← All posts

  • in-house
  • outsourcing
  • agency
  • ad-ops

Bringing ad ops in-house: what comes back, and what newly costs you

Switching ad ops from an agency to in-house removes the management fee, but the hours and the responsibility for judgment move to you. A table comparing fees against internal hours in yen at each level of spend, what happens to the account and the learning phase when you switch, the three things to take with you, and how to choose not to bring everything in-house, from where we sit building ad-ops tooling.

Bringing ad ops in-house: what comes back, and what newly costs you

Translated from the Japanese original on mureo.jp.

You are weighing whether to pull the ad ops you have handed to an agency back into your own company. Or you are already running it yourself and running out of hands. Whichever door you came in through, the two things you need in order to decide are the same. What comes back to you when you bring ad ops in-house. And what newly becomes work in its place.

Deciding on the saved management fee alone is usually not enough. The fee you save can simply turn into time inside the company, leaving the total barely changed. This article lines those two up in money and in work, then goes as far as what happens to the account and to automated bidding’s learning when you switch.

If you intend to stay outsourced and want cost benchmarks and how to pick a provider, Outsourcing search ad management should fit your situation better.

What bringing ad ops in-house gives you back

Speed of decision. The round trip between seeing the numbers and making a change disappears. Going through an agency, it can take days from noticing something to contacting them, waiting for a proposal, sending back approval, and having it applied. With permissions in your own hands you can fix it the same day. The gap tells most on one-off campaigns and on products tied to inventory.

Context from the business side. Which product you want to push this month, which enquiries are turning into orders, what sales is struggling with right now. None of this can be conveyed to an agency in full. Both sides spend time in meetings and reports trying. With the operator inside the company, that transfer stops being necessary at all.

The management fee. Around 20% of ad spend, or a minimum fee in the tens of thousands of yen, stops leaving every month. As covered below, though, this does not all turn into profit.

Accumulated knowledge. A record of which keywords worked and which angles missed stays inside the company. You escape the state of explaining everything from zero each time you change agencies.

The work that increases in its place

Behind what comes back sits work that transfers to you as it is.

Daily checks. Reviewing the search terms report and excluding terms that miss your intent, watching the pace at which budget is consumed, checking how bids are working. Each is short on its own, but it comes up daily or every other day.

Production and trafficking. Swapping ad copy, preparing banners, launching new campaigns. Each platform has its own conventions for uploading, and a review rejection means rewriting.

Keeping up with platform specs. Both Google Ads and Meta Ads change their consoles and their bidding mechanics several times a year. Stop following and you carry on running, at some point, outside the recommended settings. This is also the part where an agency’s value shows most plainly.

Responsibility for judgment. When things do not work, isolating the cause and deciding the next move becomes yours. Outsourced, you were the side asking for an explanation; in-house, you become the side giving it.

Key-person dependency. If only one person in the company can run it, everything stops the moment they leave. Part of what the management fee bought was that the work carries on when a staff member quits.

How the cost changes

Put the management fee and internal hours on the same footing and the dividing line appears. At a 20 percent rate with a fifty-thousand-yen minimum, here is the monthly cost of outsourcing alongside the time the same operations take if you run them yourself. The hours assume one or two platforms, and grow as you add more.

Monthly ad spendOutsourced (20%, ¥50k min)Internal hours, rough guideAt ¥3,000 an hour
¥100,000¥50,0008 to 12 hours a month¥24,000 to ¥36,000
¥300,000¥60,00012 to 18 hours a month¥36,000 to ¥54,000
¥500,000¥100,00015 to 25 hours a month¥45,000 to ¥75,000
¥1,000,000¥200,00020 to 30 hours a month¥60,000 to ¥90,000
¥3,000,000¥600,00030 to 50 hours a month¥90,000 to ¥150,000

While ad spend is small, the amounts are close to even. At ¥100,000 a month, outsourcing costs ¥50,000 against roughly ¥30,000 in-house, so you save about ¥20,000. Whether you set aside ten hours a month for that difference depends on what else those ten hours could go to.

Outsourcing gets relatively more expensive as ad spend grows because the fee is proportional to spend while the volume of work is not. Running ¥3,000,000 a month is not ten times the work of running ¥300,000. That gap is why advertisers who have reached some scale start thinking about bringing ad ops in-house. A further breakdown of what the fee is made of is in Are ad agency management fees too high?.

Two costs are missing from the table. Recruiting and payroll if you hire a new operator, and the cost of management tooling. If an existing employee takes it on alongside their job, the hourly conversion in the table covers the former; if you place someone dedicated on it, the order of magnitude changes.

What happens to the account and the learning when you switch

This is where in-housing trips people up most. It is worth verifying before you work out the fee arithmetic.

Whose name the account is in. If it was created in your own company’s name, removing the agency’s administrator access leaves the account with you, operational history and all. In the agency’s name, you can lose the account entirely at cancellation. In that case you rebuild from scratch and the past data does not come with you. Check the contract or the console first to see whose name it is currently in.

Automated bidding’s learning. If you take over the same account, conversion history remains, so bid learning in principle continues as it is. But change the bid strategy after taking over, or move the target CPA substantially, and learning restarts from there. Rebuilding the account naturally starts from zero, and results only become readable two weeks to a month later.

The three things to take with you. Move these three into your own hands while the contract is still running and you avoid starting from zero after the switch. Administrator access to the account, the conversion tracking configuration, and the history of changes plus the negative keyword list. Negatives in particular build up over years, and losing them means doing the same waste over again.

The numbers get noisy right after a switch. The change of operator itself and the learning restarting land on top of each other. Read the worsened numbers here as the switch having failed and you will misjudge it. For at least a month, preferably two, it is safer not to compare simply against the previous period.

What to check when you are told “we are in the learning phase”

Plenty of advertisers have asked their agency about results and had this come back. The awkward part is that the explanation is often factually correct. Being correct, it cannot be argued with, and every option but waiting disappears. And while you wait, whether operations are moving or sitting still stays unknown.

For anyone considering in-housing, this moment is evidence. Three questions establish whether there are grounds to wait.

  1. When, and what change, put it into the learning phase? Whether a date and a specific change come back. A bid strategy change, a target change, a new campaign. All of them restart learning, but with nothing done there is no learning phase to enter.
  2. What are you watching to judge when it ends? For both Google Ads automated bidding and Meta Ads ad sets, the condition for exiting is accumulating a certain number of conversions. How many have accumulated, and how many more are needed. See whether the answer comes in conversions rather than in days.
  3. After the last learning phase ended, how far did the numbers recover? If you have had the same explanation before, the result from that time should exist as a record.

If all three come back answered, the account is being run and there is a reason to wait. No date for the change, no measure for the end beyond “a little longer”, no record of last time. Hit any of those and it is the grounds for waiting that have not been verified.

These three are not questions for doubting the agency; they are questions for estimating the wait yourself. And after in-housing, you become the side answering them. Whether you can stay in a state where you can answer is what decides whether running it yourself lasts. If you are still at the stage of looking for ways to see the substance while leaving it delegated, Reading your agency’s ad report will be closer to the mark.

Choosing not to bring everything in-house

In-housing is not a choice between delegating and doing it yourself. In practice more advertisers stop partway, and we often see that working out better.

Outsource the design only. Commission account structure, keyword design, and measurement implementation as a one-off at the start, and run day-to-day operations yourself. You borrow outside hands for the hard part of getting started and hold the continuing part.

Split by platform. Search in-house, video and display outside, for instance. The thinking is to delegate only the platforms where keeping up with specs is punishing.

Keep only the supervision. The operational work stays outside while approval of changes and checking the numbers sit with you. The fee remains, but the state of not being able to see inside is resolved.

Whichever form you pick, keep account permissions and the measurement configuration with your own company. With those in hand you can change the arrangement later.

Handing the operational work to AI

One more option has appeared in the last few years. Routine work such as daily checks and adjustments goes to AI, and people concentrate on judgment and approval. Of the work that increases when you bring ad ops in-house, what eats the time is less the judgment itself than the checking, trafficking, and tallying that come before it. Shift that toward the machine and the range you can hold internally changes.

Delegate badly, though, and a different cost appears. Delegate without handing over the criteria for judgment and changes based only on the surface of the numbers get executed quickly. Two lines are worth verifying: that changes touching budgets or bids can require approval, and that a record of the changes made remains so you can revert. This is covered in Is it safe to hand ad ops to Claude Code?.

mureo, the tool we build, is this form as a product. Pricing is flat rather than a rate on ad spend, so cost does not jump when you raise budget. AI runs operations daily across platforms from Google Ads to SmartNews, and shows you directly on screen what was spent where and why each judgment was made. Changes to budgets or bids go through human approval, and operations can be stopped and reverted at any time. Before committing fully to in-house, you can take the middle position of shifting only the operational work to the machine.

Frequently asked questions

Does bringing ad ops in-house improve results?

Not necessarily. Business context being reflected and decisions getting faster work in your favour, but there are forces in the other direction too: falling behind platform specs, or stopping when a person leaves. Results improve most readily when your product lineup turns over quickly, and when the round trip with the agency really was the bottleneck. Conversely, if operations are running without particular problems and the fee amount is your only complaint, changing the pricing model solves it faster.

At what level of ad spend should I consider in-housing?

The amount alone does not settle it. What decides it is whether there is someone inside the company you can put on operations, and how much that person’s time is worth elsewhere. As a rough marker, above ¥500,000 of monthly ad spend the management fee passes ¥100,000 and the gap against the converted value of internal hours starts to widen. If it becomes a question of hiring someone dedicated, though, the order of magnitude shifts and the arithmetic is different.

Can we run it ourselves with no experience?

For search advertising at least, you can start small and learn. The hard part is continuing rather than starting, and the dividing line is whether you can build the daily checks into your working routine. Commissioning only the initial design as a one-off and taking operations on yourself is the realistic form.

When and how should I tell the agency?

Check the notice period in the contract first. Contracts commonly ask for one to three months’ notice. Before you tell them, it is safer to have the account ownership and permissions, the measurement configuration, and the negative keyword list assembled in your own hands. Ask for the handover after the relationship has soured and what you need may not come out.

Can we go back to an agency after in-housing?

You can. If you have kept the account in your own company’s name, handing over administrator access is all it takes. Put another way, rebuilding the account in your own name at the time you bring ad ops in-house leaves you free to move either way afterwards.

How do we reduce the risk of a one-person setup?

Run it in a way that leaves a history of changes and a record of why each judgment was made. Avoid the state where it exists only in one person’s head and the cost of handing over drops sharply. The problems of one person carrying multiple accounts are also covered in Why ad ops becomes key-person dependent.

Summary

What comes back when you bring ad ops in-house is speed of decision, business context, the management fee, and knowledge that stays in the company. What transfers to you in return is the daily checks, production and trafficking, keeping up with platform specs, responsibility for judgment, and the risk of key-person dependency. On amounts alone, outsourcing turns expensive from around ¥500,000 of ad spend, but since the fee you save turns into time inside the company, you cannot decide without adding what else that time could go to.

Once you have decided to switch, confirm whose name the account is in first, and move administrator access, the conversion tracking configuration, and the change history into your own hands while the contract is still running. The numbers get noisy for the first month or two after the switch, so do not compare simply against the previous period. And you do not need to bring everything in-house. Outsource the design only, split by platform, keep only the supervision. In any of those forms, as long as account permissions and the measurement configuration sit with your own company, you can rearrange it later.

For cost benchmarks and how to pick a provider if you stay outsourced, see Outsourcing search ad management; for the checks specific to Google Ads, Outsourcing Google Ads management.