How I use AI to qualify leads with confidence (and without looking like a jackass when reaching out) Those of you who know me personally, know that I don’t like making mistakes in general... Let alone in cold outreach! Some people can live with 80% fact-based outreach, but that just ain’t me! Yesterday 🦾Eric Nowoslawski released a list with 100k+ suspected SaaS companies. (link in comments). (If you are interested in improving your lead gen skills... Follow him, trust me!👌) The only problem was that the current ranking with SaaS keywords wasn’t perfect. He shared and reached out to his network to help improve the ranking. So... Here is how I solved his problem → Step by Step Step 1 - Grabbed 150 random domains from the list → This should ensure statistically significant outcomes. Step 2 - Scraped the websites with Firecrawl → I love this scraper as it returns the page in clean markdown. Step 3 - Used OpenAI to check if a website is a SaaS company → Used a prompt combining Eric’s keyword ranking with the scraped data → Returning if the company is SaaS, a likelihood score, and reasoning Step 4 - Evaluate the results: → How accurate are the results? → Is this the right model? → Can the prompt be improved? And that’s it → now you can do cold outreach without looking like a jackass... BAM💥 PS - What can I improve in my method? P.P.S - Have a great weekend!🙏 #ai #salesautomation #leadgeneration
Thanks for turning me on to Firecrawl. I'm very impressed. I'm a little unclear, however, in the difference between the Crawl data and their new LLM Extract. Assuming I'm not a developer and am just going to be feeding this into ChatGPT, is there a decided advantage of the LLM Extract? I'm thinking there is, but with the LLM costing 50x the scrape, seems worthwhile to ask. Your use case is similar to what I'm attempting, for what it's worth.
This is really cool! I'm using firecrawl a bunch myself - if you try to scale and use a cheaper model you should take a look at using Claude Haiku, has larger window and super cheap. It'll be wrong more often but I just layer the break up the classification into api calls to mitigate (does this company provide products or services return yes or no -> send all passing companies -> is this company a tech company yes or no -> send all passing companies -> is this a SaaS company yes or no and return reasoning)
Mark de Jonge check out the one i shared as well.
Wow love to see this!
that's really smart
There ya go! Awesome to see
Wow, this is impressive! Great work, Mark! 👏 It's awesome to see how AI can be used to enhance lead generation techniques.
Amazing, great stuff Mark!
Founder @ LinkedUp | LinkedIn Lead Generation | Startups & Scale-ups
1mo📌Ranking results: https://docs.google.com/spreadsheets/d/1Am2xIsEltkbkfcyop4bSWpG4fYTY9DOysfh5TCZl_gA/edit#gid=0