When we were building the featured queue, the obvious move was to automate it. Pull engagement data, weight by recent sales, maybe factor in click-through rate, and let the numbers decide. It would have been cheaper to run and easier to scale. We chose not to, and it is worth explaining why.
What an algorithm would actually optimise for
Any ranking system learns from what already works. On a marketplace, that means images with a track record of sales get surfaced more, which generates more sales, which pushes them higher. Work that is genuinely original — unusual colour palettes, niche subjects, photography that does not fit a familiar commercial mould — tends to sit near the bottom because it has no history to show the model. The algorithm is not malicious; it is just compounding whatever bias existed on day one.
A human curator can look at a new submission and recognise that it is interesting precisely because nothing like it has sold before.
How the featured queue actually works
Our team reviews submissions manually and rotates the featured spots regularly. We are looking at composition, originality, and whether the work adds something the queue does not already have. We are not looking at an artist's sales rank or follower count, because we do not think those are good proxies for quality.
This takes more time. We are a small team, so there is a queue, and not every submission makes it in. If your work is passed over, it is not a permanent verdict — you can resubmit, and the person reviewing it next time may simply be looking for something different that week.
How this connects to the rest of how Picster works
The curation decision fits alongside the other deliberate choices we made in how the marketplace is structured:
- Credits are priced in Euro at €0.15 each, with other currencies such as AUD converted from the live EUR rate, so pricing is consistent and transparent rather than set by regional guesswork.
- Artists keep a flat 50% commission on every sale, with no tiers, no exclusivity requirements, and no subscription eating into that figure.
- Buyers top up credits once when they need them — there is no recurring charge sitting on a card.
- Artists request payouts directly from their in-app wallet when they want them, rather than waiting for an arbitrary monthly cycle.
None of these are particularly flashy features. They are just structures that try to be straightforward about where the money goes and who makes decisions.
The featured queue is the same idea applied to visibility. We think the artist whose work is shown to buyers should be chosen by someone who looked at the work, not by a system that looked at last month's conversion data. It will not always be the most efficient approach, but we think it is the more honest one.
