> ## Documentation Index > Fetch the complete documentation index at: https://docs.upriver.ai/llms.txt > Use this file to discover all available pages before exploring further. # Sorting & Ranking > How the trend sort modes differ, what each one is for, and how the freshness and rising signals behave The [Trends List endpoint](/trends/overview-1) supports three sort modes via `sort_by`. Each answers a **different question**. Most confusion about trend ordering comes from expecting one mode to answer another mode's question. | `sort_by` | The question it answers | Use it for | | ------------------------- | ------------------------------------ | --------------------------------------- | | `recommended` *(default)* | "What's worth my attention overall?" | Browsing; a balanced view of what's hot | | `newest` | "What's new since I last checked?" | Daily/weekly catch-up on new arrivals | | `rising` | "What's inflecting **right now**?" | Timing: catching a wave before it peaks | ## Biggest is not the same as rising `rising` ranks by **acceleration of posting activity**: is this trend's content being created faster than its own recent baseline? It does not rank by total size. A trend with millions of views and a large, steady stream of videos is *established*: it tops `recommended`, but its curve is flat, so it sits low on `rising`. A small trend whose activity is compounding week over week is the one `rising` is built to surface. That's the window where creating content for it puts you early. Two trend activity charts: a large steady trend that ranks first on Recommended but reads Steady on Rising, next to a small accelerating trend that reads Rising Two trend activity charts: a large steady trend that ranks first on Recommended but reads Steady on Rising, next to a small accelerating trend that reads Rising If a 6.1M-view trend appears below a 43K-view trend on `rising`, that is the sort working as intended: the smaller trend is accelerating harder *relative to its own baseline*. For "biggest overall", use `recommended`. ## Rising means rising *now* The rising signal is anchored to a trend's **latest evidence** and fades as that evidence ages. Two trends with the same climb rank very differently if one climbed this week and the other went quiet three weeks ago: Two identical climbing curves; the one whose latest video is 2 days old reads Rising +0.29 while the one quiet for 3 weeks reads roughly zero Two identical climbing curves; the one whose latest video is 2 days old reads Rising +0.29 while the one quiet for 3 weeks reads roughly zero Each trend's `momentum` value ranges from **−1 to +1**: positive = building, negative = fading, near zero = steady, too little evidence to judge, or evidence that has aged out of relevance. There is no cliff: the value decays smoothly, so a trend that was hot three weeks ago drifts down the list rather than vanishing. ## Freshness follows the most recent wave `freshness_score` (0–1) measures **how recent a trend's content activity is**, weighted toward its newest wave of videos. It is a *recency level*, independent of size. A trend can be small and fresh, or huge and stale. Trends often resurface: a format that had early buzz and then a strong second wave reads as fresh again, because the score follows the recent wave rather than averaging over the trend's whole lifetime. A trend with early buzz plus a new second wave scoring 0.62 active, next to a trend with early buzz only scoring 0.18 stale A trend with early buzz plus a new second wave scoring 0.62 active, next to a trend with early buzz only scoring 0.18 stale `timeliness_status` bands the same signal into `recent` / `active` / `stale` / `cold` if you prefer labels over thresholds. If you're rendering your own "fresh" or "hot" badge, decide which question it should answer. **"Recently active content"** → threshold on `freshness_score` (or use `timeliness_status`). **"Act on this now"** → threshold on `momentum` (e.g. ≥ 0.1). **"New to the catalog"** → sort by `newest` and compare `emerged_at` against your last visit. Mixing these produces badges that feel wrong on big-but-established trends. ## Field quick reference | Field | Meaning | | ------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `freshness_score` | 0–1 recency of content activity, weighted toward the newest wave. Size-blind. | | `timeliness_status` | The same signal as labels: `recent`, `active`, `stale`, `cold`. | | `momentum` | −1..+1 trajectory of posting activity, decaying as the latest evidence ages. | | `emerged_at` | When the trend first accumulated enough confirmed videos to count as a real trend in our catalog. This is when *we confirmed it*; for formats where we backfill older example videos, it can be later than the oldest video you see on the trend. | ## Which sort for which workflow * **"I check in every morning"**: use `newest`, and stop at the first trend you recognize from yesterday. * **"I plan content weekly"**: use `recommended`, filtered to your categories. * **"I want to catch waves early"**: use `rising`, and prioritize entries that are also `recent`/`active` on `timeliness_status`. ## Legacy sort values Older `sort_by` values keep working forever; they resolve to their closest current equivalent instead of erroring: | Value sent | Resolves to | Notes | | ------------ | ----------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `momentum` | `rising` | Renamed; identical behavior. | | `timeliness` | `newest` | Retired as a distinct sort in July 2026. For a freshness-ordered view, use `newest` with the `timeliness_status=recent` filter, or order client-side by `freshness_score`. | | `emerging` | (unchanged) | Still accepted as its own ordering, by `emerged_at`. For most catch-up workflows `newest` is the better fit. |