Care work demand rose and platform workers fell

The smaller the task, the harder the hire

There is an odd line in the 2023 Platform Worker Survey that Korea’s Ministry of Employment and Labor and the Korea Employment Information Service published on August 5, 2024. Platform workers overall reached 883,000, up 11.1 percent from 795,000 the year before, but household and care work fell 1.9 percent. The release notes this happened “despite rising demand for care services driven by dual-income households and an aging population” and adds that “appropriate labor supply is needed.” Supply shrank in a category where demand grew.

Another line in the same survey reads like an answer. By working pattern, those for whom platform work is a main job fell from 57.7 to 55.6 percent, while occasional participants rose from 21.2 to 22.6 percent. More people want to work in short bursts. They are simply not meeting the short work in their own neighborhood.

Look from the task side and the reason shows. Thirty minutes to help lift a box. An hour to do someone’s grocery run. Two hours at a shop that suddenly got busy. These are too small to justify a job posting, below the minimum a company will send someone out for, and posting them to an online community means a stranger arrives with nothing attached to judge them by. So most of the time nobody is called at all.

Density needs free help on the same board

There is a common design mistake here: turning every one of these requests into a paid transaction.

Run it paid-only and the low-value requests die first. Put minimum wage, a platform fee and travel time on top of a thirty minute task and the number becomes too much to ask for and too little to go out for. When both sides pick “not today,” nothing happens in that neighborhood at all. In a two-sided marketplace the expensive failure is not bad transactions, it is no transactions.

Putting free help on the same board opens that band. Dropping off bottled water for an elderly neighbor while you are already at the convenience store does not need a price. What that person gets instead is a record: identity verified, tasks completed, promises kept, reports filed against them or not. When a paid request comes later, that record stands in for the price.

So free help is not a charity feature. It is the on-ramp that creates density. Remove the friction on the first interaction so people accumulate, then let paid work attach once trust exists. This also connects to the top complaint in the survey, “being asked to do work not in the agreement” at 12.2 percent. Early on, accept only requests whose scope fits in one line: grocery runs, deliveries, small lifting, short shifts at a shop. Childcare, anything medical, and high-value goods belong outside the launch categories.

The money is in the route, not the commission

Try to earn from the connection fee and this model eats itself. The moment you charge on free help it is not free, and every point you add to the paid fee comes out of what the helper receives, which pulls density back out.

The money is elsewhere: in the routes people are already moving along to help each other. If someone is traveling from Seongsu to Konkuk University to help, the platform can optionally surface a delivery or a grocery request that sits on that route with almost no detour. Where a delivery platform dispatches a new person to move one object, this attaches work to a person already in motion. The requester pays less and the helper earns more from the same trip. Most of the small brokerage revenue that comes out of this goes to the helper; only a slice stays for operations.

flowchart LR
  A[Request posted] --> B[Identity and trust record checked]
  B --> C{Free or paid}
  C --> D[Help performed]
  D --> E[Route-matched extra request offered]
  E --> F[Delivery and shopping micro-revenue]
  D --> G[Completion confirmed, mutual rating]
  G --> B

Three more revenue lines attach cleanly. Affiliate deals connecting neighborhood markets, convenience stores and laundries to grocery requests. Paid features for small business owners who need short-term hands repeatedly: hiring several at once, repeat requests, recalling a favorite helper. And sponsorship from local governments and welfare bodies covering insurance and verification costs for free help in a given district. None of these earn by raising the price of connecting two people.

The hard part is not the software

The hard part of this idea is not the code.

First, density. A hundred people moving inside one neighborhood beats a thousand scattered across Seoul. Pick one living area, start only inside it, and do not move to the next one until a posted request is reliably accepted within thirty minutes. Plenty of local services died by skipping that judgment.

Second, safety and law. People meet in person, so a single incident damages trust in the whole service. Brokering errands and deliveries means checking how far the individual statutes reach, and whether the arrangement counts as short-term employment or contract work changes which insurance and injury coverage applies. Accident cover and a dispute process belong in place before the first transaction, not after the first incident.

Third, the limits the existing players already reveal. The Korean errand app Haejuseyo reported 1.4 million members and one million cumulative jobs as of February 2024, with transaction volume growing from 180 million won in 2021 to 4 billion won in 2023, and 2 million members with 300,000 partners by 2025. Divide that 4 billion won of 2023 volume by the 1.4 million members reported at the time and you get under 3,000 won per member per year. Plenty of sign-ups, very few transactions. The bottleneck in this market is not acquisition. It is how many times one person uses it in a year.

Which means the first thing to validate is not downloads either. Does someone in one neighborhood actually ask a stranger for a small favor, does another person accept it, and do the two who had a good experience come back within a month. If those three fail, every other feature is premature.