The research that changed the brief twice

I wanted to design a job platform. Companies pay monthly, job seekers pay per service. Straightforward enough to start drawing screens on day one.

I did the research first instead, and it was worth the day. Two parts of my own brief were wrong.

The market broke recently, and the numbers are specific

AI made applying free. 78% of job seekers now use AI in their applications, 56% say they apply to more jobs because of it, and 77% of hiring teams regularly see AI written applications, up from 53% in early 2024.

So employers get buried. They screen harder with their own AI. Seekers respond by applying even more. Both sides are losing and neither can stop.

Underneath that, nobody can tell what is real. Between 18% and 22% of listings are ghost jobs, positions nobody intends to fill, and 93% of HR professionals say their employer posts them. The average dead end application cycle costs a seeker about nine hours. Going the other way, Gartner expects one in four candidate profiles worldwide to be fake by 2028, and 41% of organisations say they have already hired someone fraudulent.

Three wounds, and every platform I looked at treats one and ignores the other two.

Mistake one: the build order

A two sided marketplace needs jobs to attract seekers and seekers to attract employers. Indeed, LinkedIn, Bayt and Seek already own that liquidity in every market I could reach. Launching as a marketplace means launching into an empty room.

The way through is not a feature. It is to design a seeker product that works against jobs sourced from elsewhere, so it has value on day one with zero employers signed up, and open the employer side alongside it. Simplify built a real business on free autofill over other people's listings.

That reverses the build order in the brief, and it is the only version with a survivable first year.

Mistake two: arming both sides of the same war

This one took longer to see.

As briefed, the platform sells seekers volume applications and sells employers filtering. Those are the two sides of one fight. The moment the apply-for-me product works at volume, the employer customers are the ones complaining about the volume. The moment employer screening works, the seeker customers stop getting interviews.

The evidence is public. LazyApply is the volume leader in that category and sits at 2.4 out of 5 across 105 verified reviews, with about a quarter of reviewers citing refund or cancellation failures. Volume auto-apply is a refund machine, because the product promises interviews and delivers submissions.

So the seeker product sells fit and preparation rather than volume, and the apply-for-me service is capped at 10 to 25 a week with every application approved by the seeker before it sends. Both customers now want the same outcome, which is a hire that holds.

What the pricing research settled

Seekers will pay, but only for a visible edge. They treat listings as a free commodity and refuse to pay for access, then happily pay for advantage. The ceiling for self-serve seeker software is 29 to 40 dollars a month, based on Teal at 29, Huntr at 40, Simplify at 39.99 and JobRight at 39.99. Above that, buyers expect human hours, which is why reverse recruiters charge 150 to 720 a month self-serve and agencies charge 1,500 to 4,500.

The other half matters more. Seeker revenue churns by design, because the customer's goal is to stop being your customer. Average lifetime in this category is two to four months. That is why the seeker side sells non expiring credit packs rather than a subscription: a credit pack produces a purchase and no bad feeling when someone gets hired, while a subscription produces a cancellation.

Employers are the durable side. An employer subscription renews for years. SMBs under 100 staff spend 900 to 3,000 dollars a year on hiring software, which sets the ceiling.

So the seeker side funds the early months and the employer side is what makes the company worth something.

The position nobody had taken

Here is the gap. Ghost jobs are 18 to 22% of listings, 93% of HR people admit their employer posts them, and not one platform labels them.

Verified listings plus a published employer response rate is a position none of the incumbents can copy without admitting they were the problem. Regulators are already moving: Ontario legislated disclosure for 2026 and California introduced AB 1251 in March 2025.

That became the promise the whole product is built to keep. Every job here is real, and every employer answers.

The part that changes the architecture

One finding is not a positioning decision, it is a build constraint. The moment the platform ranks candidates for an employer, it is a high risk AI system under the EU AI Act and an automated employment decision tool under NYC Local Law 144, which requires an annual independent bias audit, ten business days of candidate notice, and penalties up to 1,500 dollars per violation per day.

That means audit logging, human override and a hard rule that no code path rejects anyone automatically. All of it is cheap to design in and expensive to retrofit, which is exactly why it belonged in the research and not in a later sprint.

Next up: picking a name, and finding one shape to sign the whole system with.