Work Out What Manual Agency Ops Cost You Before You Hire Anyone

The number is what your agency pays every month to move work between people and systems by hand. Not what the work is worth. What the moving costs.

It has three legs, and they get counted separately because they behave differently:

1. Labor. The hours your people spend on repeatable handoffs, priced at what an hour of them really costs you.
2. Rework. The things that come back wrong, and the hours spent putting them right.
3. Foregone billing. The billable capacity you could have sold if those hours were free, which is often zero and should be written as zero when it is.

Every free calculator on the internet does the first leg. Most of them do it wrong, because they ask for a salary and multiply by hours, and a salary is not what an hour costs. Almost none of them do the second or third leg at all, which is convenient for whoever built the calculator and useless for you, because the second leg is where the client relationship actually gets damaged and the third is where the argument for change either holds up or falls apart.

So we are going to do all three, in order, with the sourcing for every rate. Then run the whole thing on a real-shaped agency so you can see the arithmetic end to end rather than a widget spitting out a number you have no way to check.

Divide the wage by 0.699. That is the whole trick, and it is the single most common error in this calculation.

The Bureau of Labor Statistics Employer Costs for Employee Compensation release for March 2026 puts total employer compensation for private industry workers at $46.60 per hour worked. Wages and salaries are $32.60 of that, which is 69.9 percent. The other $14.01 is benefits, paid leave, insurance and legally required contributions. It is money you spend on the hour whether the hour goes into a client deliverable or into renaming CSV exports.

So take the wage, divide by 0.699, and use that. It is a multiplier of about 1.43, and the gap it opens is not small.

For grounded wage figures, O*NET OnLine republishes BLS Occupational Employment and Wage Statistics by occupation. Three that map onto agency staffing, at 2025 medians:

• Search marketing strategists: $37.87 an hour, $78,760 a year. Loaded, $54.18.
• Project management specialists: $49.19 an hour, $102,320 a year. Loaded, $70.37.
• Marketing managers: $80.19 an hour, $166,790 a year. Loaded, $114.72.

Use your own payroll if you have it, and use these if you do not or if you want a sanity check on what you are paying. Either way, load it before you multiply. Sixteen dollars an hour of invisible cost on a strategist, times thirteen hours a week, is $918 a month that never shows up anywhere you would look for it.

Count the handoff, not the task. A handoff is any moment a person moves something out of one system and into another by hand, or chases somebody so a piece of work can continue. Building the deck is the task. Exporting four dashboards, renaming the files, pasting into the sheet and refitting the formulas is the handoff, and the handoff is the part that repeats identically forever.

This distinction is the reason your gut number is too low. Ask anyone what the monthly client report takes and they will tell you about the analysis. Watch the calendar and the hours are on the arrows.

There is survey evidence for how big the gap runs. Asana's Anatomy of Work Global Index 2023, which surveyed 9,615 knowledge workers across six countries, found people spending 58 percent of the day on work coordination rather than the skilled work they were hired for, using an average of 8.8 apps to do it, and estimating they could save 4.9 hours a week with better processes.

Do not use 58 percent in your model. Self-reported time is soft in both directions, and a number that large will make your total look absurd to the one person you need to convince, which is often yourself. Use it as permission to go looking, then count for real.

How to count for real, in one week. Pick the five workflows that repeat per client or per project: the reporting pack, the weekly status roll-up, new client or new brief onboarding, invoicing and timesheet chasing, and asset delivery. For each, name the steps, then name the gaps between the steps. Put a stopwatch on the gaps once, or read them off calendar entries and Slack timestamps, and multiply by how often they happen. One week of honest counting gives you a weekly hours figure you can defend.

The answer to whether you should consolidate the stack or wire up what you already pay for falls out of this count too, because the workflows with the most arrows are the ones crossing the most tools.

Rework is errors times hours to fix times the loaded rate, and the reason it belongs in the model is that manual handoffs do not fail rarely. They fail at a rate that is well documented and that nobody believes about their own team.

Ray Panko at the University of Hawaii keeps the running summary of field audits of operational spreadsheets, the real ones businesses were actually running when the auditors arrived. Across nine audits covering 163 spreadsheets, 84 percent contained at least one error. Narrow it to the five studies that counted only serious errors and it is 91 percent of 55 spreadsheets. Individual audits range from 44 percent to 100 percent.

That is the artifact sitting in the middle of your reporting chain. It is not a story about careless people. Panko's point across thirty years is that humans are around 95 to 98 percent accurate entering data and far worse at spotting an error once it exists, so a process built on hands will produce a predictable error rate no matter who you hire.

So do not estimate this leg in your head. Count it: open the last two months of client Slack channels and retro notes and count the things that came back. A wrong figure in a report. An asset delivered in the wrong format. A duplicated line on an invoice. Then put an honest number of hours against fixing each one, and include the conversation, not just the correction. The apology email is part of the cost.

One caveat worth stating plainly. Automating a handoff does not take this leg to zero, it changes its shape, and anybody who tells you otherwise is selling. What actually breaks in production automations is a different failure set with a different frequency. Assume you recover most of this leg, not all of it.

Write zero here unless you turned work away. This is the leg that makes automation business cases look ridiculous, and it is the leg most calculators lean on hardest, because it is the biggest and the least checkable.

The logic is sound in principle. If a strategist spends nine hours a week on handoffs and you had demand for those nine hours at $145, the handoffs cost you $5,651 a month in revenue you never invoiced. The logic collapses the moment you did not have the demand, because then the alternative to the handoff was not billable work, it was slack.

There is a real gap to reason about. SPI Research's 2026 Professional Services Maturity Benchmark, summarized by Certinia from 509 professional services organizations, put billable utilization at 66.4 percent in 2025, the lowest point in the history of their survey. Average EBITDA sat at 9.9 percent. So the sector is carrying a meaningful block of non-billable time and thin margins to absorb it.

But a sector average does not tell you whether your pipeline could have eaten those hours. So use this rule and hold to it:

• You turned work away or quoted a start date further out than the client wanted, in the last quarter: book the share of hours you could actually have sold. Be specific. One month in three with real unmet demand is a third of them, not all of them.
• You did not: book zero, and say so in the model.

A calculation with an honest zero in it is more persuasive than one with a big unverifiable number, because the person you are showing it to can tell the difference immediately.

Four inputs, three lines, one total. Copy the table, fill the last column, add it up.

Monthly cost of manual ops
  = (handoff hrs/week  x  loaded rate  x  4.33)
  + (rework items/month  x  hrs to fix  x  loaded rate)
  + (billable handoff hrs/week  x  sellable share  x  billing rate  x  4.33)

The 4.33 is weeks per month, which matters more than it looks: multiplying weekly hours by 4 quietly loses you four weeks a year.

What you need before you start, and where each one comes from:

• Loaded rate. Your payroll wage divided by 0.699. If you do not have payroll to hand, take the O*NET median for the closest role and load that.
• Handoff hours per week. Counted over one week from calendars and timestamps, summed across the team, not estimated.
• Rework items per month and hours to fix. Counted backwards out of two months of client channels and retro notes.
• Sellable share and billing rate. Your own effective rate, and a share you can defend from the last quarter's pipeline. Zero is a legitimate answer.

If you have mixed roles doing the handoffs, blend the rate by hours rather than by headcount. Ten strategist hours at $54.18 and three project manager hours at $70.37 come to $752.91 a week, which is a blended $57.92 across thirteen hours. That blended figure is what goes into legs one and two.

It means the hire is probably the wrong move, and that is the whole reason to run this before writing the job ad.

Compare it against what it would take to make it go away. Start with the hire, since that is the instinct.

A search marketing strategist at the $78,760 median wage costs $112,675 a year fully loaded. Add recruiting on top: SHRM's 2026 recruiting benchmarking data, from 4,657 members surveyed between November 2025 and January 2026, puts median cost-per-hire for a nonexecutive role at $1,300 and median time-to-fill at 39 days. So the first year is roughly $114,000 and the relief starts in about six weeks, assuming you get the hire right.

That is $114,000 against a $70,484 problem. And the new person inherits the same handoffs, which means in eighteen months you are running this calculation again with a bigger team in it. Hiring to absorb repeatable handoffs converts a fixed cost you could have removed into a permanent one.

Now the other direction. Our own numbers, published on what we charge rather than sitting behind a form: a one-off starter build is $1,500 to $2,500 fixed, a two-week production sprint that puts a full workflow into production is $5,000 fixed, and retainers start at $5,000 a month for continuous build-and-run work. Against a conservative $3,990 a month, a $5,000 sprint that removes most of one workflow pays for itself inside two months. Against the same number, a $5,000 monthly retainer does not pay for itself on this one workflow and should not be sold to you as if it does. A retainer is for a queue of workflows, not a single one, and if your queue is one item long the honest answer is a fixed-price build and then goodbye.

And there is a third answer that nobody with something to sell will volunteer.

Below roughly $1,000 a month, leave it alone. At that level a fixed-price build has a payback measured in quarters, the workflow probably is not stable enough to be worth encoding, and you have better things to spend an afternoon on.

Three other conditions that should stop you regardless of the number:

• The process changes every quarter. Automating a workflow that is still moving means rebuilding it every time it moves. Let it settle first.
• The hours are spread thin across many people. Thirteen hours concentrated in two people is a system. Thirteen hours spread across eight people at ninety minutes each is a habit, and habits respond to a template or a checklist far more cheaply than to a build.
• One person is doing all of it and they are leaving. Then your problem is documentation and continuity, not automation, and you should fix that first.

When the number does clear the bar, the shape of the work and the shapes it comes in are covered properly by our guide to what an agency automation consultant does and when you need one, and the price bands for the work itself are in what it costs to automate agency operations.

Every one of these makes the number bigger, which is exactly why they are easy to leave in.

Double counting leg one and leg three. If you book a strategist hour as labor cost and then book the same hour again as foregone billing, you have counted it twice. They are different questions. Leg one asks what you paid. Leg three asks what you could have earned instead. Both are real, but a reader who spots the overlap will discount the whole model, so state plainly that the same hours appear in both and that you are treating them as separate consequences rather than adding a fictional third.

Assuming full recovery. No automation removes 100 percent of a handoff. There are exceptions, an approval step, a client who wants a phone call, the month the export format changes. Sixty to seventy-five percent recovery on a well-scoped workflow is a realistic planning assumption, and anything above that needs to be a demonstrated result rather than a projection.

Using the rate card instead of the realized rate. Your rate card says $195. What actually lands after scope creep, write-offs and the discount you gave to win the account is lower, and leg three should use the lower one.

Counting hours nobody was ever going to sell. The 4.9 hours a week that Asana's respondents believed better processes would save them are the hours they believed they would save. Belief is a good reason to go and count. It is not a line in a business case.

Five steps, and the first four take one afternoon each.

1. Pick the week and count. One ordinary week, no launches. List the five repeating workflows, name the arrows in each, and log the minutes.
2. Load the rate. Payroll wage divided by 0.699, blended by hours across whoever is doing the work.
3. Count the rework backwards. Two months of client channels and retro notes, one line per thing that came back.
4. Decide leg three honestly. Look at the last quarter's pipeline and write the share you could genuinely have sold. Write zero if that is the truth.
5. Rank by number, then by stability. Take the workflow with the biggest monthly figure that has not changed shape in six months. That is the one to fix first, and it is almost always client reporting, because it repeats per client per month and gets worse with exactly the thing you are trying to grow.

One more thing worth saying about what happens after step five. The gap between what a no-code tool can hold and what needs building properly is a real fork, and picking wrong is expensive in both directions. Where Zapier and Make stop working is the honest version of that decision, and cost is usually the wrong reason to move.

When we run a free audit, this is the count we do, and we do it with your calendars rather than a survey average. Sometimes the number comes back small and we say so and there is no build. That is a fine outcome for everyone except the consultant who needed the sale.