The first part of this guide focused on the mental shift, setup, planning discipline, and the Tools/Skills/MCP framework. Part 2 delves into the nuts and bolts of getting going with your first Cannonball GTM project.
5. The 8 Tool Categories You’ll Ever Need
Your first few Claude Code sessions probably looked like this: pull some data, write an output, done. That is the right way to start. But real GTM work does not stay that simple for long. Once you move beyond one-off tasks, you are no longer running a single job. You are running a chain of dependent steps, where the output of one module becomes the input to the next, and a bad decision early can contaminate everything downstream.
If you are building your first real pipeline, start with the minimum viable version: data source discovery, web capture, and email verification. That is enough to produce useful output and learn the operating model. Add the other layers as your workflow gets more complex and your quality bar rises.
The Operating Model: Economic Discipline
The operating principle for this entire framework is built on a “Ladder of Escalation.” You do not start with the most expensive model or the most robust paid API. You start with the most cost-effective local method and only escalate when that layer fails to meet your quality or coverage bar.
Free Local Methods: Scrapers you run on your machine
Cheap Paid Methods: Marketplace actors or batch APIs
Expensive APIs: Premium model calls and high-tier enrichment
Here are the eight tool categories you need:
#1. Data Source Discovery for TAM Building
The Question: Where does the data actually live for this specific vertical and TAM, and what is the cheapest, most reliable method to get it at scale?
If you choose the wrong source, everything downstream gets worse: lower coverage, worse match quality, more cleanup, and higher cost. Before building the pipeline, run a dedicated discovery pass for that specific client, industry, and geography. This is a formal research module, not ad hoc Googling.
Authoritative sources first: Start with source-of-truth datasets wherever possible, like government databases, state licensing boards, Secretary of State filings, Google Maps, and industry-specific directories. In healthcare, for example, MIMI Data Labs is a strong “hidden gem” because it is structured, inexpensive, and often better than generic paid enrichment.
The source-stack mindset: In many cases, the right answer is not one source but a stack: a primary source for core coverage, a secondary source for gap-filling, and a niche registry or directory to improve match quality.
What to test during discovery:
Does a structured public dataset already exist?
Can the data be scraped locally for free?
Is a custom scraper worth building for this source?
How to rank sources: Rank them by coverage, quality, cost, and durability. Coverage tells you how much of the TAM you can actually reach. Quality tells you how well the data maps to your ICP. Cost tells you what it will take to run at scale. Durability tells you how stable the source or scraper is likely to be over time.
Output: A ranked source plan with cost estimates, coverage expectations, and a clear recommendation for which sources to use as Primary, Fallback, and Gap-fillers.
#2. Web Scraping/Bulk Capture
Question answered: How do I capture this data at scale using the cheapest workable method before escalating to paid infrastructure?
Once you know where the data lives, you need an extraction strategy. Before committing to a full production run, run a capture test on a small sample. Prove access, estimate coverage, measure failure rates, and understand the cost profile.
