Optifeed Support Brands?

At Optifeed, we help brands manage their price, inventory, variant, and promotion data in a way that remains current, consistent, and compatible with different platforms.

Our support includes:

Creating regular and sale price rules
Automatically transferring availability data from source systems
Structuring variant and parent-product relationships
Validating product IDs, GTINs, and MPNs
Adapting feeds by country and currency
Matching promotions with eligible products
Building full-feed and partial-update workflows
Identifying inconsistencies across the feed, product pages, and checkout
Preparing channel-specific outputs for Google, Meta, TikTok, marketplaces, and AI platforms

Conclusion

Recommending the right product is not enough in AI shopping. The recommended variant must actually be available, its price must remain consistent through checkout, and the promotion terms must be applied correctly.

Price, inventory, variant, and promotion management should therefore not be treated as catalog operations alone. Together, these data areas form the commercial foundation of trust between product discovery and purchase.