Google Trends through HasData
One tool, hasdata_google_trends_search_getTrendsData. It takes q and returns one block of data, and which block you get is decided by dataType.
dataType renames the answer
timeseries returns interestOverTime, geoMap returns geoMap, relatedTopics returns relatedTopics and relatedQueries returns relatedQueries. The top-level key changes with the parameter, so reading a fixed key means three of the four calls look empty. One call gives one block, and a question that needs the curve and the regions needs two calls.
Interest is relative, never absolute
Google Trends reports interest on a 0 to 100 scale inside the requested window and set of terms. The numbers are not searches, and they are not comparable across two separate calls, because each call rescales to its own maximum.
Comparing terms means putting them in one request. q takes them comma-separated up to five, and a sixth is rejected with a 400. Each point then carries a values array with one entry per term, each holding query, extractedValue and a hasData flag, which is how a term with nothing behind it is distinguished from a term sitting at zero.
That comparison cannot be carried into the other blocks. relatedTopics and relatedQueries take a single term and answer a multi-term q with a 422 reading supportsSingleQuery.
A point also carries isPartial. The final bucket of a running window is usually partial, so treating it as a finished period turns an incomplete count into a fall.
Rising is labelled, not measured
In relatedQueries, a rising entry reads either as a percentage such as +300% or as Breakout. A Breakout still carries an extractedValue, and it is a sentinel far above any real percentage rather than a measurement. Sorting the rising list on that number puts the sentinels on top as though they were the biggest gains, so sort on the label and treat Breakout as its own category.
Targeting
geo is an enum of documented locations and a value outside it fails with a 422 naming geo, so a made-up code is rejected rather than ignored. region chooses the granularity of a geoMap between country, region, dma and city, and its default follows geo, so pass it explicitly or report which granularity you actually got. cat narrows to a category, gprop switches the corpus to images, news, froogle or youtube, and date takes Google's own range strings such as now 1-H or today 12-m.
tz is an offset in minutes and it defaults to 420, which is UTC-7. A series read without setting it is on Pacific time, and a daily boundary can land on the wrong day.
A quiet term is not an error
A term with too little volume returns a successful call with the series empty or flat at zero, and requestMetadata.status still reads ok. Report that the term has no measurable interest rather than presenting an empty chart as a failure.
Each successful call spends credits, and each dataType is its own call, so decide which blocks the answer actually needs before running all four.
