Agency operations is the set of stages, owners and checks that turns a sold client into work that is delivered, reported and renewed. In my own agency it broke in one place: a deliverable took three days to produce and eleven to reach the partner. Eight of those days were queue time between two senior reviewers, and no report we had flagged them.
How delivery works, and where it waits
Whatever stages your shop runs, for example intake, production, review and send, each deliverable lives on two clocks. The first counts hours someone actually worked on it. The second counts calendar days from entering a stage to leaving it. Time tracking records the first clock. Nobody logs hours for waiting, so the second clock can run for days without appearing in any report. That is how my eight days stayed hidden.
A wait-time log makes the second clock visible. It is a sheet with one row per deliverable per stage:
- Record the date the deliverable entered the stage and the date it left.
- Record the days someone actually worked on it there.
- Subtract: days waiting equals days in the stage minus days worked.
- Add up the waiting days by stage across ten or more deliverables.
Here is how it looked in my own team before the change.
| Stage | Days | Kind of time |
|---|---|---|
| Production | 3 | Work |
| Waiting in the senior review queue | 8 | Queue |
| Total from start to delivered | 11 | Waiting is 8 of the 11 days |
Review held 8 of the 11 days, and all 8 were waiting, so review was where to look. Then ask what kind of wait it is. If items sit in a line before the stage, the stage is short of capacity or is checking things that an earlier stage could check. If items go in and come back, the standard is unclear and the loop is rework. My review stage was a line: two senior reviewers in sequence, with every deliverable queued behind the last. SEO agency operations goes through the failure signals stage by stage.
Where mine broke
About two years ago a white-label partner asked my agency to take on five more accounts. Seven people were running 30 active accounts, and I could not say whether we could absorb five more, so we said no.
Five junior producers did the production, and it took three days. Every deliverable then went to two senior reviewers in sequence, and delivery took eleven. I added a third reviewer. Production got worse: the reviewers held inconsistent standards, and the inconsistency created rework. The standard for correct work lived inside two people's heads, and a third person brought a third version of it.
The full account is in my LinkedIn article and on the agency utilization consultant page.
What senior time was spent on
I had both senior reviewers log every change they requested for three weeks. Then I sorted each entry: a rule violation, which a written line could have caught, or a judgment call. About 70% of senior time went on rule compliance: template versions, data ranges, disclaimers, links, labeling consistency and brand specs. About 30% was judgment.
| Type | Example | What catches it |
|---|---|---|
| Rule violation | A report with the wrong date range selected. The report itself was sound, but the range made the results look bad. | The SOP and checklist caught it before the report went to the client |
| Judgment call | A senior decides how to explain a drop and what to recommend | No checklist can decide this; a senior does |
Mostly rule violations means the standards are undefined, so write them down before you hire anyone. Mostly judgment means your seniors are the limit and a senior hire is justified. A mixed log means rules first, because they shrink the pile you would be sizing the hire against. A senior salary spent on date ranges is the costly way to run a checklist. To run this split on your own numbers, use the Agency Review Bottleneck Test.
The fix: a spec, a gate and a log
Specification. For each deliverable type, write the required sections, data sources, naming convention and any partner-specific requirements. A line from a spec like this: "Date range: previous full calendar month, compared with the same month last year, shown on page one."
Gate. Work cannot enter review until it passes a mechanical checklist. Each line is pass or fail and asks for evidence. A line from a gate like this: "Paste the date range shown on page one next to the range in the spec." In my team the template and layout checks are automated, and everything else is a checklist and people. My SEO checklist includes noindex and canonical checks, because those failures are silent. The split between work to automate and work to keep with a person is covered in building the AI-augmented agency, and AI SEO automation goes into the rule-checking share of review time.
Exception log. When a reviewer asks for a change no spec covers, it goes in the log. When the same issue appears a few times, it becomes a permanent line in the spec or checklist. Review then holds more judgment and fewer rule checks as time passes.
| Before | After two years | |
|---|---|---|
| Turnaround | 11 days | 3-4 days |
| Active accounts | 30 | 37, with the same seven people |
| New hire productive | Six or more weeks | Under three weeks |
| Senior review volume | Baseline | Down by about two-thirds |
| Capacity ceiling | Could not be stated | Stated as a number of accounts |
The written side of this is in SEO agency SOPs and SEO delivery management.
What each number hides
Each number below hides something. The right-hand column is the number that exposes it.
| Number | What it sees | What it misses | Pair it with |
|---|---|---|---|
| Utilization | Share of paid hours logged to client work | Rework, waiting in a queue, who is doing the logging | Age of the oldest item in review |
| Turnaround | Days from start to send | Whether the work was right the first time | Share of items returned from review |
| Margin per account | Revenue against cost | Which stage consumed the hours | Hours by stage |
| Capacity in accounts | How many more accounts fit | Goes stale when a process changes | A recount after any change |
A utilization reading can look healthy while margin falls, because the hours land on rework and checking. The employee utilization rate for marketing agencies post explains the mechanism, and the post on a good utilization rate for a marketing or SEO agency gives rough ranges, which vary by agency.
I count capacity in accounts. To state a ceiling that way, find the stage that limits you, measure what one account costs there, and divide. After my fixes review stopped being the stage that limited us, which meant the old number no longer applied and I had to recount. SEO agency capacity planning works through the arithmetic.
How the operation shapes a quote
I do not quote work I cannot count capacity for. Four situations make me decline a quote or reshape it:
- The partner wants link building or PR in the scope.
- The partner wants guaranteed rankings or traffic.
- The partner has too few clients to start with.
- I cannot confirm I have the capacity.
New white-label partners start with five clients. Below that, setup per client is not cost-effective for either side, and a handful spreads the risk of one bad month.
Partners make four mistakes in month one: promising rankings by a date, giving late or incomplete access, leaving scope vague, and approving work slowly. Three of the four start at intake. Slow approvals start later, in review.
I also think a flat retainer punishes an agency for getting faster, because every hour saved is revenue left on the table unless pricing changes with it. The longer version is in every efficiency gain is costing you money, and why we don't publish pricing explains how I handle quotes.
Who owns what
Ask this of your own team: if your three most senior strategists left tomorrow, what percentage of your delivery would still ship on time? If the answer is under 70%, my staffing arrangement test says you have a staffing arrangement more than an agency. Each rule you write down should raise that number, since it is one less thing that leaves with a senior person.
In my setup the partner talks to me first, and each account has a named owner on my team behind me. Escalations go to the account owner first, then to me, and I answer within a day or two. My team checks the work before it goes out. I never contact a partner's end clients, because the partner owns that relationship.
SEO, AEO and GEO pages
For AEO and GEO I track a fixed list of questions on a schedule, with a professional AI-visibility tool, Search Console and analytics, and Google Analytics for referral traffic from language models. Results in all three move from month to month, and I tell partners so up front. I have delivered AEO and GEO on only a few client sites so far; the AEO agency service page describes how that delivery works. For what AEO and GEO are and how to measure them, see the guide to AI search optimization.
If you run an AEO or GEO agency, the utilization work has its own pages: AEO agency utilization consultant and GEO agency utilization consultant.
The table links each topic once per service line.
| Topic | SEO | AEO | GEO |
|---|---|---|---|
| Overall operations | SEO agency operations | AEO agency operations | GEO agency operations |
| Written procedures | SEO agency SOPs | AEO agency SOPs | GEO agency SOPs |
| Delivery management | SEO delivery management | AEO delivery management | GEO delivery management |
| Capacity | SEO agency capacity planning | AEO agency capacity planning | GEO agency capacity planning |
| Profitability | SEO agency profitability | AEO agency profitability | GEO agency profitability |
| Team structure | SEO team structure | AEO team structure | GEO team structure |
| Pricing | SEO agency pricing | AEO agency pricing | GEO agency pricing |
After the log: write the specs or hand off the work
After the log you can write the specs yourself or have my team deliver the work under your name. Hiring another reviewer before the log is the step I would not take.
For an outside read on your workflow, my agency utilization consultant work starts by mapping the workflow and then running the same three-week review log. For delivery, white-label fulfillment is done by my in-house team, not subcontractors, and I am your main contact; white-label SEO versus an in-house team covers the trade-off. For ongoing access to me on operations decisions, see fractional support.
FAQ
How do I know whether I need another reviewer or better standards?
Log every change your reviewers request for three weeks and sort each one as a rule violation or a judgment call. Mostly rules means write the standards first. Mostly judgment means a senior hire is justified.
How many accounts can my team handle?
There is no universal number. Find the stage that limits you, measure what one account costs there, divide, and recount after any process change.
Is running an AEO or GEO agency different from running an SEO agency?
Mostly no. The stages are identical. The difference is the proof: sampled question lists instead of rank checks.
When does white-label fulfillment make more sense than building this myself?
When the log shows undefined standards and you lack the time to write and enforce them, or when you cannot confirm capacity for work already sold. I ask new partners to start with five clients.
Keep reading
- All case studies
- Case study: 30 to 37 accounts with the same seven people
- Case study: how I got out of my agency's delivery queue
- Case study: two years of our exception log
- SEO agency SOPs
- SEO delivery management
- SEO agency capacity planning
- AEO agency operations
- GEO agency operations
- Employee utilization rate for marketing agencies
- Good utilization rate for a marketing or SEO agency
- Building the AI-augmented agency
- Every efficiency gain is costing you money
