Agency Capacity Planning: Spot Overload Before Work Slips

Most agencies discover they are over capacity the same way: a client emails asking where the audit is, the account manager checks with the specialist, and the specialist says they have been underwater for three weeks. Nobody lied. Nobody was lazy. The overload was simply invisible until it turned into a missed deadline.

Agency capacity planning is the discipline of comparing committed delivery hours against realistic available hours, several weeks ahead, so overload shows up as a number before it shows up as a slipped deliverable. It is not a resourcing tool purchase. It is a weekly habit backed by two honest numbers per person.

Why overload stays invisible in agencies

Agency work hides load better than almost any other type of work. Three reasons:

• Retainers have soft edges. A "10 hours/month SEO retainer" rarely means ten hours. It means ten hours plus a Slack thread, plus a call, plus the "quick question" that costs forty minutes.
• Task counts lie. One specialist has 14 open tasks and is fine. Another has 4 open tasks — a technical audit, a migration QA, a content model rebuild — and is drowning. Counting cards tells you nothing about hours.
• People absorb first, complain later. Good delivery people compress their own buffer before raising a hand. By the time someone says "I'm overloaded," they have already been overloaded for two to three weeks.

So capacity planning has to measure hours against a forecast, not vibes against a backlog.

Step 1: Calculate real capacity, not headcount capacity

Start with the only number that matters: deliverable hours per person per week. A full-time delivery role does not have 40. Subtract honestly:

• Internal meetings, standups, retros: 3 hours
• Admin, timesheets, tooling, Slack: 3 hours
• Sales support, pitch help, onboarding: 2 hours
• Learning, testing, algorithm-update firefighting: 2 hours

That leaves 30 hours of real delivery capacity — 75% utilization. For account managers and leads who carry client calls and internal management, use 20-22 hours. For a senior who also reviews everyone else's work, assume 4-6 hours per week goes to review alone and subtract it before you assign them production work.

A seven-person delivery team does not have 280 hours a week. It has roughly 190. If your plan assumes 240, you are already 25% over before a single client sends an urgent request.

The 15% volatility reserve

Agency work has a predictable amount of unpredictability: a Google core update, a client's site going down, a last-minute pitch. Track it for a month and you will usually land somewhere between 10% and 20% of total hours. Reserve 15% explicitly. If you plan to 100% of adjusted capacity, every surprise becomes an evening.

Step 2: Convert every commitment into hours

You cannot plan capacity against a client list. You plan it against hours. Build a simple commitments table with three categories:

1. Retainer baseline. What recurring work ships every month, in hours? Not what the contract says — what it actually takes. Pull three months of time tracking and use the median, not the best month.
2. Project phases. A site migration isn't 60 hours spread evenly across six weeks. It's 8, 12, 25, 20, 6, 4. Load the curve, not the average, or you will plan a crunch week as if it were a normal one.
3. Known ad-hoc. Every agency has a client who sends two "quick" requests a week. That's 3 hours. Put it in the plan with their name on it.

The gap between contracted retainer hours and actual delivered hours is where most agencies quietly lose margin. If a 10-hour retainer consistently consumes 14 hours, you are not over capacity — you are under-priced, and capacity planning is what finally makes that visible. Time tracking that ties hours to the specific client and phase is what makes that comparison possible; this is the part platforms like PeakKR are built around, but a disciplined spreadsheet gets you 80% of the way there.

Step 3: Build a six-week rolling view

Six weeks is the sweet spot. Shorter and you cannot react; longer and client scope changes make the forecast fiction.

Build a grid: people down the side, weeks across the top. In each cell, planned hours versus capacity hours. Then color it by load:

• Under 70% — available; this is where new work goes
• 70-90% — healthy
• 90-100% — tight; no new work, no surprises absorbed
• Over 100% — already slipping, whether or not anyone has said so

The forecast's value is not precision. It's the pattern. When you see a specialist at 118% in week 3 and 124% in week 4, you have a four-week warning instead of a four-day panic. The same logic that makes hill charts more honest than Gantt charts applies here: you want a view that shows uncertainty and trajectory, not a tidy bar that implies false confidence.

The five leading indicators of team overload

Deadlines slipping is a lagging indicator. These five show up weeks earlier:

1. Cycle time creeping up

Track median days from "in progress" to "done" for a standard task type — a monthly report, a content brief. If your content briefs normally close in 2 days and they are now taking 5, the person is context-switching, not slowing down. Cycle time is the single most reliable early signal.

2. Work in progress inflating

An overloaded person starts more than they finish. If someone's open "in progress" count goes from 3 to 9 while completions stay flat, they are juggling, and something will drop. Set a WIP ceiling of 3-4 active items per person and treat breaches as an alert.

3. Review and QA steps getting skipped

The first thing a squeezed team cuts is internal review — because it feels like the only step the client can't see. When a technical audit goes straight to the client without a second set of eyes, you are not saving two hours, you are buying a rework cycle plus a credibility hit.

4. Rescheduled-to-completed ratio rising

Count how many items had their due date pushed versus how many closed. A ratio above roughly 1:4 in any given week means the plan is already fiction and the team is managing the calendar instead of the work.

5. Senior people doing production

When your head of SEO is building keyword maps at 9pm, capacity has already failed upstream. That's the most expensive hour in the agency being spent on the cheapest work, and it means the review layer has evaporated.

What to do when the forecast goes red

Finding overload is the easy half. Here's the triage ladder, in order — always start at the top:

1. Cut scope inside the deliverable. Ship the audit covering the top 50 priority issues instead of all 180. Most clients prefer a focused deliverable on time.
2. Re-sequence, don't compress. Move a phase start by a week in the plan, and tell the client now. A proactive date change four weeks out costs nothing. The same change announced on the due date costs trust.
3. Rebalance across the team. Look for people under 70%. This only works if your capacity view is current — which is why the weekly review matters more than the model.
4. Bring in a contractor. Works for production (content, dev, link outreach). Rarely works for strategy or client-fac