How We Track the GTM-Engineer Market
GTM Engineer Search's market data comes from job postings collected daily, each read by an AI classifier that answers the same fixed set of written questions about it, with uncertain answers sent to a person for review. This page documents the sources, how classification works, the limits of the data, and which figures are held back for now.
Where the postings come from
New postings are pulled every day from two live sources, JobHive and LinkedIn, into a jobs database kept separate from the rest of our systems. JobHive supplies the broadest sweep, and the LinkedIn pulls run targeted searches by phrase and region. A posting found by both is merged into one record before it is classified. Missed collection days are not backfilled.
Older postings in the database also came from an initial bulk import in July 2026 and from Blitz, a third source we pulled from until August 2026. According to GTM Engineer Search's tracking, 9,039 postings have been read by the classifier so far, across all of these sources.
How a posting gets classified
Each posting's title and full job description go to an AI classifier that answers 20 fixed, written questions about it. The first asks whether the role is GTM engineering at all. Nine ask whether the role covers each of the nine capability domains in our taxonomy. The rest ask about technical requirements, the level of AI skill the employer wants, seniority, people management, and whether the role is a founding hire.
The questions are frozen: every posting is asked exactly the same ones, word for word, and changing a question means deploying a new version of the whole set. For each answer the classifier also reports how confident it is, as a probability. A yes-or-no answer counts as yes when that probability is at least 50%.
The nine domains themselves come out of an earlier human review loop: 7 batches of postings judged by hand and 35 rulings on edge cases, such as a title that says "GTM" while the role does another job. That taxonomy was locked on 2026-07-25. The classifier in use since September 2026 asks about those same nine domains, with each question written fresh for it.
What happens when the classifier is unsure
Some answers go to a person before they count. That happens when the classifier says a role needs review, when it gives two options the same probability, or when it is less than 50% confident in the answer it picked.
The consequence depends on which question was uncertain. If it is whether the role is GTM engineering, the role stays off this site until a person rules on it. If it is anything else, such as one of the domains, only that one answer is held back, and the role counts as not having that domain until someone rules on it.
A person's ruling is stored next to the classifier's answer and never overwrites it, so we can still see where the two disagreed. Reviewers can also overturn answers the classifier was confident about, and in the first review of this classifier they did so seven times, for roles it had excluded that were in fact GTM engineering.
Scope and reliability
The corpus is English-only. Postings written in another language are filtered out before classification, because the taxonomy was built and reviewed on English job descriptions.
Totals across the whole tracked population are steadier than the labels on any one posting, since an occasional wrong answer on a single posting moves a total only slightly. That is the level this site publishes at. Every figure here is a population statistic ("N% of tracked roles…") and never a claim about what one particular posting requires.
Why some numbers are held back right now
We switched to the current classifier in September 2026. How often it tags each capability domain has not yet been checked against people reviewing the same postings, and its domain figures differ noticeably from the previous classifier's. Until that check is done, domain percentages, the average number of domains per role, and which domains tend to appear together stay off this site.
Two figures are published now. The first is postings analyzed, which counts every posting the classifier has read, whatever its answer. The second is the number of open GTM-engineering roles at direct employers that our tracker lists. That count depends on the classifier's answer to the first question, so it is labeled for what it is: an operational count of what the tracker lists today, not a precision claim.
Salary benchmarks are held for a separate reason. Pay figures are being prepared as their own dataset, under stricter rules that tie every salary to the exact wording of its posting, and they will be published from that dataset.
Frequently asked questions
Where does GTM Engineer Search's data come from?
Postings are pulled daily from two live sources, JobHive and LinkedIn, into a jobs database kept separate from the rest of our systems. Older postings also came from a bulk import in July 2026 and from a third source used until August 2026. Each posting is merged across sources and then classified before anything reaches this site.
How is a posting classified as a GTM-engineer role?
An AI classifier reads the title and full job description and answers 20 fixed questions, starting with whether the role is GTM engineering and which of the 9 capability domains it covers. Answers it is unsure about go to a person, and a role whose GTM engineering status is uncertain stays off the site until someone rules on it.
How reliable are the numbers?
Totals across the whole tracked population are steadier than the labels on any single posting, which is why this site only publishes population statistics. Domain figures from the current classifier are held until they have been checked against human review of the same postings.
Why don't you publish salary benchmarks yet?
Salary figures are being prepared as their own dataset, under stricter rules that tie every salary to the exact wording of its posting. They will be published from that dataset once it is ready.
Hiring a GTM engineer?
Are you a GTM engineer?
Get on the radar