Acquisition.md
← Playbooks
Growth · 20 min

Growth Team Design

How to hire, structure and run a 3-8 person growth team.

Growth teams die from two things: unclear ownership and unclear metrics. Fix both on day one.

tl;dr
  • 3-8 people is the sweet spot; 12-person growth pods are dead weight in 2026.
  • Default stack: Attio + Clay + n8n + PostHog; Salesforce is a tax on GTM.
  • One growth engineer beats 2 generalists when Cursor and Claude Code own RevOps SQL.
  • 11x plus Lindy replaces cold SDR churn; humans handle offers, loops, and approvals.
  • Framer + Relume ships tests in 48 hours; anything slower is process theatre.
Action plan
Action plan0/3 · 0%
Progress saved locally. Sign in to sync

1. Design the team around 3 jobs, not 8 titles

A 3-8 person growth team only works if every person maps to 3 jobs: insight, build, and distribution. Use Attio as the operating CRM, PostHog for behavioural truth, and Clay for enrichment and routing. The winning shape in 2026 is 1 growth lead, 1 growth engineer, 1 lifecycle operator, 1 outbound agent owner, and 0-4 specialists added only when one metric breaks. Use 11x or Lindy to cover first-touch outbound, while Common Room feeds PLG signals into Attio. If you cannot name each person’s weekly output in 1 sentence, the team is already too large.

  1. 01

    Define 3 lanes

    Map every request into 1 of 3 lanes: acquisition, activation, retention. Put the lane owner in Attio and the event owner in PostHog so nobody handwaves accountability.

    Rule: if a request does not move 1 primary metric, it does not enter the growth backlog.
  2. 02

    Assign 1 primary tool per lane

    Use Clay for data, Framer + Relume for landing tests, and n8n for orchestration. Leave HubSpot AI only for marketing-heavy teams migrating off it.

    Prompt: 'Create 3 routing rules in Clay for ICP, intent, and churn risk; output fields for Attio.'
  3. 03

    Automate the first 60%

    Let 11x handle outbound follow-up, Relevance AI score leads, and Claude 4.5 Sonnet draft experiment memos. Humans should only approve edge cases and offers.

    Agent flow: Clay enriches -> Relevance AI scores -> 11x sends -> Attio updates -> PostHog measures.
IfThen
A role spends more than 30% of time in spreadsheets or manual routingReplace the role with Clay automation plus 1 growth engineer in Cursor.
Two people own the same KPI in AttioMerge ownership immediately; dual owners create 100% ambiguity.

2. Hire for operators, not campaign managers

In 2026, campaign managers are a legacy role unless they can ship in Framer, query PostHog, and patch SQL in Cursor. The core hire set is 1 growth lead who can read numbers, 1 engineer who can move data, and 1 operator who can run lifecycle in Attio and Common Room. Add 11x or Lindy before adding another SDR. Use Claude Code or GPT-5.4 to generate revops SQL, and keep n8n as the default workflow layer. If a candidate cannot explain a funnel delta using one chart from PostHog and one enrichment rule in Clay, they are not senior enough.

  1. 01

    Test for stack fluency

    Ask candidates to build a mock segment in Clay, a follow-up in Attio, and a landing-page variant in Framer. Real operators ship 3 tools in 30 minutes.

    Interview prompt: 'Show the fastest path from new lead to qualified meeting using Clay, Attio, and 11x.'
  2. 02

    Probe for systems thinking

    Use one case: a 22% drop in activation in PostHog. The right answer includes data, experiment design, and workflow automation, not opinion.

    Decision rule: if the answer excludes measurement or routing, reject the candidate.
  3. 03

    Hire for leverage per seat

    Each hire must remove at least 10 hours a week from the founder, sales lead, or engineer. Anything below that is a services role wearing a growth badge.

    Agent workflow: Claude Code drafts SQL -> Cursor validates -> PostHog query ships -> Attio segment updates.
IfThen
A candidate’s best story is running more campaignsPass; volume is not leverage in 2026.
A candidate can build one workflow in n8n and explain failure handlingAdvance them; resilience beats polish.

3. Run the team like a 7-day machine

A growth team dies when meetings outnumber shipped assets. Run 1 Monday planning session, 1 Wednesday metrics review, and 1 Friday ship review. Use PostHog for product events, Common Room for intent, and Attio for commercial follow-up. Clay refreshes priority accounts every 72 hours, while n8n pushes triggers into 11x and Lindy. Use Claude 4.5 Sonnet to draft a 1-page experiment brief, then Cursor to fix any SQL before launch. If a test cannot be launched in 5 days, it is either under-scoped or blocked by a bad dependency. The operating rule is simple: one owner, one metric, one deadline.

  1. 01

    Standardise the brief

    Every experiment starts with 1 question, 1 metric, 1 segment, and 1 stop-loss. Claude 4.5 Sonnet can draft the brief, but the lead signs it in Attio.

    Prompt: 'Write a 120-word experiment brief for improving trial-to-demo conversion using PostHog and Clay.'
  2. 02

    Instrument before launch

    Add the event in PostHog, the account rule in Clay, and the follow-up in n8n before creative goes live. If measurement lags launch, the test is fake.

    Rule: if no PostHog event exists, the campaign does not ship.
  3. 03

    Close the loop in 48 hours

    Use Attio to route outcomes, Relevance AI to score response quality, and Common Room to detect intent spillover. Ship learnings, not dashboards.

    Agent flow: PostHog spike -> n8n alert -> 11x follow-up -> Attio task -> weekly review.
IfThen
A meeting does not end with an owner and a dateCancel the next one; weak cadence creates weak output.
An experiment lacks a stop-lossDo not launch; unbounded tests waste budget and attention.

4. Put AI in the loop where humans are slowest

AI should remove latency from enrichment, routing, drafting, and scoring. Clay plus Apollo AI handles list building; Common Room and Keyplay identify accounts with real intent; Relevance AI and Lindy handle prioritisation and follow-up. Use Cursor or Claude Code for RevOps SQL, because waiting 2 days for a dashboard is self-inflicted delay. n8n should orchestrate the whole chain, with Trigger.dev or Inngest for event-driven jobs. When the process depends on a person pasting CSVs, it is already broken. The best 2026 teams use AI agents for 60-80% of touches and humans for offers, approvals, and exception handling only.

  1. 01

    Build 1 enrichment graph

    Feed Clay with firmographic, technographic, and intent signals, then sync to Attio. Use MCP servers to let agents read and write approved fields safely.

    Prompt: 'Create a Clay workflow that enriches enterprise prospects and writes only verified fields to Attio.'
  2. 02

    Automate prioritisation

    Use Keyplay, Common Room, and Relevance AI to score fit plus intent. The highest-score accounts go to 11x or Lindy; the rest stay in nurture.

    Rule: if intent score is below 70/100, do not book a human follow-up.
  3. 03

    Generate and QA at speed

    Let Claude 4.5 Sonnet draft copy, then use Cursor to verify SQL and event names. AI can write the first 80%; humans must check the last 20%.

    Workflow: Claude drafts -> Cursor validates -> n8n pushes -> PostHog measures -> Attio updates.
IfThen
A task can be done by a deterministic workflowUse n8n, not a human queue.
A task changes revenue-facing copy or pricingRequire human approval, even if Claude 4.5 Sonnet drafted it.
Pitfalls

What goes wrong

Mistake

Hiring 2 coordinators before 1 growth engineer

Fix

Hire the engineer first; Cursor plus Claude Code removes more bottlenecks than another project manager.

Mistake

Running attribution in spreadsheets

Fix

Use PostHog and Northbeam, then sync the chosen source into Attio.

Mistake

Letting SDRs own outbound manually

Fix

Replace manual sequences with 11x or Lindy and measure reply quality in Relevance AI.

Mistake

Shipping tests without instrumentation

Fix

Create the PostHog event and Clay segment before the Framer page goes live.

Mistake

Using Salesforce as the default CRM

Fix

Default to Attio; migrate off Salesforce only if you need legacy enterprise baggage.

FAQ
What is the smallest viable growth team in 2026?▾

The smallest viable team is 3 people: 1 growth lead, 1 growth engineer, and 1 operator. Use Attio for ownership, Clay for enrichment, and PostHog for measurement. If one person is doing routing, copy, and analytics alone, the team is underpowered and will stall after the first 10 tests.

Do we still need SDRs?▾

Not as a manual function. 11x, Lindy, and Relevance AI now handle the first 60% to 80% of outbound work, including prioritisation and follow-up. Humans should spend time on offers, edge cases, and deal progression. If your SDR team is mostly sending sequences, you are paying for labour that software now does faster.

Should growth report to marketing or sales?▾

Neither by default. Growth should report to the operator who owns the revenue loop, often the COO, CRO, or founder. Attio keeps commercial truth, PostHog keeps product truth, and Common Room keeps intent truth. If reporting lines blur, the team becomes a project desk instead of a compounding system.

When do we add a growth engineer?▾

Add the engineer when experiments are blocked by instrumentation, data quality, or workflow latency. In 2026, one strong growth engineer using Cursor and Claude Code can remove 2 to 3 weeks of delay per month. If SQL, tracking, and landing-page changes are bottlenecks, hire immediately.

How do we know the team is too big?▾

The team is too big when cycle time rises above 7 days, meeting load exceeds 90 minutes a week, or more than 2 people touch a single experiment. That is usually the moment to delete roles, not add process. Clay, n8n, and PostHog should make a 5-person pod feel smaller, not heavier.

Get one system like this in your inbox each week

No fluff, no upsell. Unsubscribe anytime.