Target Anomaly Detector
What it reads, what it flags, what comes back, what you approve.
- Keyword and product-target performance, daily, eight weeks back
- Search terms behind any flagged target
- The account change log, and target ACoS or CPA by campaign
- Stock, Buy Box and price on every advertised ASIN
- CPC, CVR or ACoS outside the target's four-week same-weekday range
- Competition split from product: CPC up with CVR steady, or CVR down with CPC steady
- CVR drops checked against stock, Buy Box and price before the target is blamed
- Bleeders: 10+ clicks, no orders, and spend past twice target CPA
- Every flagged target, marked CPC-driven or CVR-driven
- A bid cut or pause for each bleeder, staged for today
- The skill that handles each of the rest
- Bid cuts capped at 30% per target per day
- No more than 20 pauses in a run
- Branded and hero terms capped, never paused
- The previous bid and status logged, so every change reverses
Where a target broke overnight, and what drove it
Targeting report + change log + stock · components
The skill's components as built in Qore, with Split CPC from CVR open.
Frequently Asked Questions
This skill runs every morning on yesterday's data. Any target with 10 or more clicks, no orders, and spend past twice its target CPA gets a bid cut or pause staged the same day, so one bad night does not become a bad week.
By which number moved. CPC up 20% or more with CVR steady reads as competition. CVR down 30% or more with CPC steady reads as a product or relevance problem, and the skill checks stock, Buy Box and price on the ASIN before it blames the target.
No. Every bid cut and pause is staged and batched into one approval the same day. Nothing moves until you approve it, and each change is capped at 30% per target per day.
Will it pause my branded keywords?
A target has to spend $25 or more yesterday, or twice its average daily spend, to be scanned. Targets under seven days old have no stable baseline yet. Everything else waits for the weekly review.
They are routed. CPC-driven flags go to bid adjustments on clear outliers, repeat flags go to the chronic target review, and product faults go to the product anomaly scan.
Every one of these runs on your own account.