Working with trends & topics

A single article rarely tells you much. A trend does: it is the cluster of publications that are all covering the same evolving story, grouped so you can watch it rise, peak and fade as one thing. D2TR builds these clusters automatically — every article we index is embedded and assigned to a trend — then gives each trend a lifecycle state and a momentum reading so you can tell an emerging story from a spent one at a glance. This guide covers how that clustering works, the theme → topic → trend hierarchy it lives in, and how to use the Trends view and the Top trends rail to act before a story peaks.

#What a trend is

As each publication is parsed, summarized and classified, it is turned into a vector (an embedding of its text) and compared against existing trends. If it is close enough to a cluster, it joins that trend; if nothing matches, it seeds a new one. So a trend is not a keyword or a hashtag — it is a living group of articles about one story, and its shape over time is what makes it useful. Two properties matter most when you read one:

• Lifecycle state — where the story is in its arc. A trend moves from just-appearing, to peaking, to holding on, to going quiet. You will see states such as emerging, hot, persistent, seasonal_dormant and decayed. The state is what lets you separate a story that is still climbing from one that has already crested.
• Momentum — how fast the trend is accelerating right now, read from its daily article counts. Rising momentum on an emerging or hot trend is the window you want: the story is gathering coverage but has not saturated yet. Momentum that has rolled over means the peak is behind you.

Each trend also carries a label in several languages, a short summary, its member article count, first-seen and last-article timestamps, a peak date, and the key entities and topics that define it. Together these turn a raw cluster into something you can scan and decide on.

#Themes, topics and trends

Trends do not sit in a flat list. They roll up into a two-level hierarchy so you can navigate from the whole agenda down to a single story:

• Themes are the broadest level — the big subject areas, like sports or elections.
• Topics group related trends inside a theme — for example a specific competition or a single race.
• Trends are the leaves: one story each, with their own lifecycle and momentum.

Read it as a tree: theme → topic → trend → articles. A theme answers "what broad area is busy", a topic narrows it to a sub-story, and a trend is the story itself with the publications underneath it. Use the higher levels to orient — which themes are heating up — then drill into the topic and its trends to find the specific angle worth acting on.

#Where to find trends in the app

There are three main touch points for trends, and they serve different jobs:

• The Trends view (in the ANALYSIS group of the left sidebar) — the full, filterable list of trends for your project's markets. This is where you go to browse, filter by state and drill into a single trend.
• The Top trends rail in the Explorer's right rail — a ranked shortlist of the trends moving most right now, each with a value, a delta versus the prior window, and a View all link into the full Trends view. This is your at-a-glance "what is moving" list while you read the feed.
• The Assigned trend column in the Explorer's publications table — every publication shows the trend it was clustered into, so you can jump from a single article to the whole story around it.

There is also a Trends entry under the TRACKING group with a live count. Tracking a trend keeps it on your radar and feeds the mention counts and alerts across the app — see the Tracking guide for how tracked items work.

#Act before a story peaks

The whole point of watching trends is timing: you want to be publishing while a story is still climbing, not after it has saturated. Here is the flow.

1. Scan the Top trends rail in the Explorer
   Open the Explorer for the market and window you care about (start with 24H for breaking stories, 7D for the broader agenda). In the right rail, read Top trends: the ordering and the deltas versus the prior window tell you which stories are accelerating. A large positive delta is a story on the way up.
2. Open the trend and read its state and momentum
   Click through to the trend. Check its lifecycle state first: an emerging or hot trend with rising momentum is the opportunity. A persistent trend is a durable story you can plan around; decayed or seasonal_dormant means the moment has passed (or has not come back yet). The trend's daily counts show the shape of the curve — you want to be on the left side of the peak.
3. Read the member articles to find the angle and the gap
   The trend lists its member publications — each with its publisher, language, impressions and the role it plays in the story. See Reading a trend below. Skim these to understand what has already been said, which publishers are covering it, and what angle is still open for you.
4. Check reach and entity context
   Look at the trend's reach (its rolled-up impressions) to size the opportunity, and its key entities to see exactly who and what the story is about. A rising trend with real reach and entities you cover is a strong signal to move.
5. Track it, then turn it into a brief
   Track the trend so it keeps feeding your alerts and counts as it evolves. When you are ready to act on it, hand it to the Publisher Center to turn the trend into a brief and a draft.

#Reading a trend

Once you open a single trend, there are three things to read together. Each answers a different question about the story.

#Member articles

The publications clustered into the trend — this is the story's coverage. Each article carries its title (with an English translation), publisher and domain, language, publish and first-seen times, a summary, and its impressions. Articles also carry a role in the story: the piece that kicked it off, the pile-on coverage that followed, the long-tail that keeps it alive, and one-off mentions. Reading the members in order tells you how the story broke, who amplified it, and where it still has room to run.

#Entities

The key entities a trend is about — the people, organizations, places and products at the center of the story, extracted from its articles and linked to the Google Knowledge Graph. Entities are the fastest way to confirm what a trend actually is (two trends with similar labels can be about completely different people), and to connect a story to the entities you already track. For the entity side of the app in depth, see the Tracking guide.

#Reach

A trend's reach is its estimated impressions in the Discover feed — the per-URL reach of its member articles, rolled up to the trend. It is the same impressions-style metric you see on the Est Impressions chart and the Impressions column in the Explorer, just aggregated to the story level. Reach tells you how big the story actually is in the feed, and lets you compare one trend against another on the same footing. Read it alongside momentum: rising momentum tells you the story is accelerating; reach tells you whether it is big enough to matter.

#Trends in the API

Everything in the Trends view is available programmatically — the app and the API read the same trend graph. The Trends API lets you list trends filtered by lifecycle state, walk the theme → topic → trend hierarchy, and pull a trend's member articles, key entities and daily counts. Use it to wire trend signals into your own dashboards, alerts or content pipeline. Note that trend endpoints require an API key and are scoped to the markets your key allows.