Highlights

Highlights is the feed at the top of your Overview — “what matters now”. After each daily sync, TheAdsHub scans every active campaign, detects the most important things happening, and shows you a curated top-5: what happened, why it matters, and a concrete, quantified action to take. Where it helps, it also brings in real-world context from the web (a policy change, a conflict, a demand surge).

It’s advisory: Highlights tell you where to look and what to do — nothing is changed on the platforms.

How it works

Highlights is a two-stage engine.

1. Signal detectors (deterministic)

First, a set of detectors run over your recent data (Google + Meta) and emit candidate signals, each scored by business impact (money at stake × magnitude). The detectors:

Signal	What it catches
Mover	Campaigns whose conversions changed the most vs the previous period.
Wasted spend	Campaigns spending real money with zero conversions.
Geo-mismatch	Spend leaking into countries the campaign never targeted (targeted geo vs actual spend).
Delivery drop	Campaigns whose impressions collapsed vs their baseline.
Anomaly	Statistically unusual days in spend, conversions or ROAS (see Alerts).

Most detectors look at a short recent window (~7 days) for timeliness; geo-mismatch uses a wider ~30-day window, because location leakage is a structural configuration issue that accumulates over time.

2. AI curation (with world context)

The candidates are ranked and de-duplicated into a diverse shortlist — deliberately spanning signal types, so you don’t just get five variations of the same thing. That shortlist is passed to TheAdsHub’s AI, which:

1. Picks and writes the top 5 most business-relevant highlights.
2. For the highest-impact geo-mismatch and delivery-drop signals, uses web search to check for current real-world events in that country or segment (regulation, conflict, disaster, seasonality, market disruption) that could explain the signal or create an opportunity — and folds that into the recommendation, with source links.
3. Attaches a quantified action to every highlight (e.g. “add ID, UZ, JP as negative locations to stop the ~€560/30d leak”, or “raise budget +40% to capture the demand window before CPMs rise”).

Web search runs only for the top few signals, so cost stays bounded (roughly one AI call per workspace per day).

What’s in a highlight

Each of the five cards shows:

• Type — Geo mismatch, Delivery drop, Wasted spend, Anomaly, Mover, Budget.
• Title — the headline, at a glance.
• Narrative — what happened and why, including any real-world context.
• Recommended action — a specific, quantified next step.
• Impact — an estimate of the € or % at stake.
• Source & country — the Source (Google/Meta) and, where relevant, the country.
• Links — a link to the affected campaign, and the web sources the AI consulted (when it used external context).

Example

▸ Geo mismatch · APAC Search P1 — spent €563 (30d) in countries not in this campaign’s targeting: Indonesia (€305), Uzbekistan (€98), Japan (€81), Philippines (€55)… Indonesia’s new e-commerce regulation (Permendag 19/2026, effective June 2026) may be driving unexpected delivery there, and this spend also carries compliance risk. → Action: add ID, UZ, JP, PH as negative locations to stop the ~€563/30d leak; if Indonesia is strategic, brief a separate compliant ID campaign. (~€563 saved / 30 days)

When it runs & things to keep in mind

• Highlights are regenerated once per day, after the sync completes — so the feed reflects your latest data.
• The feed is portfolio-wide and not affected by the date or country filters in the header (those drive the metrics below it).
• Geo-mismatch needs a campaign’s targeting to be known; it appears once targeting has been ingested by a sync.
• Everything is advisory — apply the recommended changes in Google Ads / Meta Ads Manager. External context is a helpful signal to investigate, not a guarantee.
• The AI model that powers Highlights is configurable per workspace (it needs a web-search-capable model).