Building Niche Job Boards (Web3 & AI) | HR Tech Consultant | Expertise in Job Sites SEO, Google Jobs, NLP & AI Solutions, Job Scraping | HR Tech Blogger
Today, I want to talk about an important history lesson. In 2017, Google released one of the best job-matching products. Still, it failed miserably. The “Cloud Jobs API,” later renamed “Cloud Talent Solution,” was the best-matching product in 2017. I am confident it was still the best matching product until around 2022-2023 when new parsing and matching startups emerged. What made the product so good? Already back in 2017, they had an extensive layer of over 250K normalized occupations, a testament to the depth and breadth of their product. They had a skill ontology of 50K hard and soft skills and their relationships. Back then, most job parsing software providers had a list of skills with some alternative names or simple connections, but none of them had even 5% of what Google had. They used ML and embeddings to calculate similarities between occupations (something that job boards are “discovering” today thanks to OpenAI). This stack was cutting edge. It was beyond anything the job board and matching world had seen back then. Trust me; I evaluated every service provided on the market back then, and Google created something marvelous. Still, the product failed. Yes, access to the product was so limited that it was a pain to get in, and in the usual Google arrogance, this drove people away. The language support was limited, and the product never exceeded EN-US, EN UK and Spanish. The business model was expensive as hell. The lead of the talent solutions team at Google quit, which did not help either. Still, the product failed because integrating it took much work. You had to push all your jobs and company data to Google and constantly sync them. You had to maintain a clean relationship between companies and jobs (good luck with aggregators here). These requirements were challenging for your average job board. Pushing hundreds of thousands of jobs via API every second requires solid resources that most job boards did not have back then (and still don’t today). “Thanks for the industry lesson, but why do I care?” Sometimes, it is about something other than having the best technology; it is about understanding how an industry works and building your tools around the functionality everyone is used to. Did you figure out which $300MM gorilla I refer to in the paragraph above? #textkernel #bullhorn #ai #matching #parsing
