portfolio / 03 · research
3 years of gtm content
I scraped and classified 10,172 LinkedIn posts from 35 well-known GTM creators, September 2023 to September 2026, to see which kinds of posts earn engagement and which do not.
tool stack
- Apify
- Python
- LLM classification
- OCR
explore the corpus
Group the posts, pick a metric, filter by format and year. Click a row to see its trend.
over time
All posts in the current filter, by quarter.
what the data says
Computed live from the same dataset.
how it was built
- scrape10,210 posts pulled from 35 creator profiles with an Apify LinkedIn actor.
- cleanDuplicates, group pages and empty posts removed: 10,172 posts kept.
- read the imagesText inside images and carousels extracted with OCR, so the classifier sees what readers see.
- classifyAn LLM labels each post: content type, format, proof and call to action. A blind holdout check measured 90% accuracy.
Engagement is reactions plus comments. It measures attention, not revenue. Creator names, captions and links are left out on purpose: only anonymous counts ship to this page.