The problem fixed pricing cannot solve
A fixed price is the right instrument when both sides can describe the thing being bought. You want missed calls answered and booked into the calendar; we can name the build, price it, and carry the risk of having estimated it badly. That is a clean trade and it is how most of our work is sold.
It stops being clean the moment the work stops being one thing. A client who has automated their intake usually finds three more jobs worth automating within a month, and none of them were in the original scope. Every one of those becomes a small quote, a small delay, and a small conversation about whether it was covered. Multiply that by a year and the dominant cost is not engineering. It is the friction of re-contracting for work everybody already knew was coming.
What a retainer actually buys
A retainer buys a block of our week, reserved for you, whether or not you fill it. That last clause is the whole product. We are not selling discounted hours; we are selling the guarantee that the hours exist when you want them, which means we cannot sell that same capacity to anyone else. The discount is compensation for that constraint, which is why it scales with the length of your commitment rather than the number of hours you buy.
The bands are published on the pricing page and generated from the same code that produces every quote, so the number you read is the number you would be charged. Four hours a week keeps what is running healthy and absorbs the small requests. Eight gets you one meaningful improvement a fortnight on top of that. Sixteen is a standing half-team working a backlog you control. Thirty is us functioning as your automation department, and at that point you should be asking whether hiring is cheaper, because sometimes it is.
The arithmetic that decides it
Take a real shape: a build we would quote at $15,000, against the eight-hour band at roughly $5,400 a month. The retainer passes the build's total cost after about three months. If the work genuinely stops at launch, the fixed price is cheaper and you should take it. The retainer only wins when there is a fourth month, and a fifth, with real work in them.
This is stated on the quote itself, in the same card as the retainer bands, because the alternative is letting a customer talk themselves into recurring revenue we did not earn. We would rather sell one honest build and be called back than book a subscription that quietly stops delivering value in month four and gets cancelled in month seven with bad feeling attached.
The terms, and why each one exists
Hours are reserved, not consumed. They are booked to your week regardless of whether you fill them. If that sounds unfavourable, compare it against the alternative: capacity you can call on only when we happen to be free is not capacity, it is luck.
Rollover is capped at one week. This is the term clients push back on most and the one we will not move much on. Unlimited rollover sounds generous and behaves like a liability: a client who under-uses for five months arrives in month six with a hundred and sixty banked hours and a deadline, and nobody can serve that without cancelling someone else's reserved week. One week of carry absorbs an ordinary quiet patch. A quarter of carry is a promise we would eventually break.
Overflow bills at list. Work past the allocation is charged at the full hourly rate. If we discounted it, the reservation would be worth nothing and the retainer would just be a worse way to buy hourly work.
You see the ledger. Every hour is logged against a named task and visible to you. A retainer without a visible ledger is an invitation to wonder what you bought, and that wondering ends the relationship more reliably than any invoice.
Platform costs sit outside it. Vendor usage is billed at cost, unmarked up, the same as on every other engagement. Bundling it into a retainer would hide a number you should be watching.
When not to buy one
If you have one defined job and no follow-on work, a fixed-price build is cheaper and you should take it. If your requests come in bursts three times a year, buy builds and pay list; you will spend less than reserving a week you use twice a quarter. If you are not sure yet, start with the build. Retainers are easy to start and mildly annoying to unwind, so the order of operations should favour the reversible option.
The honest summary: a retainer is a good deal for clients whose automation work never finishes, and a slightly expensive one for everybody else. Most businesses are in the second group when they first ask, and in the first group about six months later.
How to find out which you are
Describe the work on the plain-English page. You get a scoped build price and the retainer bands side by side, with the crossover month calculated against your own scope rather than against an example. If the answer is that the fixed price wins, that is what it will say.