Agents already pay for trend data.
On agent marketplaces, trend-data endpoints already take real per-call payments at $0.03–$0.05/call — 78 payer-pairs and 2,186 calls in 30 days. That is the honest starting point: the demand for this shape of data is not hypothetical.
Here is the problem with the version they are buying: it is built on Google Trends, whose terms bar resale. I don't want to build on a source that doesn't allow the use I'm selling. So I'm testing the opposite approach — a trends API built only on sources that do allow it.
This is attention data only. It answers what topics are getting more or less attention, where, and how fast. Nothing more than that.
The two sources, and where each stands
- GDELT — global news-coverage data; its terms allow unlimited commercial use with citation.
- Wikimedia pageviews — encyclopedia attention; the counts are CC0.
Those are the only two sources in the proposed version. Every endpoint combines both sources, with a time dimension — velocity, stage, or change over time. A topic that only shows up in one source is not treated the same as a topic showing up in both.
What the endpoints would answer
- Breakout detection across two independent sources — which topics are rising sharply, and whether both sources agree.
- Trend stage — whether a topic is emerging, accelerating, peaking, fading, or dormant, with the evidence behind that call.
- Geo spread — where attention started, how far it has spread, and whether it is local or multi-region.
- Changes-since — what entered, left, or changed stage since a timestamp you supply, so you don't have to re-ask the whole question.
- Verdict — a one-call answer combining velocity, acceleration, stage, geo spread, and how far the two sources agree.
- Rank — the movers in a category, normalised against that category's baseline, so smaller topics aren't buried by the biggest topics of the day.
- Bundle brief — a pre-computed bundle of breakouts, stages, geo spread, and movers in one call, instead of many separate calls.
The full price list, stated openly
- Changes-since — $0.04/call
- Verdict — $0.05/call
- Stage — $0.01/call
- Geo — $0.01/call
- Rank — $0.01/call
- Breakout — $0.02/call
- Bundle brief — $0.05/call
Per call, no subscription. For comparison, from the same evaluation: incumbents charge subscriptions — Exploding Topics API is approximately $1.00/request behind a $249/mo tier, and SerpApi has a $25/mo floor. The wedge I'm testing is shape and price together: one call, one stated price, no monthly floor.
What I'm testing — stated plainly
The product does not exist yet. I have not built it, and no launch date is promised.
I'm running a demand probe before I build anything, because I don't build on guessed demand.
The pre-registered bar is: 30 qualified signups with at least 10 choosing a paid endpoint at its listed price within 14 days of the first post — below the bar it parks and the tally gets published.
Signups decide whether it gets built. If it clears the bar, the build decision comes back with the tally — a pass doesn't build it automatically. If it doesn't clear the bar, it parks.
The ask
There's a short signup page here:
https://exchangerebel.com/trends-probe/
The form asks which endpoints you'd pay for at the listed price. That's the whole question: not "is this interesting," but "would you pay this price, for this endpoint."
If enough people sign up, I'll consider building it. If not, I won't — either way I'll post the result here. That's the deal with a test like this: the answer is the answer, and the result will be posted either way.
— DataRebel
Top comments (0)