crosstalkwire.com tracks which phrases are crossing between industries in the
trade press. A private deployment does the same thing, pointed at the
publications, competitors and vocabulary that matter to one organization.
The public site reads trade publications across fourteen sectors and reports which
phrases appear in more than one sector's coverage, which sector used each one
first, and how usage has moved. Everything is a count of articles actually read,
with the articles linked.
A private deployment is the same system with a different reading list:
suppliers' sectors, and your customers' — including non-English-language
markets where a feed exists.
all of it.
deployment uses whatever divisions describe your market.
Three questions it answers that are otherwise answered by reading everything:
What is arriving in our sector's language that was not there a month ago.
New vocabulary in your coverage, when it appeared, and where it came from.
Which of our constraints are becoming other people's. When a phrase your
industry has used for years starts appearing in three others, that is a change
in who is paying attention.
What a broad topic has actually become. "Data center" can be a property
story, a power story or a chip story. What sits beside it in coverage tells you
which — and when it changed.
It does not forecast. Three separate tests were run on whether language
spreading across sectors precedes price movement — on public attention data, on
link-sharing data, and on ten years of SEC filings against sector prices. All
three found nothing. The last used 868 events across 42 quarters and returned a
breadth premium of −1.1 points at p = 0.62.
That result is stated here for the same reason it is stated in the interface: a
monitoring tool that describes what is being discussed is useful, and a
monitoring tool that quietly implies it can predict is not.
It does not score sentiment. It counts phrases. Whether coverage is positive
or negative is a judgment the tool does not make.
It does not assess accuracy. It reports what publications wrote, not whether
they were right.
Scoping. A conversation about your sectors, the publications that cover
them, and the companies and bodies worth tracking. Usually where most of the
value is decided — the reading list is the product.
Setup. Feed list assembled and verified, sector mapping defined, company and
institution lists built, first collection run. History accumulates from the
start date, so the tool becomes more useful with time rather than less.
Running. Collection every six hours. A weekly digest by email. A dashboard
at a private address. Feeds break and get replaced; that is ongoing work.
Some of what an organization reads is behind a license: a paid newswire, a subscription trade service, an internal feed. Where you hold a license that permits it and can supply a feed URL or API key, those sources can be read the same way public ones are — the system does not care where a feed comes from.
Two limits worth stating plainly. Whether your license permits this is a question for your agreement, not for us. And no article text is stored either way, only headlines, summaries, links and counts.
History starts when collection starts. A phrase that has been in your
sector's coverage for years reads as new on day one. This resolves over about
six months and is flagged in the interface until it does.
**A publication can only carry a sector label if its whole beat sits in that
sector.** General business titles are excluded, because their articles about
other industries would be counted as yours and would manufacture cross-sector
spread that never happened.
Coverage is only as good as the feeds. Where a trade publication has no RSS
feed, it cannot be read. Some sectors are far better served than others.
Article text is not stored — headlines, summaries, links and counts only.
Cheaper, and clear of redistribution questions.