You know the spiral: six tabs of breakdowns, a spreadsheet you built at 11pm, and three competing theories — fatigue, CPMs, that budget change Tuesday. The client wants an explanation tomorrow. Sentrum decomposes the spike into its actual drivers, gates them by statistical significance, and hands you the verdict with the lever attached.
The same decomposition a senior buyer runs by hand — six tabs collapsed into one significance-gated read.
Spend first — did budget or pacing change? Then result volume, then efficiency, then the supporting metrics (CPM, CTR, CVR), then mix shifts across ad sets, creatives, and placements, then the edits made in the account. No skipping ahead to a favorite theory.
Every candidate driver is tested against real thresholds — about ±10% on spend, KPIs, and efficiency, 20% swings for mix shifts, stricter floors before anything gets called a winner or loser. The CTR wobble that’s actually noise gets labeled noise.
The output is a sentence you can forward: what changed, the one or two drivers behind it, what it’s costing, and the specific Ads Manager move tied to the diagnosed cause — with the numbers that prove it underneath.
A spike has six possible drivers. The diagnosis isn’t finding a story that fits — it’s eliminating the five that don’t.
Sentrum checks CTR decay against the significance gate and your creatives’ own fatigue curves — so “refresh everything” only gets recommended when wear-out is actually the driver.
CPM moves get decomposed: account-wide drift versus self-inflicted pressure from overlapping audiences or placement-mix changes. The difference decides whether you wait it out or restructure.
Reallocation across ad sets, creatives, and placements is the spike cause buyers miss most — blended CPA rises while every segment looks stable. Sentrum reads the allocation, not just the averages.
Budget changes, audience edits, new creatives, bid adjustments — lined up against the timeline so cause can precede effect. The drop that started Thursday gets checked against what happened Tuesday.
The framework behind this page is public. Read it, use it — and when you’re tired of running it manually, the product is here.
The full manual framework this page automates — every driver, every check, in the order a senior buyer runs them.
READ THE GUIDEThe broader companion: how to read any performance drop — not just CPA — without guessing.
READ THE GUIDEOne or two of six drivers, almost always: a spend change (pacing or scaling), CPM pressure (auction competition, targeting, or placement mix), a CTR drop (creative resonance or fatigue), a CVR drop (site or funnel quality), a mix shift (spend reallocating across ad sets, creatives, or placements), or an optimization edit someone made in the account. The diagnosis is figuring out which one or two actually moved — and ruling out the rest.
It decomposes the change in a fixed diagnostic order — spend, then result volume, then efficiency, then the supporting metrics (CPM, CTR, CVR), then mix shifts, then the edits made in the account — and gates every candidate driver by significance thresholds. What survives is the verdict: what changed, why, whether it matters, and the specific Ads Manager lever tied to the diagnosed cause.
Then Sentrum says so. Changes below the significance gates (roughly ±10% on spend, KPIs, and efficiency, with stricter rules for mix shifts and rate metrics) don’t get called as findings. “This is within normal variance — don’t change anything yet” is an honest verdict, and sometimes the most valuable one.
No. The connection is read-only by architecture with a standing never-touch-budgets rule. Sentrum reads, diagnoses, and recommends; every lever stays in your hands.
About 60 seconds to connect via Meta OAuth, and the first diagnosis typically lands within minutes — Sentrum analyzes your historical data immediately, so the spike you’re staring at tonight gets decomposed tonight.