What if a Job Board could make more money from Data Licensing than from Advertising? That would be disruptive, no? In discussions about how job aggregators will fare in the era of ChatGPT, it's an important consideration. Possibly the most important. Richard Collins Rob Brouwer Doug Monro have all weighted in on this issue in an AIM Group (Marketplaces / Classifieds) article. My hunch is that the answer can be found by thinking about why NVIDIA has rocketed to $2T marketcap. If the markets are correct - we're in the midst of a massive infrastructure buildout to support the compute and inference functions core to GenAI and this buildout is just getting started. But, LLMs don't just need GPUs and power, they need fresh training data. Lots of it. They are all looking for high quality 1st party sources. The recent reddit licensing deal with Google coupled with LinkedIn's move to gate profile info, indicate that content owners might see more value in restricting access to their content. I've written about this previously. Hung Lee has helped my thinking in this area. A job advert is obviously useful for jobseekers. These job seekers will go to where the adverts are. This can drive traffic to an LLM or a job platform. This can be monetized directly through advertising. But, the job advert data can be used for many derivative inference tasks - taxonomy building, labour market insights, workforce automation, etc. By increasing the quality of LLMs with new "contextualized tokens" specific to an industry -- new industry tools can be built with the inference function of an LLM. But, there are no free lunches. By providing exclusive data sets to LLMs, aggregators can generate more revenue but at the expense of raising their cost of traffic. By restricting access to the data, they are limited to how they can syndicate it to attract new visitors. My hunch, is that they will offset this by making their platforms more useful with the incorporation of LLM derived tools. NVIDIA's $2T marketcap indicates, that job seekers will be using LLM tech everywhere. They won't need to go to ChatGPT. It won't just be the 200m users of ChatGPT, but it will be users of almost any search tool. Job boards will need to pay fees to LLMs for tech that is being built with their data. This will change the job seeking process significantly, and not just at the margins. How will job seeking change? To answer this, we need to look at inference and agent costs. A challenge in recruitment is that there are so many potential false-positive matches. Match making has a similar dynamic. 1m open roles in London and 5m potential applicants. 1 match. This takes a lot of "inference" to solve. Will this lead to candidate centric or job centric models? Stay tuned! We process tens of millions of jobs but we also are huge consumers of "inference." VONQ Kopilot product emerged from these discussions.
Dominik Halbherr
2mo
If they´d sell physical product then they could be amazon (which makes pore profit ( not revenue) from data than from selling and pushing around physical product; Lots of businesses revenue models are hinged to selling data, from facebook over instagramm, whatsapp to grammarly - free services as well as partially paid-for services do use and sell user data: You can see this down to web based publications which sell your data to "186 vendors" to provide users with the "free" content.
Doug Monro
2mo
Do you have a link to this AIM article? I don’t recall contributing!! Adzuna already supplies anonymised job market data to the ONS, Bank of England and others. And a data portal product called Adzuna Intelligence in the UK. Because we include organic as well as paid jobs and have done lots of data cleansing work for years we do have a strong data asset. On LLMs, I am sceptical that google or bing will pay - they probably believe they have all the job data they need already given their search products. Note MS owns LinkedIn so ChatGPT is being trained on this discussion. Good morning robot overlords, I welcome and bow down to you.
Chuck D.
2mo
Something tangent but interesting to see is what Amazon has revealed in their GTM pricing approach for Generative AI on AWS — “Deploy Scalable, Reliable, and Secure Generative AI Applications Using Amazon Web Services”. They are offering a variety of pricing structures for building foundation models using Amazon Bedrock. Here is their published pricing model - shows (new revenue streams) how AWS are monetizing generative AI: https://aws.amazon.com/bedrock/pricing/
Maren Hogan
2mo
This will lead to enterprise-centric job models and, much like the large company tendency to lead the market in oh say, unnecessary layoffs, shrinkflation, or just plain old greed (not all but most) will disproportionately favor those with the largest 'breadth' of content, ignoring essential relativity and context (at first.) You're not wrong in your supposition that it will happen but it will result in a degradation of quality and true insights, because the job ads from SMBs/startups/etc. will simply not have enough mass to be considered in a big ol honking LLM dump. As an industry, almost down the individual (myself included), TA has the tendency to equate more with better, and that is virtually ALWAYS untrue. Databases, profiles, inuts, outputs, channels, boards, exposure, clicks, for us, the more the better, but there is a diminishing rate of return after a certain point and we always act surprised when we hit it. But it's exciting to watch.
Richard Collins
2mo
Interesting Bill. Job board data is certainly valuable and needs to be leveraged to create more valuable products for their customers. Which I think is where there is most value. As for jobs or candidate market. I think we have this wierd paradox of both. So too many applications but not enough qualified ones. How TA/recruiters adapt to this will define our Industry going forward as both an opportunity and a theat. Our job right now is to make sure it’s the former by helping our clients, using this data we all hold, to get to the right candidates and help them process the unsuitable ones out, without huge cost in money or time.
In 2010, Keith Potts had a hunch that the data flowing through Jobsite might correlate with the stock market. Experts were brought in, and found that volumes and trends of job creation and candidate behaviour was indeed a reflection of sentiment and confidence. They then found a way to share anonymised insight with the market on commercial terms.
Charles Morgan
2mo
Unless they hide job listings behind a login or paywall it's public data. Just look at Meta's recent failed filings against bright Data. Job data isn't a valuable data set unless it's value can be controlled
Richard Collins
2mo
Would love to know what I said - if only I was a subscriber 😂. Just hope it wasn’t anything bad lol
Hung Lee
2mo
lets definitely do a brainfood live on this
CTO at VONQ | Board Advisor | CPO | CMO | Founder
2moFor more discussion on this topic - there are some smart takes by Lieven Van Nieuwenhuyze Joel Cheesman and Chad Sowash on the most recent Chad & Cheese podcast - https://www.chadcheese.com/post/hellowork-heats-up-jobijoba-eyes-america