You will decide this on cash flow. Nearly everybody does.
The retainer wins, because a number that arrives on the first of every month is the most calming thing that can happen to a service business. That is a real consideration and I’m not going to pretend otherwise.
It is just not a consideration about the work. Which is why, eighteen months later, the model is fighting the work and nobody can quite say when that started.
So here is the thing to decide on instead. A retainer sells reserved capacity, a claim on your calendar. A project sells a finished thing with an edge on it. Pick based on whether the work has a finish line your client will recognize, and if it doesn’t, you are selling capacity and you had better say that out loud.
The two things being sold, and why it matters
Almost all the pain in both models comes from selling one thing and being measured on the other.
The client who asks “so what did we actually get for $6,000 this month” is not being difficult. They bought output in their head. You sold availability. Both of you signed the same document believing different things about what it was, and neither of you lied.
Projects have the cleaner version of this problem, because a deliverable either exists or it doesn’t. Which is also their weakness: you get judged on the artifact rather than the judgment behind it, and the artifact is the cheap part.
So the model you pick decides what you spend your energy proving. Retainers push you toward demonstrating activity, because activity is the only visible evidence that capacity got used. Projects push you toward shipping the thing on the date, whether or not it is still the right thing.
Neither is a flaw you can fix with better communication. It’s the shape of the deal.
The clever answer had a bad year
Somebody will tell you the real answer is to charge on value instead. Price the outcome, stop selling hours. The advice has been everywhere for a decade and it sounds like the grown-up position.
The one measurement I can find of what actually happened to the firms doing it does not support it, and it is recent enough to matter.
Promethean Research’s 2026 State of Digital Services report found that value-based pricing fell from 31% of agencies in 2024 to 18% in 2025, and that the agencies still using it grew more slowly than those on any other pricing model. The time and materials, fixed bid and retainer mix stayed dominant and grew faster.
Two caveats before anybody quotes that at a conference. The author says plainly it comes from a single survey wave and he isn’t claiming it settles the question, and I’d hold it the same way. Adoption falling is also not the same as the model being wrong. It can mean the people who tried it were not the people it was built for.
The explanation he offers is the part I find persuasive, though. AI compressed the execution layer faster than agencies built the strategy layer above it, so a lot of firms were pricing strategy work while still delivering through execution muscle, and clients read that as replaceable. You cannot charge for an outcome you only influence.
Which leaves you with the ordinary question, honestly answered, instead of an escape from it.
The renewal nobody tells you about
Here is the asymmetry that decides how a retainer feels to run.
A project is re-sold once, out loud, at the end, and you know you are in that conversation because somebody schedules it.
A retainer is re-sold silently, every month, inside somebody’s head. You are never told you are in the renewal conversation. The first time you hear about it is the cancellation, which is how losing a retainer always seems to arrive without warning and never actually does.
So a retainer is a relationship product on a monthly cadence, and the work that keeps it is frequently not the work in the scope. It is being reachable, having the context, saying the useful thing in a meeting you weren’t strictly paid to attend. That’s not cynical, it is what reserved capacity is worth to somebody.
The structural danger sits on top of it: the person who signed is also the person being asked internally to justify you, and when they change jobs the justification leaves with them. I went through the four dates that decide whether an agency keeps an account and will not repeat it here, except to say the retainer model is the one that carries that risk.
The ratchet, which is how a retainer actually dies
Retainers don’t die of price. They die of scope, and scope only ever moves one way.
Every month adds something. A small request while you’re in there. A report somebody wants kept up. A call that became recurring. Each one is reasonable on its own, and nothing ever comes off the list, because removing something takes a conversation and adding something takes an email.
By month fourteen you are doing eleven things for the price of four. The resentment shows up a long time before the conversation does, and what the client eventually hears is a complaint about money, which is not what it is.
Projects don’t ratchet, for the dull reason that they stop.
So if you run retainers you need a subtraction mechanism, and almost nobody has one. Not a scope document, those exist and get ignored. A standing conversation every quarter whose stated purpose is to take something off the list. The point of naming it is that the client can say no and you will have asked, which beats both of you discovering the answer in year two.
What the project model costs you
In fairness to retainers, projects have a real and underrated problem.
A project ends roughly when you have finally learned the business. Six weeks in you know who actually decides things, which of their numbers to trust, what the last three failures were. Then you hand all of that back and go and learn it somewhere else.
Which means the second and third projects with the same client are the profitable ones, and the first one is tuition you paid for. That is an argument for selling a sequence of projects rather than for converting everybody to a retainer, and those are not the same move.
The other cost is quieter. A project has an end date, so the last third of it is partly spent selling the next one. Clients can feel that, and it makes the close of good work faintly salesy in a way that nobody enjoys.
The shape that holds, and what I do
A defined project first, priced to be profitable on its own, followed by a named smaller retainer with a stated capacity.
Stated capacity, not “ongoing support”. Two days a month, what that covers, what it doesn’t, reviewed on a date. The honest sentence is “this reserves two days a month of my time”, and a client who decides that isn’t worth it should get to find that out at the start rather than in month nine. The exclusion line does most of the work here and I’ve argued that case already in how to write a proposal that wins.
The reason the sequence matters more than the structure: the project tells both of you whether this works before either of you commits to a year of it. That is worth more than the predictability you gave up.
For what it’s worth, that is the shape I use, and the part I got wrong for years was pricing the project low to win the retainer behind it. It makes the project a loss you then need the retainer to repay, which means you need it rather than deserve it, and clients can tell the difference.
If you’re setting terms and want to talk through which shape your work actually wants, drop me a line on email, WhatsApp, phone or LinkedIn and we can have a quick chat, and my notes on how I work cover the rest. I’m taking on a small number of contracts at the moment, so I’ll tell you straight whether it’s something I could help with.