Writing
Field notes from inside the pipeline.
Short, specific, written for operators. No growth-hacking, no listicles - just the engineering that decides whether your GTM stack produces.
-
What I learned reading 50 companies' job postings
A job posting is the most honest document a company publishes - a dated, signed statement of what's broken and budgeted. Learn to read it like an engineer.
3 min read- #signals
- #sourcing
- #intel
-
Cold email isn't dead, bad infrastructure is
Cold email isn't dead; bad infrastructure is. The channel still works when it's built like a production system - filtered data, separate domains, a human gate.
2 min read- #deliverability
- #cold-email
- #infrastructure
-
The sending domain is production infrastructure
A sending domain is production infrastructure, not a marketing channel. Treat it like a server: monitored, hardened, separated, and never assumed fine.
3 min read- #deliverability
- #infrastructure
-
DMARC p=none: the silent killer nobody checks
A DMARC policy set to p=none quietly damages every cold email you send. It's the thirty-second DNS check nobody runs, and it's more common than you think.
3 min read- #deliverability
- #dmarc
-
AI SDRs don't have a quality problem, they have a data problem
AI outreach reads like slop not because the model is bad, but because the data layer upstream of it is. Filter before you enrich, verify before you send.
3 min read- #data-quality
- #ai
- #enrichment
-
Signals don't produce pipeline without routing
A signal that fires into a Slack channel is a notification, not pipeline. Routing - prioritization, ownership, SLAs, suppression - is the part nobody builds.
3 min read- #signals
- #routing
- #pipeline
-
Tool sprawl is the new technical debt
Every GTM tool solves one problem and creates three integration problems. The real cost isn't the subscriptions - it's the reconciliation work between tools.
2 min read- #architecture
- #integrations
- #technical-debt
-
Enrichment credits are a tax on bad filtering
Enrichment credits aren't a data cost, they're a tax on bad filtering. Filter before you enrich and your cost per qualified contact drops by a multiple.
3 min read- #enrichment
- #filtering
- #cost
-
Your CRM is a data graveyard (and it's killing your outbound)
The CRM was supposed to be the source of truth. It became the place data goes to die. Data quality is an engineering problem with engineering answers.
3 min read- #data-quality
- #crm
- #pipeline
-
Your forecast is fiction (until you instrument the funnel)
A forecast is only as good as the funnel instrumentation behind it. Define stages as a schema, measure transitions, and the fiction becomes a number.
2 min read- #funnel
- #instrumentation
- #forecasting
-
Treat your ICP like a predicate, not a vibe
An ICP isn't a paragraph - it's a predicate the pipeline evaluates across firmographics, technographics, triggers, and exclusions. Write the filter.
2 min read- #icp
- #targeting
- #filtering
-
The one person who knew the Clay tables left
When the one person who knew the Clay tables leaves, the motion halts. The fix isn't a hire - it's a documented pipeline, not a pile of per-campaign tables.
3 min read- #architecture
- #documentation
- #key-person-risk
-
Catch-all domains are where contact data goes to die
A catch-all domain accepts every address, so a verified-looking list still poisons your sender reputation. The correct action is to throw the address away.
3 min read- #deliverability
- #verification
- #data-quality
-
Your sequencer has quietly become your CRM
The sequencer quietly becomes the system of record while the CRM rots into a reporting database. The fix is a data-model decision, not a sync.
2 min read- #crm
- #data-model
- #integrations
-
Let AI do the labor, keep the gates human
AI is brilliant at outbound's labor and dangerous at its gates. Automate the pipeline - sourcing, enriching, drafting - and keep the send human.
3 min read- #ai
- #outbound
- #automation
-
Dedup belongs at the write path, not after
Dedup is a write-path problem. Match on stable keys at entry and suppress before the credit gets spent; post-hoc dedup is a refund you'll never get.
2 min read- #data-quality
- #dedup
- #crm
-
The human gate is a feature, not a fallback
The human gate isn't a fallback - it's the feature. Automate the pipeline, never the send: the checks that decide whether it works can't be modeled.
2 min read- #ai
- #outbound
- #deliverability
-
Run outbound like an experiment (the freeze rule)
Most outbound programs change constantly and learn nothing. Hold variables constant, change one at a time, tag exceptions, and guessing becomes a method.
2 min read- #outbound
- #experimentation
- #method
-
Attribution is a measurement problem, not a credit argument
The attribution fight isn't about credit, it's about which broken number to believe. Fix the tags, define the model, and the argument collapses into a number.
2 min read- #attribution
- #instrumentation
- #data-quality
Subscribe via RSS.