People ask how Podkon decides what is trending, so here is the honest, technical version. There is no black box, just a few well chosen signals, read carefully and ranked.
One source, read within its limits
Everything starts with Etsy’s official API. We do not scrape pages and we do not buy second hand data. We sweep continuously, bounded by our daily quota and rate limits, which works out to millions of requests a day. We do not claim to read every listing. A token based rate limiter in front of each worker only ever slows us down, and we spend the budget on the strongest signals first, then keep rolling forward. That is the honest framing: we cover as much as the API allows, prioritized, rather than pretending to capture the whole marketplace at once.
The three signals we follow
For every listing we track three things over time, and each one means something specific:
- Reviews. The review count on a listing, and how much it gained. A review can only follow a delivered order, so a new review is a confirmed sale. This is the signal that cannot be faked, which is why it drives the ranking.
- Favorites. The number of shoppers who have favorited a listing. An early sign of interest before a purchase.
- Views. How much attention the listing is pulling.
We also show favorites divided by views, which is a rough proxy for shopper interest. It is not a conversion rate, and we label it that way, because calling it conversion would be a number we cannot actually prove.
From raw counts to a confirmed sale
Reviews are deduped by transaction, so the same purchase is never counted twice. We aggregate the distinct transactions per listing per day, which gives us a clean daily sales floor. We call it a floor on purpose: plenty of real buyers never leave a review, so the true number of sales is higher than what we show, never lower.
How a number becomes a rank
We rank by movement inside the window you choose, not by all time totals. A listing that picked up reviews this week beats one that peaked two years ago. To measure that change cleanly we snapshot favorites and views once a day and compute the difference across the window, using valid readings only, so a single missing day cannot make a number explode or collapse. Reviews are aggregated per day and the gain is computed over the same window. Then we order the movers and publish.
Windows, and why they matter
The same data answers different questions depending on the window. The daily and weekly views catch fast moving categories while a trend is still rising. The fifteen day and monthly views smooth out the noise for slower categories, so you see the real direction instead of one good day. Picking the right window is half of reading a trend correctly.
What we will not claim
- Sales counts are a floor, not an exact total.
- Prices are the listed price, before any shop discount or coupon.
- Favorites per view is a proxy for interest, not a conversion rate.
- We measure demand and movement. We do not promise that copying a product will sell for you.
We would rather show you exactly how the result is built than hand you a single score with no explanation. If a number ever looks off, that openness is the whole point.