'Q2 Product Launch' campaign closed. 6 weeks, 3 channels, 2 audience segments. ✅ Technical workshop email: open 42% vs baseline 28%, conversion 6% vs 3.5%.
AGENT · CAMPAIGN OUTCOME REVIEWER
Campaign results arrive with the anomalies already highlighted.
Campaign Outcome Reviewer reads performance data at the end of a campaign — or every week for always-on campaigns — compares against the configured baseline, identifies creatives that perform above average and those to abandon. The summary reaches the marketing team's work channel with anomalies and proposed actions.
Campaign Outcome Reviewer at work.
⚠️ Pre-recorded demo video: -60% vs baseline. Cost per lead 3.2x on the mid-market CIO segment. 💡 Workshop+case study combination: replicate in Q3.
Good. Confirming all three actions. I'll prepare the Q3 brief on this basis.
Recorded. Event traced in audit.
Why it exists.
Marketing campaigns produce a lot of data: impressions, clicks, conversions, cost per click, cost per conversion, lead-to-sale rate. Marketing automation tools collect the numbers and show dashboards. The operational problem is not the absence of data — it is filtering: who looks at the dashboards every day, who identifies the anomalies, who decides what to abandon.
How it adds the structured signal.
Campaign Outcome Reviewer does not replace the dashboard. It adds a structured signal layer: the push summary that reaches the team's work channel at the declared cadence, with anomalies already ordered by weight and proposed actions. The marketing team receives the filtered signal, not the raw noise.
The decision stays with the head of marketing.
The decision to act stays with the head of marketing. The agent brings the picture; the human decides on strategic context. A brand may have strategic reasons to continue a campaign that is underperforming against the statistical baseline.
Three marketing team roles with different problems.
Head of marketing
Gets the push summary at the declared cadence (end of campaign or weekly). Reinvestment or abandonment decisions arrive grounded in structured patterns, not in subjective impressions built from dashboards opened in rotation.
Cross-channel · 7d 1,420 mentions
Site chat NPS 7.2
Reviews top theme: shipping
Social · IG sentiment +0.4
Brief to the brand owner
Media planner
Sees aggregated cross-channel budget allocation. Channels that deserve reallocation emerge from the comparison with the baseline, with granularity at the level of creative and channel together.
Brief 07:00 12 active segments
Gold segment +8% engagement
Silver segment redemption -3%
Anomaly · campaign 14 discarded
Monday action map
Content manager
Sees which creatives work and which do not, with supporting data. The learning curve on what makes content effective accelerates, because the feedback is structured and accumulates campaign after campaign.
RB RevOps Brief
Idritech SpA silent 8d · proposal expiring
Nexar Group proposal sent · no reply
Aludra Srl signature expected this week
End of campaign. The summary arrives in an hour, with decisions already ready.
The agent is scheduled on the end-of-campaign trigger.
For the head of marketing at a regulated B2B SaaS company, a six-week email nurture campaign on two audience segments closes. Campaign Outcome Reviewer is scheduled on the 'end of campaign' trigger from the marketing automation system.
Three blocks: working, anomalies, opportunities.
The agent reads data from the systems: HubSpot for email marketing, advertising platforms for associated paid campaigns (integration in delivery), CRM for conversions. The summary reaches the marketing Slack channel within an hour. Three blocks: creatives that work (workshop email 42% vs baseline 28%); negative anomalies (pre-recorded video -60%, cost per lead 3.2x); opportunities (banking compliance officer segment above average).
The lead validates in the Slack thread, the event stays in the registry.
The head of marketing validates the three proposed actions in the Slack thread. The full event stays in the runtime audit registry, queryable with a standard SQL client for campaign-after-campaign effectiveness analysis.
Configuration and technical resources.
The Campaign Outcome Reviewer rules are declarative. The customer's marketing team defines the comparison baseline (previous campaign, quarterly average, declared target, or a weighted combination), the anomaly thresholds (what is significantly above or below baseline), the source systems to read, the summary format and activation cadence. The rules live in the customer's repository, versioned.
Integration with advertising platforms (Meta Ads, Google Ads, LinkedIn Ads) is delivered during the project by the Exelab team via dedicated adapters. HubSpot marketing is natively available. For proprietary CDP or marketing automation systems, integration is delivered during the project.
- Language
- TypeScript (Node.js)
- LLM model
- customer's choice: Anthropic, OpenAI, Mistral, open source models hosted internally, AWS Bedrock for a private model
- Built-in controls used
- pii-detector, tool-rate-limit
- Native delivery channels
- Slack, Telegram, WhatsApp, OpenAI-compatible HTTP
- HubSpot marketing integration
- native
- Advertising platform integration
- dedicated adapter delivered during the project (Meta Ads, Google Ads, LinkedIn Ads)
- CDP and custom marketing automation integration
- dedicated adapter delivered during the project
- Memory
- persistent per instance, pgvector + PostgreSQL FTS on the historical campaign corpus
- Registry
- immutable, queryable with a standard SQL client
Frequently asked questions about the agent.
The baseline is declarative, configured by the customer's marketing team. Typical patterns: previous campaign on the same segment, quarterly average, target declared during planning. The baseline rule can combine multiple references (quarterly average weighted 70%, declared target weighted 30%). The rules live in the customer's repository.
Advertising platforms (Meta Ads, Google Ads, LinkedIn Ads, TikTok Ads) require dedicated adapters delivered during the project by the Exelab team. HubSpot marketing is natively integrated. For custom marketing automation platforms or the customer's CDP systems, integration is delivered during the project.
No. The agent identifies anomalies and proposes actions (reinforce, review, abandon). The final decision stays with the head of marketing. A brand may have strategic reasons to continue a campaign that is underperforming against the statistical baseline; the agent brings the signal, not the strategic judgement.
On two dimensions: time reclaimed by the marketing team on manual campaign data summarisation, and speed of reaction on negative anomalies (days ahead of the manual pattern). The pre-agent baseline is established during discovery on the real case.
From a 30-minute conversation to the squad in production.
A 30-45 minute conversation to understand how Campaign Outcome Reviewer would configure to the customer's case. Which source systems, which comparison baseline, which summary cadence.