The trouble with API keys that never expire
A key issued years ago, never rotated, and still valid today is not a convenience. It is a liability nobody has actually looked at in years.
An overage fee is a strange thing to build a pricing model around, because it only becomes visible after the fact. You pick a plan sized for your expected usage, and if a launch goes well or a marketing push works better than planned, you find out the real price of that success on the next invoice, at a rate that was never the headline number on the pricing page.
Flat, published pricing avoids this by having exactly one number for usage beyond the free allowance, not a tiered schedule where the price per request changes depending on how much you have already used that month. My Geocode charges €0.0001 per request beyond the 2,500 free requests a day, or a flat €50 a month for an Unlimited key, and that rate does not step up if you have a good month. There is no separate, higher overage rate waiting past a soft limit you did not know you were approaching.
The appeal of overage pricing, from a provider's side, is obvious: the headline number on the pricing page looks smaller than the price most customers actually end up paying, because the plan is sized to look attractive rather than to match typical usage. Customers who stay under the plan's cap get the advertised price. Customers who grow, which is to say the customers actually succeeding with the product, get quietly moved onto a worse rate that was disclosed but not emphasized.
Flat pricing removes that gap between the advertised number and the real one. Whatever rate is on the pricing page is the rate you pay at any volume beyond the free tier, whether that is your first paid request of the month or your millionth. There is nothing to discover later, because there is no second tier hiding behind the first.
This also makes budgeting for usage a genuinely simple exercise instead of a modeling problem. With a flat rate, expected monthly cost beyond the free allowance is just expected request volume times €0.0001, or a flat €50 if that number gets large enough that Unlimited is the better deal. With overage pricing, the same estimate requires knowing which tier a given month's traffic will land in and what the marginal rate is at that tier, information that is often harder to find than the headline price itself.
We understand why overage fees exist as a business model: they let a provider price a base plan aggressively while recovering margin from the customers who use the product the most. We think that trade is bad for the customer it targets, since the customers paying overage fees are, by definition, the ones the product is working best for. A flat rate treats a heavy user and a light user the same way per unit, which is the only version of pricing where growth is not quietly penalized.