operations
What an ecommerce agency retainer should actually cover
Most retainers quietly become hosting-and-hoping. What to buy instead, how to size it, and the three questions that tell you whether yours is working.
The build gets all the attention and most of the scrutiny. The retainer costs more over three years and almost nobody audits it.
A typical maintenance retainer is $1.5k–$8k a month. Over three years that is $54k–$288k — often more than the build it follows. It deserves at least the same questions.
The failure mode
Retainers decay in a specific and predictable way.
Month one is busy: snags, small fixes, things everyone agreed to do after launch. Month four is quieter. By month nine the invoice arrives, someone asks what it covered, and the honest answer is “monitoring, and we were here if you needed us”.
Nobody did anything wrong. There was no plan, so the hours went to whoever shouted, and in a quiet month nobody shouted. You have been paying for availability, which is the most expensive thing to buy and the least valuable thing to own.
What to buy instead
Named hours, with a use-it-or-lose-it rule you actually accept
Hours should be explicit: “20 development hours a month”. Whether they roll over is a real negotiating point — most agencies allow one month of carry-over, and unlimited roll-over is a red flag for both sides because it becomes a debt nobody can schedule.
A roadmap that exists before the month starts
The difference between a working retainer and a decaying one is whether there is a list. A short quarterly plan — three or four things, prioritised, with rough sizes — turns the monthly conversation from “what shall we do?” into “are we still doing this?”.
A standing block for the unglamorous work
Reserve part of the retainer for things nobody will ever request:
- Dependency and app updates, with a staging check before they touch production
- Checkout testing after any platform release
- Core Web Vitals and page-weight regression checks
- Reviewing 404s and broken internal links
- Backup and restore verification — actually restoring, not just confirming a backup exists
This is the work that prevents the emergencies you’d otherwise pay a premium to fix.
A defined response time for things that are on fire
“Checkout is broken” and “can we change this heading” are not the same request and should not have the same SLA. Agree three tiers, in writing, with hours attached:
| Severity | Example | Response |
|---|---|---|
| Critical | Checkout, payments, or the site is down | Same business day, named escalation path |
| High | A key template broken, tracking not firing | Next business day |
| Standard | Content changes, small features | Within the sprint or the month |
And agree what happens out of hours during peak trading, including what it costs.
How to size it
Work backwards from the roadmap, not from a feeling.
- A store doing under $50k a month rarely needs more than a small block — enough for updates, monitoring and one meaningful improvement a quarter.
- $50k–$250k a month justifies continuous work: a steady stream of conversion rate optimisation and merch improvements, and someone who knows the stack well enough to respond quickly.
- Above that, the question changes from “how big a retainer” to “what belongs in-house”. At some point a full-time developer is cheaper and more responsive than agency hours, and a good agency will tell you that before you work it out yourself.
The three questions
Ask these every quarter. They take five minutes and they are the whole audit.
- What shipped in the last three months? Not what was worked on — what is live. If the answer is hard to assemble, that is the answer.
- What did it do? Revenue, conversion rate, page speed, support tickets avoided. Not everything is measurable, but nothing being measurable is a pattern.
- What are the next three things, and why those? A retainer with no answer to this is being spent reactively, whatever the invoice says.
When to end one
Not when it gets expensive — when it stops being specific. The signs, in order of how much they should worry you: the monthly report becomes a list of hours rather than outcomes; the roadmap stops being updated; the same person stops being on the calls; you find yourself explaining your own stack back to them.
Ending well is easier if you read the contract first. What to check before you sign covers the handover clause that makes this a process rather than an argument.
