Agency Project Management Process: From Kickoff to Reporting
A stage list tells you what an agency project passes through. It does not tell you what has to be true before a project is allowed to leave one stage and enter the next, or what has to exist as proof that it actually did. That is a gate, and it is the one piece missing from every process piece I have written for this site so far. Below is the table I actually use: seven stages, kickoff to reporting, each with its entry criteria, its exit criteria, and the artifact that has to exist before the project is allowed to move. After the table, the two gates I watch get skipped most often, and what skipping them costs.
The seven-stage gate table
| Stage | Entry criteria | Exit criteria | Artifact produced |
|---|---|---|---|
| Kickoff | Signed proposal or agreed scope and price on record, nothing else | Named owner on each side, a definition of done stated in writing, first dated task assigned inside the project tool | The project record: owner, definition of done, first task, created live during the call |
| Scoping | Definition of done exists in the project record | Inclusions, exclusions and dependencies (revision rounds, feedback consolidation, sequencing) written next to the price | A scope note or SOW addendum stating what is explicitly not included |
| Scheduling | Scope and exclusions are written down | Work has a place against a named person’s actual capacity for the week, not just a due date in a field | A dated, assigned task visible on the assignee’s own board |
| Production | Task assigned, capacity confirmed | Every handoff names the next owner; work is not marked done until someone else has accepted it | The deliverable sitting in a review-ready state with a named next owner |
| QA | Work marked review-ready, a named QA owner distinct from the producer, time scheduled for the review | Deliverable checked against a written standard, pass or fail recorded, any fail routed back to production | A completed QA checklist tied to the specific deliverable |
| Delivery | QA passed and recorded | Deliverable reaches the client in the stated format, on the stated date, with a stated ask for what is needed back and by when | A delivery log entry: what shipped, when, what was requested in return |
| Reporting | Delivery logged, reporting period closed | Report sent on the agreed cadence and format, matching what the internal delivery record actually shows | The client report plus a stated next step, sent on schedule regardless of how the period went |
What counts as actually passing a gate?
Meeting the exit criteria and producing the artifact, not the calendar saying the stage is over. A project can sit in production for two weeks with nothing handed off, still get marked complete because the due date arrived, and never have cleared the gate at all. The artifact is what makes that checkable later. If there is no scope note, no QA checklist, no delivery log entry, there is no proof the gate was passed, whatever the status field says. I have run this table across 92 completed engagements over twelve years, and the artifact column is the part that catches problems the status field hides.
The stages are the same regardless of which tool holds them. I have built this exact gate table inside ClickUp, Asana, Monday.com, Airtable, and Notion, using n8n to move an artifact from a completed stage into an update in the next one. The table does not care which tool you use. It cares whether the artifact exists.
Scheduling, production, and delivery fail too, just less often as a first cause. A missed handoff in production is usually a downstream symptom of an unscoped exclusion, not a production failure on its own. In my experience, closing the two gates above removes most of what surfaces as a problem in the other five.
The two gates that get skipped most often
Of the seven, two get crossed without meeting their exit criteria far more often than the rest.
Kickoff and reporting fail more visibly, which is why they already have full pieces of their own on this site. Scoping and QA fail quietly, inside the stages nobody is watching, and by the time the cost shows up it gets attributed to a different stage entirely.
The scoping exit gate
Agencies are reasonably good at estimating hours. They are far worse at writing down what those hours exclude. The scoping stage gets treated as finished once a number is agreed, and the project moves to scheduling with nothing on record about revision rounds, who consolidates client feedback, or what sequence dependencies exist between copy and design. Nothing forces this gate closed, because there is no separate step that checks for it. The estimate looks like the whole job.
What it costs: scope creep that neither side sees coming, because there was never a written boundary to cross. The client is not behaving badly, and the account team is not underperforming. Nobody wrote down where the project ended, and the disagreement over what was included surfaces mid-project, when it is most expensive to resolve and hardest to bill for.
Closing this gate does not require new software. It requires a written exclusions line in the same document as the price, agreed before the project reaches scheduling, so the boundary exists before anyone needs it.
The QA exit gate
QA is the gate most likely to be skipped on purpose, under a deadline, by people who know they are skipping it. If no time is scheduled for review and no owner is named who is not the person who did the work, QA is not a stage, it is a hope that someone senior happens to glance at the deliverable before it ships. On a normal week that hope mostly holds. On the week that matters, it does not.
What it costs: a client-visible error costs more than a schedule slip ever does, because it changes what the client believes about whether anyone is watching the work. A missed internal deadline is recoverable. A mistake the client finds first is a different kind of damage, and it lands on the one gate agencies treat as optional under pressure.
Closing this gate requires naming a person who is not the producer and putting review time on the calendar before the deadline gets set, not after. A gate with no named owner and no scheduled time is not a gate. It is an intention.
Where the rest of the process lives
This table is deliberately narrow. The story behind each of these seven stages, what actually breaks and why, is in the full agency delivery process. The kickoff gate specifically, what has to happen in that first call and the first week after it, is in the agency kickoff and first-week workflow. The reporting gate, cadence, who assembles what and what should never be automated, is in the client reporting workflow. Read this table first if you want to know which stage to check. Read one of those three next if you want to know what to actually do about it.
If you would rather have the gate discipline run for you than build it yourself, that coordination across intake, scheduling, and handoffs is what I describe on the page for project management delivered as a service.
Where to start
The gate that is failing first is rarely the one making the most noise. A missed deadline reads as a scheduling problem and is frequently a scoping problem that surfaced two stages downstream. That diagnosis, which gate is actually failing and in what order to fix them, is what the Agency Ops Audit exists to do. It is $1,500 fixed, runs two weeks, and ends in a written 90-day roadmap and a 60-minute readout naming the gate that breaks first, not the one that is loudest.
