Retail advertising with stock, margin and fulfillment controlsРеклама розницы с реальными остатками и маржой

A shop can have a successful-looking product campaign and still spend money sending shoppers to the wrong size, an already reserved item or an order that costs too much to fulfill. The useful control is the saleable offer: one variant, one location or fulfillment route, a current price and a promise the shop can keep.

This guide shows a small retailer how to decide which offers are safe to promote, respond when stock becomes uncertain and evaluate the contribution left after fulfillment. It uses the fictional outdoor-clothing shop North Window. Every quantity, price, timestamp and outcome below is illustrative, not a SABSUS customer result or an industry benchmark. Platform sources were checked on October 7, 2026.

Write the advertised promise before choosing a campaign. “Navy rain jacket, medium, available for delivery in the supported area” is an offer that can be checked. “Rain jackets available” hides the difference between sizes, colors and fulfillment locations.

Give every sellable variant a stable internal SKU and a mapping to the provider's product identifier. Keep the group identifier separate. Google Merchant Center groups variants with an item group ID while retaining an individual product ID for each variant. Google: item group ID.

For each promoted offer, record:

• SKU, size, color and the exact landing-page variant
• Stock location and permitted delivery or pickup route
• Sellable quantity, existing reservations and the time the quantity was verified
• Current selling price, promotion end time and currency
• Expected contribution after variable fulfillment costs
• Person responsible for a mismatch and the advertising object they may change

Do not treat stock in a distant branch as locally available unless a transfer or delivery can meet the advertised promise. A product family may remain advertised when one variant is unavailable, but its creative and landing page must not imply that every size is ready. Open the destination from a phone and check the selected variant rather than only the default product page.

North Window has 18 navy medium jackets physically recorded at one warehouse. Six are reserved for unpaid but still valid orders, two are quarantined after a damage check, and the shop keeps a two-unit operating buffer. The reserved, quarantined and buffer groups do not overlap, and all ten deducted jackets are included in the 18 on hand.

Available to promise = 18 on hand − 6 reserved − 2 quarantined − 2 buffer = 8 jackets.

This is an operational advertising limit, not a universal inventory formula. If your stock system's “available” field already subtracts reservations, do not subtract them again. Document whether physical stock includes damaged goods, whether reservations expire and whether pending store purchases have already reduced availability. A buffer is a deliberate policy choice, not extra stock that another channel may also promise.

A return is not saleable merely because a refund was issued. Wait for receipt and inspection. A purchase order is not received stock. If backorders are allowed, manage their terms and dates separately; do not silently add expected receipts to today's ready-to-ship quantity.

The same eight jackets may be exposed through search, social, marketplace and walk-in sales. They remain eight shared units. Do not allocate eight to each channel without a reservation mechanism. When simultaneous checkout is possible, the final order acceptance step must resolve the last-unit conflict independently of ad targeting.

A fresh file does not prove that the underlying stock count is fresh. Keep separate timestamps for the stock observation, the exported record, the provider's processing status and a destination-page check. Use one time zone in the review record.

For North Window's fast-moving promotion, the owner chooses a 15-minute maximum age for a trusted stock observation. This is a fictional operating threshold. A slower catalog might justify a different limit; a last-item sale might need stricter handling.

Suppose stock was observed at 10:00, a file was generated at 10:06 and provider processing was still unconfirmed at 10:18. At 10:18 the observation is 18 minutes old, so the chosen threshold has been exceeded. At 10:18, the export file is 12 minutes old, but the underlying stock observation is still 18 minutes old.

Review the slowest part of the chain: stock collection, export, ingestion, destination update and human response. Log failed updates and missing timestamps explicitly. “Unknown” should remain unknown until checked.

Google's automatic product updates can use website information to correct certain discrepancies, but Google says these updates do not replace regular product-data updates. Treat them as a supplementary safeguard, not an inventory reservation service or a guaranteed synchronization interval. Google: automatic product updates.

These are different decisions. Availability describes whether the customer can buy the item. Price describes the actual offer. Visibility controls where the offer is promoted.

For Google product data, variant availability should agree with the destination, and price and currency should match the relevant landing-page and checkout offer. Review the applicable country and program rules instead of assuming every shopping or local-inventory surface works identically. Google: availability, Google: price.

A promotion may become unattractive while the item is still genuinely in stock. Use an appropriate, authorized advertising visibility control in that situation. Do not label the item unavailable solely to manage advertising economics. Google documents separate visibility and pause controls; the correct mechanism depends on the destination and intended duration. Google: manage product visibility.

Before a price change, check the offer record, page, structured data, checkout, discount conditions and scheduled end time. A banner saying “20% off” should not lead to an undisclosed minimum quantity or an expired checkout price. After the change, sample the actual variant path, including shipping restrictions. An accepted feed upload alone is insufficient evidence that all these surfaces agree.

Use a comparable unit and tax basis. The following USD teaching model excludes taxes collected for government and assumes one jacket per completed, retained order. Expected return-related losses are a planning allowance allocated across retained orders, including the unrecovered handling or product losses associated with returns. They are not a second deduction of the full selling price. Reconcile the allowance with actual outcomes later without charging the same loss twice.

At a $120 selling price, the fictional per-order amounts are:

• Product cost: $58
• Payment fee: $3
• Picking and packing: $6
• Shipping subsidy paid by the shop: $8
• Expected return-related loss: $5

Contribution before advertising = $120 − $58 − $3 − $6 − $8 − $5 = $40.

The owner wants to retain $16 per order toward overhead and operating surplus. The resulting advertising allowance is $40 − $16 = $24 per completed, retained order. This is not net profit and does not establish a $24 new-customer CAC unless the denominator really is unique new customers and the cost scope matches.

Now reduce the price to $100 while holding product cost, picking, shipping and return allowance constant. Assume the payment fee falls to $2.50. Contribution before advertising becomes $100 − $58 − $2.50 − $6 − $8 − $5 = $20.50. Keeping the same $16 reserve leaves only $4.50 for advertising per retained order. The discount did not preserve the previous advertising allowance.

Check the actual mix. A pickup order may have lower delivery cost and additional counter labor. A two-item basket needs an order-level calculation with shared costs allocated once. A high return rate in one size can make an aggregate product-family average misleading.

Stock does not describe packing capacity. On a teaching day, North Window's remaining preparation window has capacity for 24 one-jacket order equivalents before its promised cutoff. Fourteen are already committed and four are kept as an operating reserve. This leaves 24 − 14 − 4 = 6 incremental one-jacket orders for the same delivery promise.

There are eight available jackets but only six additional fulfillment slots. The immediate campaign planning ceiling is the smaller number: six one-jacket orders. With the $24 allowance above, six × $24 = $144 is a scenario-based spending envelope if all six eligible outcomes materialize. It is not a platform spending guarantee, a required budget or proof that six orders will be won.

Do not raise spend merely because the envelope is positive. Consider existing organic demand, pending carts, other channels and uncertainty in conversion. Check who can accept the next order at the cutoff. If tomorrow's capacity is available, advertise tomorrow's genuine promise rather than continuing to sell today's exhausted one.

Write a small operating playbook with a named owner:

• Verified and eligible: fresh stock, consistent offer, acceptable contribution and available fulfillment capacity. Promote only within approved limits.
• Low stock or near cutoff: recheck more often, narrow the promise and review the remaining allocation.
• Unknown or stale: investigate the stock source; request or execute only the already authorized visibility action. Do not invent a replacement stock value.
• Unavailable: update genuine availability consistently and remove the unavailable promise from the customer journey.
• Economically ineligible: retain truthful stock data while reviewing price, costs or advertising visibility.

A recommended action, a submitted request and a provider-confirmed status are different records. If a pause request times out, label its outcome unknown and inspect current provider state before retrying. Do not repeatedly toggle an object because one screen is delayed. Record the exact product, product group or campaign affected; broad pauses can stop valid offers too.

Restart only when the reason for stopping is resolved, the relevant provider state is checked and the restart is within authorization. A new delivery, an expired pause or a recovered feed does not by itself establish all three conditions.

In a separate fictional review cohort, $360 of advertising is associated with 20 placed one-jacket orders. Two are canceled before fulfillment. Eighteen ship; two of those are later returned and refunded. After the agreed return observation window, 16 completed, retained orders remain.

Cost per placed order = $360 ÷ 20 = $18. Cost per retained order = $360 ÷ 16 = $22.50. The $22.50 figure is the comparable measure for the $24 retained-order allowance. If all 16 buyers are not verified new customers, do not relabel it new-customer CAC.

This review cohort is separate from the six-slot planning day. Do not combine its quantities with that day's stock. Reconcile revenue, refunds and actual costs for the cohort; the planning allowance is not a substitute for that ledger. Record the unmatched orders and source limitations. Provider-attributed orders and observed business records alone do not prove that all 16 sales were caused by the campaign.

Before launch, test one normal variant, one unavailable variant, one promotion boundary and one deliberately stale input in a safe, non-live test. Confirm what the system proposes and who is allowed to act. If no test environment exists, document a supervised check without changing customer-facing stock to false values.

During the promotion, review exceptions when the business's chosen freshness limit or operational risk requires it. At closing, reconcile failed reservations, wrong-size landings, oversells, late dispatches and cancellations by cause. Weekly, compare actual contribution across variants and fulfillment routes using a mature cohort. An attractive click-through rate cannot repair a recurring oversell.

A practical first deliverable is a one-page offer-control sheet containing the chosen SKU, shared-stock definition, maximum data age, remaining packing slots, per-order allowance, stop condition, approver and restart test. Start with a manageable set of offers before extending the method across the whole catalog.

This article defines the retail operating rules. Use the established guides for channel-specific planning: Meta lead campaigns when an inquiry must precede a custom order, and Pinterest product discovery when people explore products before purchase. Their event and audience choices should inherit the genuine retail offer.

For the wider cost model, see advertising budget and contribution. Before connecting a stock signal to any write action, use the platform capability checklist and the provider confirmation guide.

Using SABSUS or another business system does not by itself establish a live product-feed connection, automatic inventory-based pauses, complete provider history or permission to spend. Verify the specific account, readable fields, supported action, authorization and returned status. Until that route is proven, this playbook can be operated as a reviewed manual process.