"we ran ads to a sold-out product all weekend. google doesn't care that the shelf is empty."
Monday morning, and the first thing you see is Saturday's ad spend on an out-of-stock product that sold out Friday night. The ad performed beautifully right up to the moment it started wasting 100% of its budget, and nothing in Ads Manager changed colour. A Shopify Community merchant put the routine that follows in one line: "sick of manually pausing campaigns every morning (and still missing things on weekends)".
This is for the founder or operator on Meta and Google who has just found that leak, or suspects it is running and wants a routine that catches it. It explains why no ad platform flags it, gives you the weekly check that does, and sizes what the leak costs when you skip the check. It does not claim a robot does this for you. Today, it is a check you run, or ask Drew to run.
Key Takeaways
- → Spend lives in the ad account; stock lives in Shopify. No ROAS column connects them, so a sold-out product keeps earning ad spend until a human joins the two.
- → Shopping and PMax stop serving an item once the feed says unavailable. Search ads, Meta catalog items with sold-out variants and products set to keep selling when out of stock do not stop.
- → The weekly check is a four-step join: active spend by product, Shopify inventory by variant, the intersection, then the near-sold-out hand-off. Twenty minutes on a Monday.
- → The cost is spend rate times days undetected. Found in the monthly review means 30 days of leak.
Why No Platform Catches Ad Spend on Out-of-Stock Products
The platform optimises the objective it was given and reads the feed it was sent. Meta's objective is purchases; Google's is conversions at a target. Neither objective includes "and the shelf is not empty", because neither platform holds your inventory as a decision input. The leak lives in the join between two systems that never talk: spend is in the ad account, stock is in Shopify, and the only table that connects them is the one you build.
The mechanics differ by format, which is why "Google pauses out-of-stock products" is half true and expensive.
| Format | What happens when the product sells out | Keeps spending? |
|---|---|---|
| Google Shopping / Performance Max (feed-driven) | Stops serving the item once the feed reports it unavailable (Google's automatic item updates help here). Feed-sync lag applies, so "once" can be hours or a day | Briefly, then no |
| Google Search ads pointing at the product page | Nothing. The keyword, the ad and the landing page are all still live; the page just says sold out | Yes |
| Meta catalog / Advantage+ items with a sold-out variant | The product often stays eligible while one size or colour remains, and shoppers land on the size they wanted, which is gone | Yes |
| Single-product or hero ads (any platform) | Nothing. The ad is not tied to a feed row, so no availability signal reaches it | Yes |
| Any product set to "continue selling when out of stock" in Shopify | Shopify reports it as available (the inventory setting is per variant), so every feed says available, and every format keeps going | Yes, everywhere |
The worst case is a winner. The ad is at its most efficient right before the stock runs out, so the algorithm is feeding it budget at exactly the moment every click starts bouncing off a sold-out page.
Check What the Business Can Sell Before Judging What the Ads Deliver
Here is the habit that separates operators who find this leak from the ones who find a mystery instead. When a product's ROAS falls off a cliff, the ad account is the last place to look. Check what the business can actually sell: is the variant in stock, is the size run intact, did the price change, did a discount code break in the cart. Only after that do you judge the ad.
Skip the order and you will "fix" a media-buying problem that was a warehouse problem, resetting learning on a campaign that was doing its job, then be baffled when it does not recover after the restock. Business reality first. The ad account can wait twenty minutes.
The Weekly Check: Active Spend × Current Inventory, Step by Step
Four steps, and the first is the one most guides skip.
- → 1. List active spend by product, not by campaign. For catalog and dynamic formats, the platforms' product reports give spend per item: Meta's product-ID breakdown and Google's product report, both an export you run. For single-product and hero ads, map ad to product by landing page or by the product in the creative. This step is the reason the leak survives: spend is reported by campaign and ad, and a campaign is not a shelf.
- → 2. Pull Shopify inventory by variant. Not by product. A sold-out size run counts: if the medium is gone and the ad's audience buys mediums, that product is out of stock for the purpose of the ad. Flag every variant at zero units, and every product set to continue selling when out of stock, because those will never show as unavailable anywhere.
- → 3. Intersect the two lists. Any product with spend in the last seven days and a zero or a broken size run is the leak. Sort by spend. The top three are usually 80% of it.
- → 4. Hand off the near-sold-out cases. Products with days of cover left and rising spend are a different problem with a different rule (the forward-looking version of this check is its own piece in this series). Here, just note them.
Two honesty notes before you size anything. Product-level ad spend is an approximate split: the platforms' product reports give spend per item only for catalog and Shopping formats, single-product ads map to a product only by landing page or creative subject, and Datadrew holds ad spend at campaign, ad and account level and allocates it to products by approximation. So treat the leak figure as a range, not an invoice. And on cadence: this is a weekly routine on daily-synced data. Datadrew syncs spend and stock once a day, so a leak that opens on Saturday shows up on Monday morning. We say so plainly, because the alternative is pretending to an intraday check nobody actually runs.
What It Costs When You Skip It
The arithmetic is spend rate times days undetected, and the second term is the one you control.
A hypothetical to make it concrete: a hero product carrying $120 a day of catalog and Search spend sells out on a Friday evening. Caught Monday, the leak is roughly two and a half days, about $300. Caught at the following Monday's review, it is ten days, $1,200. Caught in the monthly review, it is thirty days, $3,600, and the Quality Score and the conversion rate of the campaign it lived in have both taken damage that outlasts the restock. Found in the monthly review means 30 days of leak. That is the whole case for a calendar entry.
The leak isn't the scandal; the detection lag is.
The Adjacent Leaks the Same Join Finds
Once you have spend next to stock, three more things fall out of the same table.
- → Low-stock winners about to strand their budget. The product with nine days of cover and a rising spend line. The rule for those, and the thresholds, live in the inventory-aware piece in this series.
- → Back-in-stock products with ads still paused. The inverse leak: you paused the winner when it sold out, the restock landed, and nobody un-paused it. Demand you paid to build, unserved.
- → The margin sibling. Some products lose money at a good ROAS once shipping and returns are in. Run that check against the margins you've shared with Drew; Datadrew doesn't ingest COGS or compute contribution margin, and finding below-break-even SKUs is an analysis you run or ask for, not a sweep Drew runs by itself. The method is in contribution margin before scaling.
How to Make It Stick
Calendar it. Monday, 9am, twenty minutes, before you touch a budget. The four steps above, a sorted list, a pause or a product-set exclusion for anything at zero, a note for anything close.
Or hold both sides of the join in one place. Today this is a weekly check you run, or ask Drew to run against your connected Meta and Google spend and Shopify stock. The automatic out-of-stock × spend check is on our build list; Drew doesn't run it on its own yet. Pausing the ad is a change Drew doesn't make on your account today; approved changes with guardrails are part of execution, which is rolling out. If you would rather run the join yourself in Claude or ChatGPT, Datadrew's MCP connector gives an AI assistant read-only access to your spend, your Shopify catalog and stock on paid plans, and the weekly check is a single prompt.
Single-purpose apps exist for the pause itself (AdStockGuard, for example, auto-pauses on stockout and resumes on restock). They are proof the pain is real. What they cannot tell you is which of the products still in stock are the next leak, or whether the one you just paused was worth the spend in the first place. That needs stock, margin and spend in the same view, which is where product-level performance starts.
How Drew fits. Ask Drew to run the out-of-stock spend check. It holds your Meta and Google spend, your Shopify catalog and stock, and the margins you've shared in one place, and answers in one pass, conversationally. Today it surfaces the leak when asked; it doesn't pause anything. (Product-level spend is an approximate split.) The same view is what the product intelligence side of Datadrew is built on, with stock coming straight from your Shopify connection.
To run it on your own catalog, pricing is published and flat and the free plan doesn't need a card.
Key Takeaway
Ad spend on out-of-stock products leaks because spend is reported by campaign in the ad account and stock is held by variant in Shopify, and no platform joins the two before it spends. Shopping and PMax stop once the feed catches up; Search ads, Meta catalog items with a sold-out variant, single-product ads and anything set to keep selling when out of stock do not. Run the four-step join weekly: spend by product, inventory by variant, the intersection, the near-sold-out hand-off. Size the leak as spend rate times days undetected, and remember that product-level spend is a range. Found Monday is a $300 problem. Found in the monthly review is a $3,600 one.
Frequently Asked Questions
How do I catch ads pointing at an out-of-stock product before they burn a week of budget?
Run the spend-times-stock join on a fixed day each week: list last week's ad spend by product (platform product reports for catalog formats, landing page for single-product ads), pull Shopify inventory by variant, and intersect. Anything with spend and a zero, or a broken size run, gets paused or excluded from the product set that morning. On daily-synced data a Friday-night stockout surfaces Monday, so the weekly cadence caps the leak at a few days rather than a month.
Do Google Shopping or Meta pause ads automatically when a product sells out?
Partly. Google Shopping and Performance Max stop serving an item once your feed reports it unavailable, with sync lag. Google Search ads pointing at the product page keep running. Meta catalog items often stay eligible while any variant is in stock, and single-product ads on either platform have no availability signal at all. Products set to "continue selling when out of stock" in Shopify report as available everywhere, so nothing pauses them.
Should I pause the ad or exclude the product when it sells out?
For catalog and Shopping formats, exclude the product from the product set or listing group so the campaign keeps learning on the rest. For single-product or Search ads, pause the ad or ad group rather than deleting it, so the history survives and you can resume on restock. Either way, put the restock date in the same calendar entry as the pause, because the back-in-stock product with ads still paused is the next leak.
How do I find wasted ad spend on sold-out products in Shopify?
Shopify holds the stock and the ad platforms hold the spend, so you need both exports side by side: Meta's product-ID breakdown or Google's product report for spend per item, and a Shopify inventory export by variant. Join on product or SKU, filter to zero units or a missing size, and sort by spend. Datadrew holds both sides for connected stores and runs the check conversationally when you ask; it doesn't run it automatically yet.
How much does an out-of-stock ad leak cost?
Daily spend on the product multiplied by the days before anyone notices, plus the slower damage to conversion rate and Quality Score in the campaign it lived in. A $120-a-day product caught on Monday after a Friday stockout costs about $300; caught at the monthly review it costs about $3,600. The detection lag, not the stockout, sets the bill.
Written by Sumit Bansal, co-founder of Datadrew. Published 3 September 2026. The merchant lines are quoted as written from our operator research. Datadrew's product-level spend figures are an approximate split; the automatic out-of-stock check is on our build list and is not live.