How we rank what you see

• Discovery — the algorithmic feed

  Discovery is the default mode in your feed. Posts are ranked by an algorithm — here's what it weighs.

  • Engagement rate matters more than raw counts. What counts is the share of viewers who engaged, not the absolute number — so a smaller-account post that hits hard can rank ahead of a larger-account post that doesn't. Tiny samples don't dominate either: a post needs real exposure before its rate counts for much.
  • Engagement signals are weighted by quality. A save is worth +4 points, a comment +2, a share +1.5, a like +1. A save (you want to come back to a post) is the strongest signal of quality; a like is the easiest, weakest one. Self-engagement — your own likes, comments, and saves on your own post — is excluded.
  • Track record gives a small uplift, breakouts a small bonus on top. Creators with a strong engagement history get a modest boost on every post; creators with a weaker history get a modest dampening. A post that clearly outperforms its own creator's typical work earns an extra Progress Bonus. Both effects are bounded — they nudge the feed, never decide it.
  • Posts with real reach get a small Volume Bonus, so high-traffic work isn't beaten by similar-rate posts that simply had fewer views.
  • Brand-new posts get a 3-day visibility floor that fades gradually. Without it, posts that start cold would never warm up.
  • Older posts decay smoothly. A post's score halves every five days. Tattoos are timeless work, not news — a great piece deserves to be seen for a week, not a single news cycle.
  • The order shifts on every refresh. A small per-refresh shuffle reorders posts with similar scores, and posts you've already scrolled past in this session get pushed down so new ones can surface. Pushed down, not removed — so you can re-discover great work on a later refresh.

• Following — the chronological feed

  Posts from accounts you follow appear newest-first, full stop. No quality filter, no hidden ordering, no suggested posts mixed in. The day Following starts re-ordering posts behind your back is the day it stops being yours.

• Ranks — the monthly competition

  Tap the trophy at the top of the feed to open Ranks. Each calendar month is its own independent competition.

  What you see when you open it: last month's leaderboard. The rankings on screen are the final, frozen results of last month's competition — not whatever's happening this month. The current month's competition runs in the background; its leaderboard isn't revealed until the month closes.

  Why we wait until the month ends. If we showed live rankings during the month, posts published early would have a structural advantage over posts published later — they've had more days to accumulate views and engagement than something posted on the 28th. Putting early posts at the top of a visible leaderboard would amplify that gap and quietly punish anyone who posts later in the month. By revealing the leaderboard only after the month closes, every post in that month has had the same window to perform. (Hiding live standings also discourages gaming — timing posts, asking for engagement bursts.)

  Each month starts fresh. No carryover. Last month's winners get their permanent badges, but they don't start this month with any advantage. The same artist can win every month if their best post each month still tops the platform — but they have to earn it every month.

  One entry per artist. Each artist appears at most once on the monthly leaderboard. Top 100 means 100 different artists, not 100 posts. We made this choice for diversity at the top, fair prize money once the prize pool ships, and giving every artist a real shot at being seen. Tie-break: highest engagement wins; if two of your posts tie on engagement, the one with more views wins; if those are equal too, the one posted earliest wins.

  All your posts this month are in the running. Every piece you publish accumulates views, engagement, and a quality score. They all compete throughout the month — when the month closes, the system looks across all of yours and picks the highest-scoring one to take your slot on the leaderboard. So posting more great work this month is better, not worse: every additional piece is another shot at having one of yours land near the top — three solid posts give you three chances of one ranking high, versus a single chance with one post. Your non-winning posts still keep their score and still appear in feeds, on your profile, and in search.

  The scoring itself is a stripped-down version of Discovery — same engagement-rate calculation (with the same anti-fluke smoothing), extra weight on reach, no time decay, no jitter, no seen-scaling. The in-app Our Ranking System button (at the top of the Ranks screen) opens a visual breakdown of the algorithm and the vision behind the future prize pool.

• The math, for the curious

  For readers who want the actual formula.

  Pre-computed (cached on the post; recomputed every time anyone engages):

  • weightedEngagement = saves × 4 + comments × 2 + shares × 1.5 + likes × 1 (creator's own engagement subtracted)
  • engagementRate = (weightedEngagement + 100 × platformAverageRate) / (views + 100) — the +100 cushion is what stops a 5-likes-on-10-views post from looking like a 50%-engagement rocket
  • baseScore = engagementRate × progressBonus × authorQualityMultiplier × volumeBonus
  • baseScore = max(baseScore, newnessBonus) — 3-day visibility floor for brand-new posts

  Per-user, computed on every feed load:

  • timeDecay = 1 / (1 + ageHours / 120) — 5-day half-life
  • finalScore = baseScore × timeDecay × jitter × seenMultiplier

  …where jitter is a deterministic per-refresh random factor in [0.85, 1.15] (same refresh = same order, new refresh reshuffles ties), and seenMultiplier is 0.3 if you've already seen the post this session, otherwise 1.0. Posts are sorted by finalScore, descending.

  Ranks uses just engagementRate × volumeBonus (with extra weight on volume) and skips time decay, jitter, and seen-scaling.

Frequently asked questions

What's not in the Discovery score?

Follower count, account age, posting time, posting frequency, and anything paid. The score only looks at how the post performs relative to its views, the creator's own track record, the post's age, and whether you've already seen it. None of those excluded factors affect what surfaces.

Does a brand-new post stand a chance against established work?

Yes. New posts get a 3-day visibility floor that fades gradually, then they compete on engagement rate — which normalises for exposure. A small-account post that performs well can out-rank a large-account post that performs averagely.

Do my own likes, saves, and comments on my post count toward the score?

No. The displayed counts include them (so the number you see matches what other viewers see), but the algorithm subtracts them before computing the score. Self-views are excluded too — opening your own post doesn't inflate its view count.

Can someone game it by buying views or repeat-viewing?

Repeat views from the same account on the same post are capped per day. Saves — the heaviest weight at +4 — are also much harder to fake than likes. Beyond that, the formula is published: there's no hidden lever to pull.

I posted something this month — why isn't it on the Ranks leaderboard?

What you see when you open Ranks is last month's final leaderboard, not this month's in-progress one. This month's competition is running in the background; its leaderboard won't be revealed until the month closes. So a post you uploaded yesterday is currently competing — its result will appear next month, when this month's rankings get frozen and published. The reason we don't show live mid-month rankings is fairness, not secrecy: a post published on the 2nd has 28 days to accumulate engagement, a post on the 28th has 3 days, and showing them on a visible leaderboard mid-month would lock in a structural advantage for early posts (which would also get more visibility from being at the top, compounding the gap). Hiding live standings until the month closes gives every post the same window to perform. It also discourages gaming the system through engagement-burst timing.

Why can't the same artist take multiple positions on the monthly Ranks leaderboard?

Each artist appears at most once per month. We do this for three reasons. Diversity at the top: a Top 100 of 100 different artists is a better discovery surface than a Top 100 of 20 artists with their portfolios. Fair money once the prize pool ships: multiple ranked posts from the same artist would let a few elite artists capture most of the prize money — that's not the competition we want to build. Encouraging everyone to compete: if the leaderboard is "always the same 3 artists' portfolios," other artists conclude posting is pointless. There's no cross-month cap — the same artist can win every month if their best post each month is still the best on the platform.

To be clear: this rule is not a reason to post less. All your posts compete throughout the month — every piece you publish accumulates views, engagement, and a quality score, and they're all in the running. When the month closes, the system picks your highest-scoring one to take your slot on the leaderboard. So posting more great work in a month is better for your odds, not worse: three solid posts give you three chances of one ranking high, not one. We want active artists shipping their best work.

Will the algorithm change?

When it does, we'll publish what changed. The in-app Our Ranking System page already includes a version-history section. We'd rather argue about a documented change than have you guess what shifted.