Campaign Building

---
name: email-infra-campaign-building
description: "Build cold-email campaigns and route by ESP and SEG. Use for campaign setup, the Lead-ESP by sending-vendor decision matrix (ESP matching is dead as a fixed rule), profiling a lead list by recipient ESP with the dns-auth-audit playbook in --esp-mix mode, isolating SEG leads onto dedicated domains, and the launch gate. Triggers on build campaign, ESP matching, provider matching, ESP mix, profile the lead list, what ESP are these leads on, SEG, Mimecast, Proofpoint, Barracuda, campaign routing, sending-vendor matrix. Do NOT use for writing copy or sequences (use gt-cold-email) or reading the dashboard (use the dashboard-reading sub-skill)."
---

# Campaign Building & ESP/SEG Routing · [GTM Engineer]

> **Reads:** `{SKILL_BASE}/resources/reference.md` §1, §2, §7, §8 · **Related:** dashboard-reading, bounce-audit · gt-list-building.

> 🔒 **Read-only area.** The campaign build and the routing are done from the **email infra management system**, not by hand in the sequencer. Use this sub-skill to decide *what* the campaign and routing should be, read the live config to check it, and report the gap. Do not edit a campaign or a routing rule in Instantly / EmailBison / Smartlead / Lemlist directly — that creates a second source of truth. The `--esp-mix` profiling below is read-only and safe to run.

How to build a campaign that routes to the right inboxes and gets optimized from data, not from a 2024 rule of thumb. Numbers and taxonomy live in `{SKILL_BASE}/resources/reference.md` §1, §2, §8.

You are not A/B-testing copy. You are reading a **matrix of already-segmented sends** and pushing volume toward what works. The winning combination is `lead list × sending vendor/ESP × recipient ESP × SEG`, and it's on the dashboard.

---
### Determining the recipient ESP mix on a lead list

**Fastest path — run the playbook, no Clay credits:**

```bash
cd {SKILL_BASE}/playbooks/dns-auth-audit/scripts
uv run execute.py --esp-mix --file lead_domains.txt --csv acme_esp_mix.csv
```

Queries MX directly and prints the distribution, the SEG share, and a `no-email`
count (domains with no MX at all — guaranteed hard bounces, strip them before sending).
It uses the **same provider list** as the Clay formula below; the two are kept in
lockstep in `MX_PROVIDERS` inside `execute.py`. Add a provider in one place, add it
in the other.

Use Clay instead when the domains already live in a Clay table and you want the ESP
as a column alongside the rest of the enrichment.

### Clay version — How To Run MX Analysis in Clay to Determine ESP Mix on Lead Lists

Run this on a target account list or an existing customer list — you want the real inbox mix, not a sample.
### Tools Needed:
- Clay workspace
- Company domains normalized (e.g. `growthtoday.co`)
- Target account list or existing customer list

Step 1: Add Enrichment Column in Clay
- Column name: **`HTTP API`**

Step 2: Set Up the HTTP API**

- Paste this endpoint:
    
    ```
    https://dns.google/resolve?name=domain&type=mx
    ```
    
- Replace `domain` with the normalized company domain field in Clay.
    - Example: https://dns.google/resolve?name=`{{company_domain}}`&type=mx

Step 3: Rename This Column

- Rename the `HTTP API` column to: **`get_mx`**

Step 4: Add a Formula Column (copy-paste)
- Column name: MX Provider

```
  javascript
// A null MX ("0 .") is the domain saying it accepts no mail at all (RFC 7505).
// Same as having no MX: a guaranteed hard bounce. Check it FIRST.
!{{get_mx}}?.Answer?.length || {{get_mx}}.Answer.every(data => /^\s*\d+\s+\.\s*$/.test(data?.data || "")) ? "no-email" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("google")) ? "google" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("outlook.com") || data?.data?.includes("office365")) ? "microsoft" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("pphosted.com") || data?.data?.includes("ppe-hosted") || data?.data?.includes("ppsmtp")) ? "proofpoint" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("sophos.com")) ? "sophos" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("mimecast")) ? "mimecast" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("barracuda")) ? "barracuda" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("fortimail") || data?.data?.includes("fortimailcloud.com")) ? "fortinet" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("emailsrvr.com")) ? "rackspace" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("trendmicro.com")) ? "trendmicro" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("securemx")) ? "securemx" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("mxthunder.net")) ? "mxthunder" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("mtaroutes.com")) ? "mtaroutes" :
{{get_mx}}?.Answer?.some(data => data?.data?.includes("zoho")) ? "zoho" :
"other"
```  

## Part 1, Don't hard-code ESP matching

Sending from the same provider the recipient uses (Google→Gmail, Outlook→Outlook) was a 2024 band-aid. **Do not hard-code it.** It can still *turn out* to be the right call for a given segment, but only the data decides, per segment, per week.

- Do **not** set "ESP matching" as a global campaign rule.
- **Do** read the Lead-ESP × sending-vendor matrix (Part 2) and route toward the cells with the best **human** reply rate.
- We are already heavily segmented, so this is a **diagnostic matrix we read**, not a set of experiments we run.

---

## Part 2, The decision matrix (read it, then route)

On the dashboard's vendor-performance panel (the dashboard-reading sub-skill), read **human reply rate** per cell. Rows = the recipient's ESP; columns = the sending vendor / inbox ESP.

| | Sending: Google | Sending: Outlook | Sending: SMTP/Custom |
|---|---|---|---|
| **Recipient: Google** | cell | cell | cell |
| **Recipient: Microsoft/Outlook** | cell | cell | cell |
| **Recipient: Enterprise / SEG** | → Part 3 (isolate) | → Part 3 | → Part 3 |
| **Recipient: Other (Zoho/custom)** | cell | cell | cell |

**How to read it**
- **Metric = human reply, never total reply.** Automated replies (OOO/auto-responders) inflate the number and mean nothing. Use the Automated-vs-Human toggle and read Human. (Bounce and positive-reply don't belong in this matrix, see `reference.md` §7 and the dashboard-reading sub-skill.)
- **Winning cell** (high human reply, bounce in range) → push more leads/volume there.
- **Losing cell** (near-zero human reply, or rising bounce) → pull leads out of that pairing; don't force it.
- **Don't assume the diagonal wins.** Google→Google or Outlook→Outlook may or may not be best; let the cell decide.
- **Data hygiene:** flag untagged inboxes and any large "Custom Mail Server / SMTP" share, an untagged column makes the matrix lie.

**Iterate weekly.** Week over week you get clear winners per `lead list + sending vendor + recipient ESP` combo. Shift weight to winners; on a pay-per-lead engagement, volume on winners is revenue.

---

## Part 3, SEG leads go in their own campaigns, on dedicated domains

Enterprise recipients behind a **Secure Email Gateway (Mimecast, Proofpoint, Barracuda)** are a separate track, not a cell in the normal matrix. A single spam complaint on a SEG can poison a whole domain, so SEG traffic must never share domains with normal leads.

**Rules**
1. **Isolate SEG leads** into their own campaign(s).
2. Run them on **dedicated, never-reused sending domains**, used for nothing else.
3. **Low concurrency into one org**: don't fire many inboxes at a single company at once (SEGs block a domain that pushes many messages into the org in a short window).
4. **No links, no tracking** (already our default), SEGs weight these heavily.
5. **Go multi-channel**: pair with LinkedIn/phone; email won't be your only path into SEG accounts.
6. **Recycle, don't waste.** When bounce climbs on a SEG campaign, swap the domain out, then re-test it on easy **Google/Outlook** segments before retiring. A SEG-burnt domain often still performs on regular leads. Use placement + warmup score as the swap trigger (the dashboard-reading sub-skill).

Expect lower reply rates on SEG-heavy segments. That's the recipient's policy, not a broken setup. Prioritize reachable segments; shift weight off aggressively-gated ones rather than burning domains forcing them.

---

## Part 4, Building the campaign (mechanics)

Build so the campaign is **visible to and managed by the inbox-management system** and so **nothing sends from an unhealthy inbox**.

1. **Create the campaign through the management dashboard, not directly in the sequencer.** A campaign built directly in the sequencer is invisible to routing, dashboard, and health management. (If you must draft in the sequencer, attach **no inboxes** until the system equips it, so it can't send unmanaged.)
2. **Set the routing rule:** `Google` | `Microsoft` | `Both` (based on Part 2, not on ESP-matching dogma).
3. **Scope the inbox pool** with tag filters: *include-by-tag* to restrict to a chosen pool, *exclude-by-tag* to keep away from inboxes used elsewhere; set the **region** tag where a client sends by region.
4. **Let the automation attach/detach.** It attaches only eligible inboxes and maintains membership:
 - **Active / New Inbox** → eligible, but read §2 before you attach. The system stops excluding an
   inbox at `new_inbox_age_days`; **Growth Today does not attach one until `warmup_floor_days`**,
   which is longer. "The dashboard let me" is not the standard — §2 has an explicit warning box
   about exactly this gap.
 - **Warmup Needed** → throttled to cold 0–1 but kept attached.
 - **Burnt** → excluded.
5. **Naming convention:** `Segment – ESP`, e.g. `Webvisits – Google`, `Webvisits – Microsoft`; low volume (< 500 leads) → `Webvisits – All` (Both); by rep → `Webvisits – Andrew`.

> **Failover caveat (EmailBison only):** a lead being prospected by an inbox that turns Warmup Needed keeps getting sent from that throttled inbox, Bison won't move the lead to a healthy inbox on the campaign. Instantly and Smartlead can reroute the lead to a healthy inbox; EmailBison can't. Watch for leads stranded on throttled inboxes (the dashboard-reading sub-skill).

---

## Part 5, Launch gate (hard, before any send)


- [ ] **List 100% verified** (and re-verified if > 30 days old), see `gt-list-building`.
- [ ] **First email is plain text**: no HTML, no images (incl. signature), no links.
- [ ] **Spintax / variance present** on subject + body.
- [ ] **Blacklist pre-check** on any domain < 60 days old (Spamhaus DBL / URIBL), see the bounce-audit sub-skill.
- [ ] **SEG leads isolated** onto dedicated domains (Part 3), not mixed into normal campaigns.
- [ ] **Routing rule set from the matrix** (Part 2), not from ESP-matching.

---

## ✅ CAMPAIGN CHECKLIST (copy-paste)

```
ROUTING
[ ] Segment defined: lead list × recipient ESP × (SEG? yes/no)
[ ] SEG leads split into their own campaign on dedicated, never-reused domains
[ ] Routing rule (Google/Microsoft/Both) chosen from the dashboard matrix, NOT ESP-matching
[ ] Include/exclude-by-tag set; region tag set if client sends by region

BUILD
[ ] Campaign created via the management dashboard (or drafted with NO inboxes attached)
[ ] Only Active inboxes past §2 `warmup_floor_days` attached; Burnt excluded; Warmup Needed throttled+attached
[ ] Naming convention applied (Segment – ESP)

LAUNCH GATE
[ ] List 100% verified (re-verified if >30 days)
[ ] First email plain text: no HTML / images / links
[ ] Spintax / variance present
[ ] Blacklist pre-check on domains <60 days
[ ] Low concurrency into any single SEG org

OPTIMIZE (weekly)
[ ] Read human (not automated) reply per matrix cell
[ ] Push volume to winning cells; pull leads out of losing cells
[ ] Swap SEG domains on rising bounce; re-test on easy Google/Outlook before retiring
[ ] Flag untagged inboxes / large Custom-SMTP share
```

---

*Created by [Growth Today](https://www.growthtoday.co), the AI-native GTM engineering firm. Maintained by [Brigitta Ruha](https://www.linkedin.com/in/brigittaruha/). More open Claude Skills for go-to-market teams: https://www.growthtoday.co/claude-skills*