Hello friends 👋,
The Pallet Post features conversations with awesomely insightful professionals, jobs from companies they recommend, and some of our own takeaways from building a startup.
This post is going to be split into three parts, and highlights a lot of the reasons my co-founders and I built Pallet.
It all began with a simple question:
Why does job searching feel so unnecessarily hard?
When I search for jobs, for some reason or another, I often find myself staring at a listing that has literally no relevance to me. I don't fault myself, I searched for the thing I supposedly wanted, but nonetheless I'm staring at a marketing role that involves using SQL, building dashboards, and optimizing user acquisition funnels on landing pages when what I'm really looking for is more content oriented. If you're good, you'll recognize these differences by the title of the job, if you're really good, you'll know to search for something specific, like the title you want, before you start, but even the best, even the right search, wont always give you the results you need.
Sometimes, it's a casualty of the whole shebang, you click on something you know isn't right, but why the hell not? You've already looked at fifty jobs what harm could the fifty first do.
If this isn't your experience: congratulations. Truly, I envy you. If it isn't, welcome to my super secret club that only includes LinkedIn's entire user-base.
In any case, I was struck by this question (the bolded one up top), as were my co-founders, and the following essay outlines some of our answers.
For a long time, a horizontal aggregation is the product architecture that won online-job search. Despite the rise of vertical marketplaces, it still wins today. Legacy products are relying on ideas and technologies that are laughably old— but that's precisely what's built a veil on their flaws. Problems become subliminal, in many ways engrained into our way of life. When something "isn't supposed to be easy" the companies that offer solutions for that thing, are begging to be disrupted. There are so many opportunities to win in the recruiting market, whether it be discovery of roles, or of interested candidates, building truly ethical vetting practices, remote-hiring in international markets, the list goes on and on.
For the purpose of this essay, we'll start by focusing on discovery. And the damn search bar.
In 2005, sites like Monster, and Indeed were a pretty logical conclusion to the way people sourced jobs offline. If you're looking for a job, you want to see all the possible options in one place. And while I wasn't alive to fully experience the days of newspaper listings, from quick google searches it seems all the possible options, literally meant all the possible options.
With the internet, "all the possible options'' extended beyond the surface area of a page in the newspaper. All of the sudden, you needed a database where all these items could live. You needed to provide filters and a search bar so people could sort through what they wanted. And presto! Along came online job-search. For a moment, it was good. Or, at least, there was nothing better. Companies began pouring these sites with their job listings. For them, the value prop was clear:
- Reach larger audience
- Inbound > outbound
- Focus on vetting
- Among other things I'm sure but hey, rule of three
- Shit, I made it four things
- Now it's five
- You get the point...
So what happened in-between? What happened in the time between job search's illustrious birth and when I first laid eyes on an Indeed search in the summer of 2016? Because honestly, I'm sure that the beginning was awesome. I'm sure that everybody who was mailing in resumés, looking through different newspapers, finding people on yellow pages, we're happy to get a search bar. (Apologies if that is nothing like the days of pre-internet life). But then what? What comes after the search bar? Why has the whole thing bloated itself ad-infinitum with noise and lackluster results?
Job aggregators are not really designed to understand your interests
Online jobs products are destinations, they take you away from the rest of your life so you can "be on the hunt"
Recruiting marketplaces have negative network effects— the larger they are, the noisier they get
Online jobs products fail in three key ways:
- Product Architecture - How it works.
- Product Market Fit - How it fits into our lives.
- Network Effects - How it grows.
Part One: Product Architecture - Users can't search if they don't know what they want.
Almost all large-scale job aggregators employ a search-based product architecture. Search-based platforms (think Amazon, Google, Indeed) are, exactly as they sound, products that rely on queries from users to output specific information. This model works great if a user knows exactly what they want.
If you're looking for a fridge, you go on Amazon, and buy a fridge. The product experience is usually frictionless.
But what if you have a broad set of interests and don't know exactly what you want? What if you're unsure on which searches will deliver relevant results?
Search doesn’t satisfy this user-persona. Tik Tok, Youtube, Spotify, Netflix all understand that discovery-based architectures win when it comes to content-delivery.
Discovery-based products create both broad and niche channels that users can subscribe and follow— and these usually become the mechanisms through which they're delivered content.
When it comes to the search for entertainment, videos, songs, or shows, you usually aren't looking for something hyper specific. You're looking for some general themes that allow you to sequester your search. A superior search. You're looking for a set of algorithmically pre-selected options to choose from. Sometimes you want dance videos, sometimes you want comedy sketches, sometimes educational content. Whatever it is, you usually want the platform to do some of the "searching" for you. Then, once you've used the product a couple times, you want it to begin to understand you. You want it to pick up on your interests as you go along.
How does a product understand you?
Discovery-based products are designed to understand you. That's the whole value prop, otherwise, how are you supposed to find the things you like? Whatever was in your instagram explore page in 2018 (they may have fallen off a bit) was probably a pretty good reflection of what you like.
You can interact with them in a multitude of ways. Think of Twitter. You can like something, retweet, quote retweet, view someone's profile (without following), follow, set a notification to keep track of them, and of course, tweet. Twitter might not be the best example of a product leveraging signals (over something like Tik Tok), but you can see how the architecture of the product itself— how it invites you to use it, has the potential to create a robust understanding of who you are and what you like.
Search is particularly weak when it comes to picking up signals. The most tangible reality I can think of— I often find myself staring at ads based on things I searched for, but they only seem to reaffirm how much I don't want that thing. It’s like we’re doomed to a future we don’t want, like our noses get rubbed in the fact that we weren’t certain in the first place. Simultaneously, I can't really search for things that I know nothing about. So any algorithm that is "learning" off the basis of my search queries, is receiving an incomplete picture of what I might want.
Most people, even if they’re very knowledgeable probably have the following relationship with information in the world:
There are more things the average person would consider than there are things they directly know about. Any search-based platform loses out on an entire world of content that could be applicable to a person. In an age where algorithmic recommendations are responsible for a large portion of what we see, this is a pretty damning pitfall.
To illustrate the point: The first three seasons of the show Breaking Bad had moderate viewership and generally positive reviews. It wasn't really that huge. What happened?
Between the third and fourth seasons, Breaking Bad saw a huge increase in viewership due to Netflix's recommendation engine— Netflix recognized there were users who were interested in that type of content and pushed it in their feeds. It quickly became one of the most-watched cable shows on American TV.
People didn't search for Breaking Bad, Netflix showed it to them.
When it comes to job-search, discoverability becomes a pervasive problem, on the B2B side, for any company whose name isn't Facebook or Google. In spaces where new and unknown players emerge quickly and often (like Tech), job-aggregators have no ability to help job-seekers discover worthwhile companies that have yet to generate proven search engagement. And the negative flywheel is created: companies are confined to terms people are already searching for and people are confined to search terms they know will generate results.
Job-aggregators not only predominantly use search, but also haven't tweaked (or remotely even touched) their architecture in almost two decades. Take a look at Indeed's beta product in 2005:
Now look at the product today:
Their platform is virtually identical to its initial rollout— users still need to input keywords and locations in order to receive the output of jobs. Lets superimpose this model to an area where discovery is commonly understood to be an important aspect of the user experience: Spotify.
If you wanted to listen to some jazz music, would you rather have the ability to peruse already-curated playlists that feature a large assortment of jazz artists, or search for them one-by-one? You can only imagine how difficult it would be to discover new music if Spotify looked like this:
At this point, you might be saying to yourself, "well, job listings are different than songs, or shows", but are they really? I'm proposing that job listings and entertainment actually share two key defining aspects: genre and intended audience.
The genre of a job listing can be determined by its composition, which usually looks something like this:
- Job title
- Company Stage / Size (often correlates with culture)
- Skills involved
- Responsibilities / Function
When you add all of these elements, you get a specific "type" of job. Each individual portion functions as a part of that job’s DNA.
Take two product manager jobs with the exact same title, industry, and skills involved, but one is at a FAANG and the other at a seed-stage startup— by changing one attribute, the entire scope of the job changes.
Job listings intend to speak to a specific group of people— just as a video of an athlete performing acrobatic feats intends to speak to a specific group of people. Mapping a job listing to its intended audience is one of the major failures of large-scale aggregators. If you're on the hiring side, you're sure to have received applicants that weren't quite qualified for what you're looking for. Or maybe applicants who had some interesting skills, but probably would perform best in a different function. Did you use the wrong title? Was the description not suited for the people you're looking for? Do you regret not explicitly stating the years of experience you wanted? Did you miss out on adding a remote friendly option?
So the frustration extends past the search on the job-seeker's end.
But speaking of the job-seeker...
This is all to say that at the end of the day, job listings are just like any piece of content you might find online. They have a type, a genre, they intend to speak to someone specific, yet we're stuck searching for them like you'd find a video on youtube in 2008— one-by-one, and with no prior understanding of who you are and what you like. It's funny, only in job-search are recommendations met with such hesitation. And yet, we see a very tangible world where that's no longer the case. A world where you can easily discover the companies that allow you to articulate your passions, a world where a DM from a recruiter is more likely to be an opportunity worth pursuing than a wasted moment in time, a world where the "good jobs" aren't hidden behind the veil built by legacy products. Do you think it's possible?
Isn't it pretty to think so?
If you'd like continue reading, we've published part two of this essay here.