Product sets are where dynamic ads pretend to be smart. “Bestsellers.” “New in.” “Under £50.” Then a colourway dies on Tuesday and the set still behaves like last month’s truth.
Seasonal rotation is the most common cause of set sprawl. Handling it as data rather than as duplicated sets is covered in excluding off-season stock.
How rot forms
- Someone hand-picks 40 SKUs for a launch set and never revisits.
- Bestseller rules rank on 30-day sales while inventory is already gone.
- Filters use product-level stock while ads convert on sizes.
- A sale ends. Sets still point at clearance colours you are ashamed of.
ROAS can still look fine while the set quietly trains on worse and worse availability experiences. That is the same family of lie as ROAS looks fine, store hurts.
A 40-minute audit that pays for itself
- List every set tied to active ads.
- Sample 20 items per set. Check variant cover in Shopify.
- Count how many are zero or under your paid threshold.
- Fix rules (inventory filters, exclude tags) or rebuild manual lists.
- Note owners. Unowned sets rot fastest.
If more than a handful of sampled items are unshippable, the set is not a growth asset. It is a memory.
Rules that age better
- Inventory greater than buffer (see buffer stock)
- Exclude tags like
no-paid/final-sale-onlywhen those policies exist - Separate “can prospect” from “can only remarket”
- Prefer variant IDs when size/colour is the unit of truth
Pair with variant OOS behaviour and catalogue lag. Sets cannot save a broken feed. They can make a healthy feed usable.
Name the set after the rule, not the hope. “In-stock bestsellers with buffer” ages better than “Heroes forever”.
Where operators lose the week
They rebuild creative while the set still feeds dead olive. Or they pause a whole campaign because one SKU died, when pruning the set would have been enough. Precision is cheaper than panic.
Ralph’s job on a desk like this is to put cover next to what is actually in flight, then stage the prune or pause for approval. You still need sets designed like inventory products, not like mood boards.
Worked example
Set name: “Bestsellers 30d”. Rule: top 50 products by purchase count. No inventory filter. On audit day, 11 of 50 sampled variants are under buffer, including the colour that drove last month’s UGC. Dynamic ads still shop the set. Static ads still show the colour. Refunds tick up. The team blames creative fatigue.
Fix: add inventory greater than buffer, switch ranking to variant IDs where possible, move true clearance into a labelled set with honest creative. Same budget. Less lying.
Reporting trap
Set-level ROAS can look healthy while a few dead members burn trust. Sample the members, do not only chart the set. If your tools cannot show stock on members, export and join in a sheet once a week until they can.
Questions people actually ask
Are automatic bestseller sets safe?
Only if inventory filters are real and IDs are variant-correct. Bestseller without stock truth is a nostalgia playlist.
How often should I audit sets?
Weekly at minimum when spend is live. Daily during sales.
Product ID or variant ID?
If you sell sizes and colours, variants. Parent IDs hide death.
Should I delete dead items or rely on availability?
Both. Availability is the floor. Rotten manual membership still confuses humans and reports.
If a sold-out size is the problem: Ralph reads every size against every campaign overnight, finds the ones still paying for an empty shelf, and stages the exclusion with a revert for restock day. You say yes. See Ralph · How Ralph handles out-of-stock sizes · Docs