How the market trends board is built

Every figure on the board is computed from the stories GovWire has collected from the public sector and partner technology press, and it is recomputed on every daily refresh.

The universe is the set of distinct stories in the window being measured, matched on the article's own URL so a story that appears in two feeds is still one story. Every trend's share is that trend's mentions divided by the universe, which is why the shares on a given refresh always add up sensibly against one number.

Themes and topics come from a fixed taxonomy, not from a model reading each article afresh: a story is matched to a trend when its language, headline or body touches the terms that define it. A topic always sits under one theme, and a brand is matched by the company name and the terms that identify it.

Because the taxonomy is fixed, the same theme means the same thing from one refresh to the next. A new theme is added deliberately, and a trend that was not on the board last month can therefore appear as a breakout rather than as a number that silently changed its meaning.

Mentions are counted over the last 30 days. Growth compares that count with the 30 days before it, so a trend is always read against its own recent past rather than against the board average.

The 90-day comparison measures the last 30 days against the 90 days before it, expressed per 30 days so the two windows are the same length. A trend that is up 40% on the month and flat over the quarter is a spike, not a shift, and the two numbers together are what make that visible. A trend with no prior mentions reads as a breakout rather than as infinite growth.

Each trend is placed in one of five states, read from the last three 30-day windows: emerging, accelerating, peaking, fading or steady. Steady means volume is holding; the other states describe which way it is moving and where in its own arc it sits.

The state is a reading of the trend's own history, not a prediction. It answers whether this conversation is arriving, growing, cresting or thinning out, which is the question that decides whether to act now or wait.

Tone is the share of recent stories that read positive, neutral or critical, taken from the language of the stories themselves. Framing separates stories about something opening up, such as awards, funding and pilots, from stories about something going wrong, such as breaches, lawsuits and delays. A trend can be loud and mostly risk, which is a different signal from loud and mostly opportunity.

Breadth reports how many publications carried the trend and how many stories each ran, so wide coverage across many outlets reads apart from one outlet repeating itself. First mover names the publication that carried the trend earliest in the window and how many days it led the next publication to it.

Deals are the largest spending and award figures the recent stories put on the record, shown as dollar values in millions with a link to the story that reported each one. Agencies named lists the agencies the stories name outright, most mentioned first.

The hiring signal counts open roles at Exchange partners whose titles name the same subject as the theme. It is read from the job boards the companies publish themselves, and refresh on the same daily schedule as the rest of the board. It is a real headcount reading, not a second mention count.

The board measures what is being published, which is not the same as what is being bought. A quiet quarter for a category does not mean agencies stopped buying it, only that the press wrote about something else.

Trends are matched to a fixed taxonomy, so a genuinely new phrasing can be undercounted until the taxonomy is extended. Where a story cannot be attributed to a partner, it counts toward the theme but not toward a brand.