The End of the Job Search.
As a former marketplace investor, I’m of course a big fan of the theory of unbundling of classifieds and verticalised approach to marketplace.
One of the most recent articles on this topic that I get sent a lot by my wonderful network is the one by A16Z:
Platforms vs Verticals and the Next Great Unbundling
One of the most effective forms of that competition often comes in the form of newcos who aspire to take chunks out of that emergent platform by better addressing the needs of a specific vertical within that platform — by creating a user experience or business model that’s much more tailored to the unique attributes of that vertical
And of course, identifies LinkedIn as the next big victim of this inevitable innovation cycle and goes on the list the verticals that have already been disrupted through the unbundling.
We are already beginning to see innovation bubbling up with newcos using this vertical play in the jobs platform space. From engineering (Hired) to blue collar (Merlin, Wonolo) to oil services (RigUp) to hospitality (Pared, Instawork, Qwick) to bookkeeping (Paro), each of these new companies are building a user experience and business model that works better — for both candidates and employers — than the generic LinkedIn model. In other words, the great unbundling of LinkedIn may have already begun.
The reason why the first verticals to be stripped away from LinkedIn are engineering, oil services and hospitality is not a coincidence. These verticals are the first to fall victim because they have relatively simple matching at the core, based on obvious proxies.
For example, one of the investments I worked on back at Piton Capital, was Lantum — the leading UK (and now US) platform for Locum, or temporary, doctors and nurses. It’s characterised, like many of the above platforms A16Z mentions, by two things:
- A great oversupply of jobs, and a scarcity of supply-i.e. professionals
- A clear quality indicator — after all, you’re either a GP with a qualification or not
- A need for incredibly fast turnarounds, requiring hyperlocal liquidity — e.g. local GP clinics needing a doctor the next day
What’s interesting is that in this case, and in the examples above, there is no complex matching involved, and the threshold, or the minimum matching criteria is relatively easy to meet through technology.
This is what I saw when I started looking at the space.
The question I kept asking myself is — what about that other huge market out there? What about those people you see walking down the street, going to desk jobs. Are they really happy with their job hunting experience?
For jobs like white collar roles [Linkedin] works well.
Is that really true?
(In case you’re a venture capitalist reading this… don’t think of your own job search, but that of a real white collar user — ask your receptionist, or HR manager, or marketing assistant what they think! Or better yet, try to hire one :-))
If so, why is the recruitment agency market still worth billions in any country, but particularly in the UK and the US, where most of us are keen and want to find a job online? What are those intermediaries doing that’s of such a value add that companies are willing to pay £1000s for?
What do you really need to transform this market?
You need to solve the lemon problem. You need data to bridge the gap.
The value of a platform is not solely in creating liquidity. It’s about overcoming an information gap. One of my favourite undergraduate economics papers of all time surely is about the lemon market. Akerlof’s 1970 paper is a cornerstone of modern micro-economics, and something that (thankfully) stuck with me even 10 years later.
A long story short, he outlines that when selling a used car it’s hard to assess the quality, and therefore it’s hard to establish the price. As a result a market equilibrium is only reached with poorer quality cars on the market, as the average price a consumer is willing to pay when they’re unaware of the quality of the car is too low for car-owners who have great cards to put them on the market. This pushes price down further, and so-on.
The interesting thing about this paper for me is not just the inefficient equilibrium which the market finds itself in, but more the huge business and social opportunity for an intermediary to solve the information asymmetry.
I truly belief this is one of the reasons why Auto1 is so huge and one of Piton’s best investment (definitely one of the most interesting deal I worked on). The company physically buys second hand cars from consumers and sells them onto dealers, often also leveraging international arbitrage — did you know they drive Japanese cars, with the steering wheel on the wrong side, in Eastern Russia?
The Lemon Problem in the Labour Market
It’s no longer just about building marketplaces with superior liquidity and customised user experience that fits that vertical.
Certain verticals require much more than that, as they face an information asymmetry.
In white-collar verticals, professionals do not understand what jobs are a good fit for them, both in terms of what they love doing at and what they’re good at. Businesses, in turn, have an incredibly hard time assessing professionals and making the right hires.
This is not a new idea of course, it’s not even mine. Nobel Prize winner Michael Spence wrote this already on labour market inefficiencies in 1973. In the labour market, this process of bridging information asymmetry is called recruitment, and the signals used are education, qualifications, personal presentation and the CV. Based on signals, we can draw implicit conclusions from the limited data they have, and make (often biased) decisions.
But then again, sometimes academic insight takes a while to make it to real life impact…
Replicate and Enhance Human Decision at Scale
This is the question that I’ve really been rather obsessed with for the last few years.
To replicate a human judgment on whether someone could be right for a job, you have to do several things:
- Understand what a company is looking for when making a hire for a particular role (rather than just posting a job ad, interpret what the real hiring criteria are)
- Understand what a professional is like — both in terms of what they would love to do, as well as what they are good at (not just what is written on their CV)
- Match the two to ensure both the business and the professional believe this is a good enough fit to at least schedule an interview and, eventually, make/accept an offer to start in the job (rather than semantically relating words in the job ad to the CV)
So what did we do? We started matching Personal Assistants to jobs, measuring a range of traits about the professionals and the jobs and companies. I’m not kidding, that’s it.
We did this by not just using Machine Learning techniques, but rather by incorporating Behavioural Data Science to ensure the data that we are capturing about people and businesses actually makes sense.
It took us 12 months to gather enough data (that was initially trained by myself and later by our clients, through our MVP product) to build the first model. Funded through friends & family, and lots of R&D Grant Funding, we reached a point where our models were better than us at predicting who a business would deem suitable for that particular role.
Vertically expanding, we now cover over 10 career paths and are on a mission to make this the future of job recommendation and matching.
Job searching sucks. That’s why we have killed it.
Imagine what it’s like to search, one by one, 20 million jobs. The thought is quite daunting to me.
Using data, we can eliminate search and fundamentally transform the user experience and business model. More personalised matching and platform liquidity create benefits for both users on the transaction:
- The Recruitment Assistant: Companies and HR teams can offload top of funnel recruitment tasks such as sourcing and the initial stages of phone screening, especially in high-volume white-collar hiring, freeing up time for value-add tasks.
- The Career Advisor: Professionals don’t just have access to many job opportunities, but they can actually find out which fit their experiences & interests. A helpful career coach who can point you to the right jobs for you, keep you posted on unique opportunities when you’re not looking, and give you personalised feedback.
This is just the beginning. What’s next?
Our thinking on People Data has evolved much beyond this in the recent year, and we’re now working on realising our broader vision: go beyond predicting hiring outcomes, and move into more objectively predicting employee performance outcomes. These insights, in turn, can be used to optimise job matching and help people find jobs that they will truly excel and be happy in.
If you have read this far and are curious about learning more about the category we are creating — People Analytics — click here to read my co-founders short introduction to the category.
