fivos-campaign-analysis

Fivos Campaign Analysis

Closes the loop. Pulls analytics from Instantly, Smartlead, HeyReach (and others as needed), compares against benchmarks, flags anomalies, produces weekly and monthly rollups.

Hard rules

• Never auto-pause a campaign. Flag for engineer review - pausing is a human decision.
• Always report by tier. Aggregate tier-blind numbers hide the story (Tier A carrying the average, Tier C dragging).
• Inbox health is a first-class metric. A campaign with stellar reply rate on a dying domain is a crisis, not a win.
• Positive-reply rate is the north-star metric, not reply rate. Reply rate includes OOO, negatives, unsubs.

Inputs

• Client / campaign name or campaign ID(s) (required).
• Reporting window (default: last 7 days for weekly, last 30 for monthly, full-campaign for retros).
• Tier CSV (optional) - if provided, per-tier breakdown. Otherwise aggregate.

Outputs

1. Weekly report (the standard deliverable):
   • Volume: sent / delivered / bounce rate.
   • Engagement: reply rate, positive-reply rate, meeting book rate.
   • Inbox health: warmup status, bounce spikes, spam complaints.
   • Tier breakdown.
   • Anomalies + recommendations.
2. Monthly rollup: same metrics, windowed at 30 days, with trend comparison.
3. Retrospective (on request): campaign-end teardown - what worked, what didn't, copy breakdowns.

Metrics + benchmarks

Per tier (cold outbound, first 30 days):

Metric	Tier A target	Tier B target	Tier C target	Anomaly threshold
Reply rate	>=8%	>=4%	>=2%	<50% of target
Positive-reply rate (of all replies)	>=30%	>=25%	>=20%	<15%
Meeting-book rate (of positive replies)	>=50%	>=40%	>=30%	<20%
Bounce rate	<3%	<3%	<3%	>5% (pause; investigate)
Spam complaint rate	0%	0%	0%	>0.1% (immediate pause)

Domain-health warnings:

• Any inbox with >5% bounce in a 24-hour window = pause and investigate.
• Any inbox with a spam complaint = pause for 48 hours, investigate before resuming.
• Warmup score dropping >=15 points week-over-week = regress to lower daily send volume.

Workflow

Step 1 - Resolve campaign IDs

If user gave client name, list active campaigns:

mcp__instantly__list_campaigns({})
# filter by name contains <client>

Confirm the set with the user before pulling data.

Step 2 - Pull analytics in parallel

Per sequencer:

Instantly:

mcp__instantly__get_campaign_analytics({ campaign_id: <id> })
mcp__instantly__get_daily_campaign_analytics({ campaign_id: <id> })
mcp__instantly__get_warmup_analytics({ campaign_id: <id> })

Smartlead, Lemlist, HeyReach: equivalent per provider playbook. Run all in parallel.

Step 3 - Merge with internal data

Join analytics to the internal DB campaign log (from fivos-campaign-launcher) to attach tier, segment, client metadata.

Step 4 - Compute + compare

Per campaign:

• Compute each metric.
• Compare to benchmarks.
• Flag anomalies.

Step 5 - Produce the weekly report

# Campaign Weekly Report - <Client>
Window: <start> to <end>

## Headline
- Meetings booked this week: M (target: T)
- Positive replies: P (target: T)
- New anomalies: A

## By campaign

### <campaign_name> (Tier A)
- Sent: N / Delivered: M / Bounce: X% (bench <3%) [green/red]
- Reply rate: X% (bench >=8%) [green/red]
- Positive-reply rate: X% (bench >=30%) [green/red]
- Meeting book rate: X% (bench >=50%) [green/red]
- Inbox health: all warmups stable [green] / 2 inboxes regressing [red]

### <campaign_name> (Tier B)
...

## Anomalies
- <Campaign X>: bounce rate 6.2% - investigate SPF/DKIM on sender-domain-Y.
- <Campaign Z>: positive-reply rate 12% (below 20% floor) - messaging likely off-ICP for this segment.

## Recommendations
- <Campaign X>: pause sender-domain-Y, move leads to backup cluster.
- <Campaign Y>: tighten Tier C filter - too many DQ-adjacent rows leaking through.
- <Campaign Z>: rewrite emails 2 and 3 in the Tier B sequence - engagement drops sharply after email 1.

Step 6 - Publish

Output as Markdown. If the user has requested a Google Doc output (via CLAUDE.md convention), render it as a Doc in the Personal - Fivos / Reports / <Client> folder using the gws CLI.

Step 7 - Archive

Write a copy of the report to the internal DB (campaign_reports table) tagged by client + window, so trend analysis works across weeks.

Monthly rollup differences

Same structure, but add:

• Week-over-week trend lines per metric.
• Best and worst performing subjects / first-email opens (top 5 / bottom 5).
• Inbox retirement recommendations (any inbox trending to unrecoverable bounce rate).

Retrospective differences

Campaign-end teardown only. Add:

• Meeting-to-opportunity-to-won funnel (if CRM data available).
• Copy variants that outperformed.
• Segments that over-performed - feed back to fivos-icp-segmentation for the next build.

References

• ~/.agents/skills/gtm-meta-skill/provider-playbooks/instantly.md
• ~/.agents/skills/gtm-meta-skill/provider-playbooks/smartlead.md
• ~/.agents/skills/gtm-meta-skill/provider-playbooks/heyreach.md
• Instantly MCP tools: mcp__instantly__get_campaign_analytics, get_daily_campaign_analytics, get_warmup_analytics, list_campaigns, pause_campaign.