Home / Recipes / Watch for regulatory action in a sector
Worked queryWatch for regulatory action in a sector
/stories in hybrid mode. The parameters matter less
than the reasoning behind them, so that is most of what is below.
The request
GET /stories
?q=<sector> regulator investigation ruling fine consultation
&mode=hybrid
&from=<now - 14d>Why these parameters
Regulators publish in bursts around decisions and consultations, then go quiet, so a short rolling window returns either a flood or nothing depending on where it lands. Widen the window rather than loosening the query. Hybrid mode because regulatory language is formulaic enough for keyword matching to work on the terms of art, while the surrounding coverage describing the action is not.
The obvious approach, and why it loses
The obvious approach is to watch the regulator's own feed, and where one exists it is strictly better for completeness — this recipe does not replace it. What it adds is the part the feed cannot tell you: which actions attracted attention, how they were characterised, and what the affected firms said in response. The other common approach, a keyword query on the regulator's name, returns everything the regulator is mentioned in, most of which is background reference in stories about something else. Pairing the sector with the terms of art keeps the results on actions rather than on mentions.
When it returns too much, or too little
Too much usually means the sector term is broad enough to catch general business coverage; replace it with the specific instrument, licence type or statute that the actions concern. Too little is almost always the window, because regulators publish in bursts around decision dates and then go quiet for weeks — a fortnight is the working minimum and a month is often better. Resist loosening the query to fill a quiet window; an empty fortnight in regulatory coverage usually means nothing was decided, which is itself worth knowing. If you need a specific case followed, use the story identifier rather than re-running this query; get_story returns the whole case in one call.
How you would know it is working
Compare a quarter of output against the regulator's own published decisions for the same period. Two numbers come out: how many decisions this surfaced, which is your recall, and how many of the surfaced items were commentary rather than decisions, which is your precision. Recall will be well under half in most sectors and that is expected — coverage skews to enforcement against large or newsworthy firms — but knowing the figure prevents the feed from being read as a complete record. Re-run the comparison annually, because the skew moves with how much attention a sector is getting.
When this is the wrong tool
News coverage of regulatory activity is systematically incomplete. Regulators publish a great deal that no publication covers, and coverage skews heavily towards enforcement against large or newsworthy firms. If you need completeness, the regulator's own publications are the source and this recipe is a supplement that tells you which actions attracted attention. That is useful information, but it answers a different question. This is also the wrong tool for tracking a specific case through its stages: coverage clusters at announcement and resolution with nothing in between, so a case can appear dormant for months while proceeding normally.
Running it for real
Response fields, failure modes, adaptations and scheduling cost: Watch for regulatory action in a sector, in the docs.