Design Your Own GTM System

Outcome: a one-page written system design for your own motion, and a named next course that builds it.

Surface: App and MCP server
Level: Beginner
Uses: Everything from lessons 01–09
Credits: 0
Prerequisite: Lessons 01–09, and the notes from each

Why write it down

A design that lives in your head gets rebuilt every time someone asks a question about it, and drifts every time you change one column. One page fixes that: it is the artifact you hand to the person who inherits the workflow, and the thing you check against when a number moves.

It is also short. If it runs past a page, the workflow is doing two jobs and should be two workflows.

The template

Copy this and fill it in. Every field maps to a lesson.


# System design — <workflow name>
 
## 1. Intent            (lesson 01)
Motion:
Who counts:
Why now:
What I need to know:
How they get reached:
 
## 2. Grain              (lesson 02)
One row is:
Second table needed?    yes / no — grain:
 
## 3. Stages             (lesson 03)
Find:
Research:
Enrich:
Outreach:
Boundary counts to log:
 
## 4. Contact strategy    (lesson 04)
Primary path:
Fallback path:
Catch-all policy:
Expected hit rate:
 
## 5. Criteria            (lesson 05)
Attributes (refresh cadence):
Signals (window, dedupe rule):
 
## 6. Gate                (lesson 06)
Fit model:
Timing model:
Threshold:
Derived from capacity of:      rows/week
Gate positions:
 
## 7. Hygiene             (lesson 07)
Match key:
Merge rules:
Suppression list:
Source of truth conflicts:
 
## 8. Economics           (lesson 08)
Rows per run:
Credits per run:
Cost per contactable lead:
Cost per meeting:
Biggest lever:
 
## 9. Build / buy / manual (lesson 09)
Built:
Bought:
Deliberately manual:
 
## 10. Cadence
Weekly:
Monthly:
Quarterly:
Never re-runs:
Owner:

A filled example

The developer-tool motion carried through the course, compressed:

Section	Answer
Intent	Warm outbound to engineering leaders at companies building platform teams
Grain	Table A: company. Table B: person, from qualified companies only
Stages	Find `find_companies` → Enrich company + stack + jobs → gate → Find people → email → verify → AI line → sequencer
Contact	Verified work email primary, LinkedIn fallback, catch-alls to secondary domain
Attributes	50–500 employees, eng org >20, competing CI tool — refresh quarterly
Signals	2+ infra roles in 30d; eng leader started under 90d — check weekly
Gate	`fit >= 6 AND timing >= 4`, threshold set at 80 accounts/week capacity
Hygiene	Match on root domain; verified beats unverified; suppress customers + 30d contact
Economics	5,380 credits/run, CPCL ~6, ~86 credits per meeting
Build/buy	Built: score, gate, prompt. Bought: all data. Manual: top-10 founder picks
Cadence	Signals weekly, company enrich monthly, funding/stack quarterly

Ten rows. That is the entire system, and anyone on the team can read it and know what breaks what.

Pressure-test it

Before you build, run the design against these five questions. Each maps to a failure this course has already described.

Does every expensive column sit below the gate?

If not, move it. This is the single cheapest fix available to you.

Does the “why now” have a date on it?

An attribute wearing a signal’s clothes gives you a list with no ordering and no expiry.

What happens on run two?

Walk the merge rules by hand for one row that already exists. If the answer is “it re-enriches everything,” fix the cadence before scheduling.

What is the honest denominator?

Count rows that end contactable and qualified — not rows enriched. If you cannot compute it from the design, you cannot manage the cost.

What breaks first at 10× volume?

Usually the gate threshold, because it was set against today’s capacity. Note what you would change rather than changing it now.

If a question has no answer yet, that is fine — write “unknown, resolve in first run” rather than guessing. A design with three honest unknowns is more useful than one with three invented numbers.

Do this now

Fill the template for one real motion. Aim for one page.
Run the five pressure-test questions and fix what they surface.
Estimate the first run’s cost from section 8 and check it against your balance.
Pick the course that builds your design:

If your design is mostly about	Build it with
Sourcing and market sizing	TAM Sourcing
Getting contactable data	Waterfall Enrichment
Timing and triggers	Signals & ABM
A full cold campaign	Automated Outbound
Facts no database holds	Research & Web Scraping
Fixing what you already own	CRM Enrichment
Reacting to inbound	Automated Inbound
Doing it all by prompting	AI for Sales Reps

If you have not built anything in the product yet, take SyncGTM 101 first — it is the same design executed step by step.

Check your work

The design fits on one page
Every section has an answer or an explicit “unknown”
The five pressure-test questions are answered
You have named the next course and the first thing you will build

Where this breaks

The design that never gets used is the one written for a motion nobody asked for. Pick something with a person waiting on the output — a rep who needs a list next week, a campaign with a date. A design with a customer gets built and corrected; one written as an exercise gets filed. If no motion has a waiting stakeholder, that is your finding, and it is more valuable than the document.

Further automation

Keep the page next to the workflow and update it when the workflow changes — particularly the gate threshold and the economics, which move most. Teams that keep the design current can answer “why is spend up?” in a minute; teams that do not spend a week reconstructing what the pipeline does.

Course complete

You now have the vocabulary the rest of the library assumes: grain, stages, waterfalls, attributes and signals, gates, cadence, CPCL. Every other course is one of these concepts built out in the product.

Next: SyncGTM 101 to build it, or jump straight to the outcome course your design pointed at.

Reference for this lesson: Core Concepts, Tables, Credits, Signals.