field notes · research · published October 12, 2026
What 10,172 GTM posts taught me about proof
I looked at three years of LinkedIn content from 35 GTM creators. One pattern stood out: most posts made a point, but very few showed specific evidence behind it.
“Add more proof” is familiar advice for anyone writing about go-to-market. I wanted to know what that looked like across a large body of real posts, not just the examples people remember.
So I collected posts published between September 2023 and September 2026, cleaned the data, and classified each post by format, content type, proof, and call to action. The final corpus had 10,172 posts. I used OCR to read text inside images and carousels, then checked the classification against a blind holdout set, where it reached 90% accuracy.
Proof was the exception
About 69% of posts had no proof mechanism. Only 1.6% named a client and included numbers. That gap matters: readers see a steady stream of advice, but much less evidence that helps them judge whether a claim holds up.
In this dataset, posts with a named client and numbers averaged 175 reactions and comments. Posts with no proof averaged 98. That is an association in this sample, not proof that adding a number by itself causes more engagement. Still, the difference is large enough to make specificity a useful thing to test.
Make the evidence easy to see
Proof does not need to mean a grand case study. It can be a clear before-and-after, a sample size, a screenshot of the work, or a concrete constraint. The point is to give the reader something they can inspect, instead of asking them to accept a broad claim on confidence alone.
The analysis also surfaced a format pattern: text posts averaged 112 engagements, compared with 49 for carousels. That is another descriptive result, not a universal format rule. Topic, audience, timing, and creator all matter. The useful move is to treat patterns as hypotheses, then check them against your own audience.
What I take into my own work
When I write about a GTM workflow, I try to show the actual inputs, the decision it made easier, and what changed. A number helps when it answers a real question. A number without context is just decoration.
This is the part of content research I find most useful: turn a large pile of examples into a small set of ideas worth trying, then measure what happens in your own context.