The pipeline
How Nuvelics turns articles into events, pitches, stories and outputs, and where people decide.
Nuvelics works in four stages. Each stage hands a smaller, more useful thing to the next, and the people in your newsroom decide at the points where judgement matters.
| Stage | What it produces | Where you see it |
|---|---|---|
| Articles | Everything your data sources collect | Articles |
| Events | Articles about the same occurrence, grouped | Events |
| Pitches | A scored story suggestion for one newsdesk | Pitches |
| Stories and outputs | A researched story, and the pieces written from it | Stories |
Articles
Data sources bring articles in: feeds, wire services, social accounts, council portals. An article is one document from one source at one URL. Nothing an editor works with is an article, though. Editors work with what happened, which a dozen outlets may have reported, some of them with identical wire copy.
Events
The understand stage turns the stream of articles into events. It does three things, cheapest first. It folds literal re-runs of the same text together, so a wire story carried by forty stations counts once. It groups the articles that report the same happening, even when they share almost no wording. Then it tags each new event with topics and locations.
That tagging is also the routing decision. An event becomes a candidate for a newsdesk when it overlaps every list the desk fills in: topics, locations, or both. A list the desk leaves empty places no restriction, and a desk with neither matches nothing. This is why a newsdesk's topics and locations matter so much: they are how it is matched to events.
Pitches
For each newsdesk and each event routed to it, Nuvelics writes a pitch: a title, a summary and why it matters, and a score from 0 to 100 against the desk's beat. A newsdesk is evaluated for an event at most once. Writing and scoring are separate steps so writing and judging cannot influence each other.
Pitches are ranked rather than simply listed. The ranking blends the stored score with how many sources now report the event and how recent it is, so an event that gathers corroboration after it was pitched climbs the queue without being pitched again.
Pitching is the first stage a person decides on. A pitch waits in the review queue until someone approves or dismisses it. You can also write a pitch yourself from notes, links, files or library articles; see Lifecycles for how manual pitches start.
Stories and outputs
Approving a pitch creates a story. Stories also arrive from your newsroom system, in which case they have no pitch and no newsdesk. Nuvelics then enriches the story: it researches beyond the original articles, structures what happened and checks the claims. See Enrichment and fact-checking.
An enriched story is where people pick the outputs they want. An output is a piece written from the story using an output template: a web article, a short-form video storyboard or a post for a social platform. Nothing is generated until a person asks for it. A story can stay enriched without ever getting an output. The story can then be pushed to your newsroom system, and web article outputs can be filed into your content management system.
Why it works this way
- People decide at the two points that matter. Nuvelics proposes pitches and writes outputs, but a person approves what becomes a story and chooses what to publish.
- The free step runs first. Folding duplicates costs nothing, so the paid steps do not spend on the same text twice.
- Approval is one-way. Approving creates a story an editor may already be working on, so an approved pitch cannot be reversed. Dismissing creates nothing, so it can be.
- Spending is bounded. Every step Nuvelics pays for counts against your spend caps.