The paid Pinnacle API plans sell a rate, not a pile of requests: 10 requests a second, or 30, as the plan guide lists them, checked 2026-10-09. That choice of unit is the whole pricing model, and reading it right tells you both what your bill will be and what your architecture has to survive.
What per-second pricing actually sells
A per-second ceiling is a pipe, not a tank. Ten requests a second, held for a day, is 864,000 requests, and the plan guide does that arithmetic for you. But you are not buying the 864,000; you are buying the right to ask ten times in any given second. The day-count falls out of the rate you actually sustain, which is almost always lower than the pipe.
The free key and the SSE plan's REST side are the exceptions that prove the rule: they carry calendar caps, 20 a minute and 100 a day, because they are sized for evaluation and lookups, not for pipelines.
Why the model favors the buyer
A rate ceiling makes the bill flat and the capacity predictable. There is no meter to watch at the end of the month, no overage line, no volume surprise after a busy weekend. The pricing page lists no overage pricing at all, checked 2026-10-09; the ceiling is the plan, and the plan is the price.
The trade is that the ceiling is a design constraint. The pricing page says it directly: a published per-second limit is not permission to run permanently at it. Your pipeline has to treat the rate as a budget with headroom, not a target.
How to read your own needs in per-second terms
The sizing question becomes one sentence: how many times a second does my loop need to ask at its busiest moment. Count boards times phases divided by the interval, at peak, and compare against the rate with room for retries. A loop peaking at 3 a second fits the 10-a-second plan with comfort; one peaking at 12 needs the 30-a-second plan or a slower interval.
This is also why upgrading mid-surge does not fix a surge: the rate you bought is the rate you have, and the design work happens before the busy minute, not during it.
The question the model answers for you
Rate pricing answers the buyer's real question, which is not how much data can I get but how fresh can my state be. Freshness is an interval, the interval is a rate, and the rate is the plan. Once you see the chain, the price list reads as a freshness menu with a flat price on each tier.
Sell the rate, keep the bill flat, design under the ceiling. That is the whole model, and it is a kinder one than a meter running.