Detection
Signal, not noise.
One noisy hour shouldn’t reach anyone. Additor compares each campaign with its own history, waits for the next hourly check to agree, then scores what’s left by what it could cost.
-
01SUN SUN SUN SUNExpected$813
Its own baseline
Each campaign is measured against its own recent pattern for that day of the week, over a 28-day window.
-
0213:3518:3500:35
Checked every hour
Fresh metrics from every connected ad account are compared with that baseline each hour, using a robust z-score.
-
0323:35Outside the normal range00:35Still outsideFinding opened
Confirmed, not guessed
A finding opens only when consecutive hourly checks agree. A single noisy hour never reaches your team.
-
04Score62WARNING
Scored by what’s at stake
How far the metric moved, how much history backs the comparison and how much spend is at stake.
One score, three severities
Some detectors have a floor or a cap on their score.
- How far it movedThe gap between the observed number and the expected one.
- How much history backs itThe depth of the baseline behind the comparison.
- How much spend is at stakeThe budget the move actually touches.
Anatomy of a finding
Every alert explains itself.
Open a finding and the whole case is on one screen, what moved, against what, what probably caused it, what it has cost so far and what to do next.
- 1
Observed against expected
The real number next to what this campaign’s own history predicts, and the gap between them.
- 2
Honest about fresh data
If the platform is still re-reporting the day, the finding says so before you act on it.
- 3
The days behind this
The normal range, the expected line and the day being judged, so you see the move instead of reading about it.
- 4
Why it probably happened
A plain-language explanation, with the possible contributing factors beside it.
- 5
What to do, and what it cost
A recommended next step and the estimated extra spend so far.
- 6
How it was measured
Baseline window, robust z-score, first and last seen, ad account and platform ID. Nothing hidden.
On hold
When the data is stale, it says so.
Ad platforms sometimes fall behind. When an account’s latest metrics are too old to trust, Additor puts judgement on hold instead of guessing, and shows you the last numbers it trusted.
Judgement waits for good data
No campaign is judged on metrics that may be wrong.
Last trusted, clearly labeled
The numbers you see on hold are the last ones Additor trusted, and they say so.
Sync status for every ad account
See which accounts are current and which are behind, at a glance.
Alert routing
The right alert, in the right channel.
Connect Slack once for the whole organization. Each workspace then decides which alerts leave it, and which channel they land in.
Routes per workspace
Critical and Warning to the team channel, Insights somewhere quieter, each workspace sets its own.
The fix travels with it
Each message carries the severity, the campaign, the move and the suggested next step.
A delivery log you can check
Every alert is logged as sent, failed, held or queued.
Follow-through
From alert to action in one click.
Every finding links straight to the place you fix it, and stays on record after you have.
-
WARNINGSpend spikeCampaignsDaily budget$1,400Spent today$412ROAS2.31×
Open campaign
Jump from the finding straight to the campaign it’s about, in Campaign management.
-
Why did spend jump on FB_CLM_US_EN_ROAS?Used 3 toolsThe 17 Oct budget raise explains most of it. Want me to draft a change for your approval?
Ask Ad Intelligence
It reads your anomalies too. Ask what changed. Anything that affects spend waits for your approval.
-
History14 findingsSpend spikeDismissed by Marco Ricci · 11:02
CPC surgeDismissed by Kenji Tanaka · 16 Oct
Dismiss, keep the record
Dismiss what you’ve handled. History keeps every finding, and the audit log keeps who did what.