There are 168 hours in a week. Your ad revenue is live for all of them.
A three-person ad ops team, working standard hours with slightly staggered starts, covers somewhere between 45 and 60 of those hours. The other 110 or so run unattended. Nothing about that is a criticism of the team. It is arithmetic.
The comparison worth running is not human versus machine. Nobody is replacing an ad ops manager with a model. The real comparison is between the team you have now and that same team with a monitoring and execution layer running underneath them. Same headcount, same salaries, same people making the judgment calls. The difference is what happens during the hours nobody is logged in.
This piece works through both sides of that: what the team costs, what it covers, and what the uncovered hours cost you.
What an ad ops team actually costs
Start with the number most studios and publishers already know, then add the part that usually gets left out of the mental math.
Salary data for ad operations roles varies widely by source and title, which is worth knowing before you benchmark against any single figure. ZipRecruiter puts the US average for an ad operations manager at $83,555, while Glassdoor puts the same title at $116,295 and Salary.com reports $129,968 for New York specifically. The gap comes from what each source counts as total pay and which company sizes dominate their sample.
Base salary is not the cost of the hire. BLS data for March 2026 shows wages and salaries make up 69.9% of employer compensation costs for private industry workers, with benefits accounting for the remaining 30.1%. That works out to a loaded multiplier of roughly 1.43 on base.
| Market | Typical base salary | Loaded cost per head | Three-person team, loaded |
|---|---|---|---|
| United States | $84,000 to $116,000 | $120,000 to $166,000 | $360,000 to $500,000 |
| India | ₹6.5L to ₹8.6L | ₹7.8L to ₹10.3L | ₹23L to ₹31L |
India figures come from Glassdoor’s ₹6,50,000 national average for ad operations specialists and ERI’s ₹8,60,916 for the same role in Pune, with a lighter statutory loading than the US multiplier.
Add tooling on top: reporting and BI seats, an analytics layer, whatever sits between GAM and the spreadsheet. For most teams this lands somewhere between $15,000 and $60,000 a year depending on how much of the stack is bought versus built.
What that team actually covers
Cost is the easy half. Coverage is where the interesting number hides.
Break the week into windows and mark honestly which ones have a person watching:
| Window | Hours per week | Realistically covered |
|---|---|---|
| Weekday business hours | 45 | Yes |
| Weekday evenings and overnight | 75 | No |
| Weekends | 48 | On-call at best |
That gives you roughly 27% of the week with active human attention, and that is the optimistic reading, because it assumes the team is watching performance during all 45 of those hours.
They are not. A survey of 170-plus advertisers found that three routine workflows consume 46.5 hours per strategist per month, about 26.8% of working time, before anyone gets to actual optimization. Separately, ad ops teams average 12 to 18 hours a week pulling reports across five to eight platforms. Detection competes for attention with trafficking, reconciliation, discrepancy chasing, and partner emails.
The industry knows this. A Digiday survey found 79% of ad ops professionals consider their current tools inadequate and 87% believe automation would make their companies more profitable.
So the practical coverage number is not 45 hours of monitoring. It is closer to a check every few hours during the workday and nothing outside it.
The cost hiding in the uncovered hours
Here is where the two halves meet.
When a tag decays or a demand partner pulls back, revenue does not go to zero. It degrades. The typical pattern across studios running GAM waterfalls is a drop that sits unnoticed for four to six hours before someone spots it and switches tags.
Run that as a formula:
Hourly ad revenue × hours of delay × degradation rate × incidents per week × 52
Using a five-hour delay, a 40% degradation during the window, and two off-hours incidents a week:
| Daily ad revenue | Revenue per hour | Loss per incident | Weekly | Annual |
|---|---|---|---|---|
| $5,000 | $208 | $417 | $833 | $43,300 |
| $20,000 | $833 | $1,667 | $3,333 | $173,300 |
| $50,000 | $2,083 | $4,167 | $8,333 | $433,300 |
Compare the right column against the team cost table above. At $5,000 a day the gap is an annoyance. At $20,000 a day it costs about the same as an extra senior hire. At $50,000 a day it approaches the fully loaded cost of the entire three-person team.
Adjust the assumptions to match your own incident frequency and the numbers move, but the shape holds: the loss scales with revenue while the team size stays flat. The bigger you get, the worse the uncovered hours hurt.
What the AI layer takes, and what stays human
The reason this works as an addition rather than a replacement is that the two kinds of work are genuinely different. One is pattern recognition against thresholds, running continuously. The other is judgment, negotiation, and deciding what the thresholds should be in the first place.
| Work | Better handled by the AI layer | Stays with the team |
|---|---|---|
| Continuous metric monitoring | Yes | |
| Anomaly and decay detection | Yes | |
| Floor price recalibration | Yes | |
| Tag switching on performance drops | Yes | |
| Off-hours incident response | Yes | |
| Setting thresholds and guardrails | Yes | |
| Demand partner relationships | Yes | |
| Direct deal terms and pricing strategy | Yes | |
| Ad format, placement, and frequency calls | Yes | |
| Discrepancy disputes and billing | Yes | |
| Quarterly stack and mediation decisions | Yes |
Notice that everything in the left column is execution against rules a human wrote, and everything in the right column requires knowing something the system cannot observe from GAM data. That division is the whole argument. More on how AI agents are being applied to ad operations work.
The secondary effect matters as much as the coverage one. When detection and tag switching come off the team’s plate, the 12 to 18 hours a week spent pulling reports to find problems turns into time spent on the right column. Publishers who automate these workflows report roughly 30% to 40% efficiency gains on the routine layer.
Side by side
| Team alone | Team plus AI layer | |
|---|---|---|
| Headcount | 3 | 3 |
| Annual people cost | Unchanged | Unchanged |
| Added platform cost | None | Fee or revenue share |
| Hours with active monitoring | 45 to 60 | 168 |
| Mean time to detection | Hours | Minutes |
| Time from drop to tag switch | 4 to 6 hours | Within the hour |
| Cost of adding apps or geos | More headcount | Roughly flat |
| Where team hours go | Detection and execution | Strategy and judgment |
The math changes by segment
Casual mobile game studios. This is where the coverage gap bites hardest, because the player base does not share a time zone with the ad ops team. Asia-Pacific holds around 1.48 billion mobile gamers, roughly 53% of the global player base. A studio operating out of London or Austin has its largest revenue block peaking while the office is dark. Add 44% of players reporting they play more on weekends and the highest-value hours land squarely in the uncovered column. For studios running multiple titles, the problem multiplies per app, which is the point where headcount stops being a viable answer. See also: how much mobile games make per ad and the top mobile gaming publishers by revenue.
App developers outside gaming. Ad ops is often half of one person’s job here, with monetization sitting behind subscriptions or commerce in priority order. Coverage is close to zero by design, and the loss stays invisible because nobody is looking for it. The threshold question is simpler: at what daily ad revenue does an uncovered gap cost more than the layer that closes it? For most apps that crossover sits between $3,000 and $7,000 a day. AdMob alternatives and mediation options are worth reviewing alongside this, since the mediation choice and the monitoring gap are separate problems.
Web publishers. More demand partners in the auction means more independent failure points, and header bidding setups add configuration surface where a wrapper or timeout issue can quietly cost fill for hours. Traffic peaks tie to content and news cycles rather than a predictable daily curve, so the incidents cluster at genuinely unpredictable times. Publishers running header bidding should also read the common header bidding mistakes that create these silent failures.
Run the numbers on your own setup
Before any vendor conversation, build the figure yourself. It takes about an hour.
- Pull 90 days of hourly revenue from GAM and export it.
- Flag every hour that came in more than 30% below the same hour on the prior three matching weekdays. Those are your incidents.
- For each flagged incident, find the timestamp of the corrective action in your change log. The delta is your real mean time to reaction.
- Multiply the shortfall by the duration for each incident and sum the year.
- Split the total between incidents that started inside working hours and those that started outside. The second number is what a monitoring layer addresses. The first number tells you whether you also have a process problem.
That last split is the one most teams skip, and it is the one that tells you whether you are buying coverage or buying a workaround for something you could fix internally.
FAQ
Does adding an AI layer mean we can cut ad ops headcount?
Not in any setup we would recommend. The layer handles detection and execution against rules your team defines, which means the team is still deciding thresholds, managing partner relationships, and making format and pricing calls. What changes is the mix of their week. Teams typically redirect the hours previously spent on report pulling and manual switching toward yield strategy and direct deals.
How is this different from what our mediation platform already does?
Mediation runs the auction. It decides which demand source wins each impression given the current configuration. It does not watch whether that configuration is still the right one, and it does not notice when a tag starts decaying at 2am. Switching mediation platforms does not close a monitoring gap, because the gap sits above the auction layer. See AppLovin MAX alternatives for how the two layers differ in practice.
What does the layer cost relative to a hire?
Pricing models vary between flat fee and revenue share, so the honest answer is that it depends on the vendor. The useful framing is the comparison in the table above: measure the annual cost of your uncovered hours first, then judge any quoted price against that number rather than against a salary.
Will automated tag switching break something overnight?
This is the right thing to worry about, and the answer is guardrails. Any layer worth running should let you set boundaries on what it can change, log every action in real time to a channel your team already watches, and let you reverse anything. If a vendor cannot show you the action log before you sign, that is the disqualifying answer.
How long does it take to see whether it works?
Ten to fourteen days on a single app or property is usually enough to see a directional signal, provided you baseline the same app for the same length of time beforehand. Run it on one property rather than the whole portfolio, keep the comparison on your own reporting, and look at revenue during your uncovered hours specifically rather than the blended daily number.
Find your coverage gap
The number that matters is not what your ad ops team costs. It is what your unwatched hours cost, and most teams have never calculated it.
Request a coverage gap audit. We will map your hourly revenue against your team’s actual monitoring windows and show you the annual figure sitting in between. No SDK, no app changes, and nothing to install to get the analysis.
