They do have different tiers depending on your search volume and features, so in a way they already have this. I'd hate to have to go through checkout every time i did a search.
I haven't been using kagi long enough to really understand how it works yet, but it's my understanding that they want you to pay every month, even if you had remaining searches from the previous month.
If I pay $5 for 300 searches, why does it matter if I do them within a time frame? When someone isn't' searching, they aren't really costing Kagi anything.
Alternatively, let people pay 1.6 cents per search (or 1.8 cents or something).
Basically because the product they're selling isn't "You get to do a search whenever" but "You get to do a search this month".
The reason for that, based on my experience with various web startups, is they want to maximize the predictability of their resource usage in terms of staff and servers.
If millions of people pay their $5 and then don't use their searches, then in the extreme case Kagi could be maintaining servers twenty years later in anticipation that their customers might use those searches.
It's an edge case, but it illustrates the point.
Also, on the customer side, there's a psychological benefit to free things. Free as in "already paid for; no cost to using it".
If you have something that can be used this month but not any other month, then using it is free. If using it now means you can't use it next year, then there's still a cost to it despite it already being paid for.
Contrary to what MBAs want to believe, there’s no way to actually model a business in its entirety. Instead we have to use heuristics. And that predictability — minimizing the amount of potential energy in outstanding transactions (like searches paid for but not yet used) — is of value.
One can put a number on the value, but such numbers are always a little arbitrary.
Stepping back though, even if we want to make an attempt at a purely numerically model, the present value of selling a search can go way negative if one has to keep the server running indefinitely until a person uses the search.
If I sell you a search, and then you wait ten years to actually use the search, my side of the contract ends up including ten years’ worth of server costs as my server sits there waiting for you to use the search.
Again, it’s an edge case, but it illustrates the point. The cost of delivering on my sold goods goes up the longer you take to actually collect those goods. The physical world analogy would be if I sold you a car but then you took ten years to take possession of the car: now I’m paying for a garage to store that car for ten years.
Generally speaking, that’s the monetary reason to put a cap on the time involved in delivering on a sale.
Another angle to look at it is that money tends to devalue over time, and so goods (relative to money) tend to inflate in value over time. If I sell you a baseball for $10 and then you take a hundred years to collect on my baseball promise, maybe by then a baseball costs $100 and I’ve “lost $90”.
In summary, there are numerical models that show the costs of unbounded delivery times, but even without that there is a value to the predictability boon of keeping transactions time bound.