An Open-Source CLI for
Vector-Based Semantic
TAM Segmentation
GTM still hasn't solved TAM segmentation.
Sales lives or dies on the relevance a message has to its receiver. This is why marketers use segments. At scale, that relevance is decided before a single word gets transmitted, through what bucket each account gets assigned to. What TAMage's thesis questions: is the status quo drawing those different segments on the same axis you differentiate your messaging on? Mostly, no.
If you look at how teams treat outbound nowadays, you'll find three patterns:
1) Naive teams treat the whole TAM as one bucket. One message for everyone, relevance for no one.
2) AI-eager teams are the opposite extreme. Generating unique copy per lead, effectively one new "segment" per row. The models can't yet write strong sales copy, regardless of how much prompt engineering or account context you input.
3) Disciplined teams segment w/ deterministic filters: size, funding, industry. Feels rigorous, until it's checked against the axis: is having different copy for 30-person companies vs. 300-person companies the most effective use of segmentation? Maybe if you are an HR SaaS, but for everyone else, it's suboptimal. These firmographic traits are convenient lines to draw, but they are not the lines your messaging differentiates on. They lack nuance.
The nuance lives in the unstructured data you can already find. What a company does, sells, and struggles with is written in its positioning, its product language, its website. Embedding models translate all this data into vectors, every account gets real coordinates in semantic space. Clustering algorithms find the natural groups in this space.
The result: segments defined by what makes these companies tick — the same axis your messaging differentiates on — with an explanation on how to sell to each one. Every account shows its segment, its fit to your ICP, and the model's confidence in the assignment.
TAMage is that translation layer. Feed it your enriched TAM and it hands back the map. Giving you access to this workflow without the ML learning curve, as an open-source CLI.