Home / Recipes / Catch funding rounds, acquisitions and IPOs
Worked queryCatch funding rounds, acquisitions and IPOs
/stories in semantic mode. The parameters matter less
than the reasoning behind them, so that is most of what is below.
The request
GET /stories
?q=raises funding round acquires acquisition stake IPO listing
&mode=semantic
&from=<now - 48h>Why these parameters
Deal language is formulaic but varied — raises, secures, closes, lands, acquires, takes a stake. Semantic mode covers the variants; a keyword query catches whichever verb you thought of. A short window suits this because deal news is announced rather than developed.
The obvious approach, and why it loses
The obvious approach is a keyword query on raises or acquires, and it catches whichever verb you thought of. Deal language is formulaic without being fixed: a company raises, secures, closes, lands or banks a round, and an acquirer acquires, takes a stake, buys into or agrees to purchase. Each is standard, none is universal, and the distribution varies by market and by publication. Writing the full list out as an OR query is possible and gets stale, because the vocabulary drifts and nobody revisits the list. Semantic mode covers the variants as a class, which is the difference between maintaining a list and not having one.
When it returns too much, or too little
Too much here is usually secondary coverage rather than irrelevant coverage: commentary about a deal announced last week, or listicles of the month's rounds. A short window is the cheapest fix, since deal news is announced rather than developed, and 48 hours removes most of the follow-on. Too little generally means the sector vocabulary is missing — add the sector to the query rather than more deal verbs, because the verbs were never the constraint. If you are watching a specific company, pair this with a name query instead of relying on the deal terms to surface it; a small round at a private company is often one paragraph in one outlet.
How you would know it is working
Reconcile against a source that is not news. Take a month of output and compare it to a deal database or a set of filings for the same period, in one sector where you can get both. Two numbers fall out: what the news feed missed, which is mostly undisclosed and secondary transactions, and what it reported that never completed. The second is the one people forget to measure, and it is why a count built on announcement coverage runs high. Once you know both gaps for your sector, the feed becomes useful for timeliness with a stated error bar, which is a different and more defensible claim than treating it as a record.
When this is the wrong tool
Announcement coverage is the only coverage there is here, and it is generated by parties with an interest in how the deal is perceived. Sizes are frequently rounded upward, valuations are reported as claimed rather than verified, and deals that fall apart between announcement and close are almost never covered a second time. This produces a record that overstates both volume and completion. It is also structurally blind to anything undisclosed, which includes most secondary transactions and a good deal of private credit. Use it for timeliness on what has been announced, and a filings or deal database for anything you intend to count.
Running it for real
Response fields, failure modes, adaptations and scheduling cost: Catch funding rounds, acquisitions and IPOs, in the docs.