Agency Operations Checklist: Systems, Roles, and Delivery Basics

If you run a marketing agency and you want an agency operations checklist, the useful version is not a list of everything an operation could contain. It is the short list of things that should already be true, written plainly enough that you can tick or fail each one in about ten seconds. That is what this is.

I have spent twelve years running delivery operations for marketing agencies across 79 completed engagements with a 4.9 out of 5 client rating, which means I sell the fix and you should read the list with that in mind. Every item earns its place with one sentence on why it is there, because a checklist without reasons is only a record of somebody’s preferences.

How should you use this checklist?

Answer each line as your agency runs this week, not as it runs on a good week. There is no file at the end of this and no form to fill in. The article is the checklist, and if you want a version your team can use, copy the lines that apply into wherever your processes already live.

I have not attached a score to it. A number out of the total would tell you less than the answer to the better question, which is the one at the end: of the boxes you cannot tick, which ones are expensive at your size.

What should already be true of your delivery workflow?

Start here. If this layer fails, nothing built on top of it holds.

• Every project has one named owner. Shared ownership reads as collaborative and behaves as nobody, and you notice it the first time a project stalls without anyone feeling responsible for the stall.
• Every stage has a written condition for leaving it. Without one, work moves forward on assumption, and assumption is the mechanism underneath most missed deadlines I have been called in to explain.
• Nothing becomes work without passing through intake. Requests that arrive as a direct message skip scoping and scheduling, and those are the ones that break the week they land in.
• A scope change is recorded where the work lives. If the only trace of the extra round is in an email thread, it will not reach the plan, the schedule or the invoice.
• Someone other than the maker checks the work before a client sees it. The check does not have to be the founder, and the moment it can only be the founder you have found your bottleneck.
• You can see what each person is carrying for the next fortnight. An agency runs many projects against one shared pool of people, and a per-project plan cannot show you an overload that has not happened yet.

Who should own what?

Ownership is where checklists usually get vague, so these are written as names against jobs.

• Each recurring responsibility has exactly one name against it. Intake, scheduling, quality, client communication and measurement happen whether or not they are assigned, and unassigned they land on whoever is most conscientious.
• The definition of done is written and belongs to someone other than the owner. While the founder’s review is the only standard, the standard is a person rather than a document, and a person cannot be delegated.
• Coordination appears in someone’s job description. If nobody is paid to coordinate, your strategists and designers do it at specialist rates, and you lose the work you actually hired them for.
• There is a written rule for what your team can decide without you. Three short lists cover it: decide alone, decide and tell me, bring it to me. Most owners find the third list shorter than their calendar suggests.
• Somebody owns the system itself and has protected time for it. Running delivery this week and improving how delivery runs are two jobs, and when one person holds both, live client work wins that argument every time.
• Every named owner has named cover. An owner on holiday is not an owner unless a specific person has been told, in advance, that they are standing in.

What has to be written down?

The documentation layer is the one that decides whether a process survives a resignation.

• One source of truth for status, and anything not in it does not exist. Two half-maintained systems are worse than one imperfect system, because nobody can tell which of them to believe.
• Your statuses mean the same thing to everyone. In review, blocked and done are the three that quietly mean different things on different teams, and reporting built on them is fiction until they are defined.
• The processes that repeat are written: onboarding, reporting and QA. Those three cost the most when done differently each time, and trying to document everything else is how documentation projects die.
• Every pair of roles that hands work over has a documented handoff format. The detail that goes missing in an agency almost always lives in a gap between two people who each assumed the other was holding it.
• Every document carries an owner and the date it was last reviewed. Undated documentation is ignored within a year, because no one can tell whether it describes the current process or the one before it.
• Someone competent but new could run a normal week from what exists. This is the test the other five items are working towards, and it is the only one worth running as a live exercise.

What should your reporting be able to answer?

These are the questions the operation should be able to answer without a person assembling the answer.

• Status is a view someone opens, not a question someone asks. If a weekly update requires interrupting three people, you are paying twice for information the system already had.
• On-time is measured against the date the client was given. Not the date the task was later moved to, because the gap between those two dates is the only honest version of the number.
• Rework hours are counted separately from production hours. Work done twice is the cleanest evidence that intake or scoping failed upstream, and it stays invisible for as long as it sits inside one total.
• Client-visible slips are counted separately from internal ones. A missed internal date costs hours and a missed client date costs trust, so they should never share a number.
• A named person reviews these weekly. A roughly measured number looked at every week beats a precise one calculated every quarter, because only one of them can still change a decision.

I will not publish benchmark figures for any of these, because a number I invented for a healthy utilisation rate would get repeated in somebody’s leadership meeting as though it were data. The longer version of what I watch and why is in my agency operations guide.

What does tooling hygiene look like?

I work inside ClickUp, Asana, Monday.com, Airtable and Notion, and automate across them with n8n. I have no reseller relationship with any of them, and none of these lines is about which platform you chose.

• The tool mirrors how work moves, not your org chart. Structures built from reporting lines force people to route around them, and the routing around is where your real process ends up.
• Every field is either used or deleted. Fields nobody fills in teach the team that the system is optional, and that lesson spreads to the fields you needed.
• No part of live delivery lives in a private spreadsheet or a direct message. Workarounds are the most honest signal in an agency, and each one marks a place the system failed to fit the work.
• Every automation has an owner and a known failure behaviour. The question I ask before shipping any workflow is how you will find out when it stops running, because a silent automation is worse than no automation.
• You can say what each subscription is for. When two tools cover the same job, your team chooses between them weekly, and half your history ends up in the one that loses.

What if you cannot tick most of these?

Then you are normal. In twelve years I have not seen an agency tick every one of these, including agencies delivering good work at a healthy margin. Completeness is not the goal, and agencies that chase it tend to produce documentation nobody opens. The goal is knowing which of your gaps is expensive right now.

Size decides that more than anything else. Below about five people the founder is the system, and most of the documentation and measurement lines are theatre; the workflow lines still earn their place because they are what you will hand over later. Between roughly five and fifteen, ownership stops being free, since coordination has become a real job that nobody has been given. Past fifteen, the systems, reporting and tooling lines are the expensive ones, because each of them compounds with headcount rather than getting easier.

There is a second question underneath this one, which is whether the boxes you cannot tick are already costing you. Those two are not the same, and the symptoms tend to arrive before the diagnosis, which is why I wrote them up separately in the signs your agency needs operations help.

One more honest case. If your answer to the documentation section is that there is no settled process to write down yet, that is not a failed checklist, it is a different piece of work. Defining how delivery actually runs before anything gets documented or automated is business systems consulting, and it belongs before every tooling line above.

Where to start

Pick the single unticked box that would change the most weeks, and close that one before looking at the rest. If you cannot tell which it is, that is the question the Agency Ops Audit exists to answer: $1,500 fixed, two weeks, a written 90-day roadmap and a 60-minute readout call telling you which gap is costing you most and in what order to close the others. If your answers here were mostly yes, I would rather say so in the readout than sell you a project you do not need.