Agency Margin Math: What Execution Hours Actually Cost

An execution hour costs you the fully loaded hourly cost of the person doing it, multiplied by how many of those hours exist across your book. Most agencies know the first number and have never measured the second. The model is three lines: loaded cost per delivery hour, console-and-reconciliation hours per account per week, accounts per strategist. Multiply. In the illustrative model below, that lands at roughly $7,800 a month of loaded labor per strategist spent on work that does not require a strategist's judgment. Your inputs will be different. Run them. The shape is what matters, not our numbers.

The uncomfortable part is not the size of the number. It is that the number only becomes margin if you do the harder of two possible things with it. Cutting headcount converts freed hours into margin immediately and visibly. Raising accounts per strategist converts them into margin too, slower, and only if you have demand waiting to absorb the capacity. Most agencies tell clients and staff the second story while quietly modeling the first. The rest of this post is about which one is actually true for you.

What does an execution hour actually cost?

Start with the number most agencies get wrong: loaded cost per delivery hour.

The common mistake is dividing salary by 2,080 hours and calling it done. That understates the real figure twice over. It ignores loading, payroll taxes, benefits, equipment, software seats, the share of management and ops overhead a delivery person consumes. And it uses a denominator that assumes every paid hour is a delivery hour, which no agency has ever achieved.

Here is the model. Replace every figure with yours.

Say a mid-senior strategist is paid $95,000. Apply a loading multiplier of 1.35 for taxes, benefits, seats, and overhead, and the real annual cost is about $128,000. Now the denominator: 2,080 gross hours, minus PTO, holidays, internal meetings, training, pitch work, and admin. Call it 1,700 hours that actually land on client accounts. That is $128,000 divided by 1,700, or roughly $75 per delivery hour.

That $75 is the number to carry through everything below. If your loading multiplier is 1.2 and your delivery-hour denominator is 1,500, you get a different figure, use it. The point of the model is that it is a model.

What this is NOT: this is not your billable rate. If you bill that strategist's time at $175 and their loaded delivery cost is $75, the gap is your gross delivery margin, and it is the thing execution hours quietly consume.

How do you build the margin model for your own agency?

Four inputs, in order.

1. Loaded cost per delivery hour

Done above. In this model: $75.

2. Delivery hours per strategist per week

Not 40. Use the weekly equivalent of your delivery-hour denominator, 1,700 hours a year is roughly 40 delivery weeks at 40 hours, so 40 is defensible here only because the annual denominator already absorbed the PTO and internal time. Do not subtract twice.

3. Hours per account per week, split into two buckets

Say each account consumes 5 hours a week, split 3 hours of console execution and reconciliation against 2 hours of strategy, analysis, and client contact. That split is an assumption, not a finding. The section below on measuring your own execution-hours percentage is how you replace it with something you actually counted.

4. Accounts per strategist

Say 8. That is 40 hours of delivery capacity fully consumed at 5 hours per account.

Now multiply. Console execution is 3 hours x 8 accounts = 24 hours a week. At $75, that is $1,800 a week, or about $7,800 a month of loaded labor per strategist, sitting inside work that does not require the judgment you are paying for.

What happens to the model when execution automates?

The honest version: console hours drop substantially, they do not vanish, and nothing else in the model moves at all.

Under an approval-and-verdict workflow, the strategist stops doing the mechanical steps and starts doing three different ones: reviewing proposed changes against a stated hypothesis, approving or rejecting, and reading the verdict afterward. That is real work with real hours. Say it takes 1.2 hours per account per week instead of 3. The 2 strategy-and-client hours are untouched.

Two lines are doing all the work in that table. Per-account hours fall from 5.0 to 3.2, and 40 divided by 3.2 gives 12.5 accounts, call it 12 to leave slack. Capacity per strategist rises by half without adding a person.

The edge case that breaks this table: onboarding. A newly connected account consumes setup, guardrail configuration, and a heavier approval cadence in its first weeks, because the system is proposing against an account nobody has cleaned up yet. If your book turns over quickly, your steady-state per-account hours are not your average per-account hours, and modeling with the steady-state figure will overstate capacity. Model new accounts separately for their first cycle.

Should you cut headcount or raise accounts per strategist?

These are not the same decision dressed differently. They produce different margin, on different timelines, with different risk.

Cutting headcount converts freed hours into margin immediately. In the model, 1.8 freed hours per account x 8 accounts is 14.4 hours a week, 36% of a strategist's delivery capacity. Across five strategists that is 72 hours, or roughly 1.8 FTE. The saving is real and it lands next payroll cycle.

Raising accounts per strategist converts the same hours into revenue instead of cost reduction, but only if you have demand. Capacity you cannot fill is not margin. It is idle cost with better utilization on paper.

We will say the unfashionable part plainly: the headcount version is the one most automation pitches are really selling, and the capacity version is the one they put on the slide. If your pipeline cannot feed 50% more accounts per strategist within the next two quarters, the capacity story is aspiration, not math. Say which one you are actually doing, to your team especially, because they will work out the answer faster than you present it.

There is a third option that gets skipped: hold both headcount and account count flat and spend the freed hours on the accounts you already have. That produces no margin line in the first year. It produces retention, which shows up later and is harder to attribute. A legitimate choice, just not a margin argument, and it should not be presented as one.

What doesn't change when execution automates?

Three things, and they are the three that actually justify your rate.

The client relationship. Nobody has automated the quarterly conversation where a client tells you their category is changing, or the call where a campaign underperformed and someone has to own it.

Strategy, offer, and audience definition. What to test, what the positioning is, which segment is worth chasing, a system that optimizes within your strategy does not produce your strategy.

The approval judgment. This is the one that surprises leads. Approval is not a rubber stamp you can delegate to a junior to reclaim more hours. Deciding whether a proposed change should ship, reading the hypothesis, checking it against what you know about the client's seasonality, the promo calendar, the thing the client said on Tuesday, is the senior judgment call. The model above already budgets hours for it. If you try to shrink those hours to zero, you have not automated execution. You have removed oversight, which is a different decision with a different risk profile.

How do you measure your own execution-hours percentage instead of trusting an industry figure?

Do not borrow a ratio. Measure yours. It takes one billing cycle and it is the only input in the model that anyone can credibly argue with.

Use three buckets, not two

Most agencies split time into "strategy" and "execution" and immediately lose the interesting part. Use three: console execution (making changes in platform UIs and editors), reconciliation and reporting (exports, spreadsheets, making platform numbers agree with each other and with the CRM), and strategy and client work. Reconciliation is usually the bucket that surprises people, and it is the one that scales with platform count rather than account complexity.

Sample a full billing cycle, not a week

A one-week sample is nearly useless, because month-end reporting is concentrated. Sample any week that is not month-end and reconciliation looks trivial; sample month-end week alone and it looks like the whole job. Track a full cycle per account.

Tag at the account level

Agency-wide averages hide the distribution that matters. One enterprise account with six platforms and a weekly reporting deck may carry four times the console load of five small accounts combined. You want the per-account figures, because the automation decision is made per segment of your book, not across all of it.

Exclude onboarding months and pitch work

Both distort the ratio in opposite directions. Flag them separately rather than letting them average in.

At the end of the cycle you have your own split, maybe it is 60/40, maybe it is 35/65, maybe it varies so much by account tier that a single agency number is meaningless. All three are useful answers. Any of them beats a figure you read somewhere, including this one.

When does this math NOT favor automating execution?

Four situations. If you are in one, the model above will still produce a number, and the number will be misleading.

Small account counts. Configuration, guardrail setup, and approval-workflow habits carry a fixed cost that does not shrink with book size. At three or four accounts, freed hours are a rounding error against that fixed cost, and you are better off leaving the process alone until the book grows.

Low change volume. Some accounts are genuinely stable, evergreen budgets, a mature keyword set, a client who does not want experimentation. If an account's real console load is one hour a week because there is nothing to change, the model's 3-hour assumption is fiction for that account, and automating execution saves you almost nothing. The savings scale with change volume, not with spend.

A creative bottleneck. This is the big one, and it is the one agencies misdiagnose most often. If your accounts are limited by how fast you can produce and iterate creative, freeing console hours does not increase throughput. It moves the queue to the designer, who is already the constraint. You will feel busier and ship the same amount. Fix the constraint that is actually binding; a faster non-bottleneck step is not an improvement.

Time-and-materials billing with no pipeline. If you bill hourly and cannot resell the freed capacity, automating execution reduces your revenue. That is not a paradox, it is the honest consequence of a pricing model tied to inputs rather than outcomes. That question is its own post, and it is the one linked first below.

Frequently asked questions

Related reading

• How Agencies Should Reprice Retainers When Agents Do the Execution — where an execution layer fits if you are consolidating agency tooling