The best UA teams spend less time in ad managers

Time spent inside an ad manager is now a reasonable proxy for how little a UA team has left to control. The platforms have spent three years absorbing bid decisions into their own models, then placement, then audience, and they have been open about it: Unity told investors in its February results that Vector was already 56 percent of Grow revenue, Mintegral's February trends report with Insightrackr put automated bidding spend growth above 50 percent, and AppLovin has not offered manual bidding in any meaningful sense since Axon 2.0. The console is still there. The decisions that used to live in it are not.

The claim in this piece is that the teams performing best in 2026 have stopped treating the ad manager as the workplace and started treating it as a settings page. Some experienced buyers will resist that, because the console is where they built their expertise. The mechanism below explains why the expertise has moved, and the framework at the end suggests where the hours should go instead.

What the platform took, and why it wanted it

Automated bidding usually gets described as a convenience. It works better as a transfer of information advantage. When a buyer set bids by geo, by placement, by daypart, that judgement was an input the platform could not see and could not learn from. Hand over a target ROAS and a creative set instead, and the platform sees everything at once: which impressions cleared, which of them installed, which of those paid and how the same pattern plays out across every advertiser in the auction.

The platform's model learns from the whole market. The buyer's judgement learned from one game. In a contest of bid decisions the model wins, and the platforms knew it well before the buyers did. Which is why every major network has pushed manual controls to the margins. Unity's announcement in March that the ironSource Ads network would close at the end of April, with the company concentrating on Vector, was the clearest statement yet: a legacy console with human-set waterfalls was worth less to Unity than a model with fewer knobs.

The incentive is straightforward. A platform that controls bidding captures the surplus between what an impression is worth to the advertiser and what it costs. Manual buyers used to capture some of that surplus themselves. The automated buyer gives it back in exchange for scale and lower labour cost, which for most games is a fair trade. It is still a trade, and knowing what you gave away is the first step to deciding where to compete instead.

The second-order effect: console skill has become a liability

Now the part that gets less coverage. A team whose senior people are excellent at ad-manager operations has every incentive to keep working in the ad manager, because that is where their skill shows up and gets rewarded. So they keep adjusting things: splitting campaigns, tweaking targets, pausing ad sets, relaunching them a day later. On an automated platform most of those actions are noise at best and harmful at worst, because every reset discards what the model had learned about that campaign.

The cost shows up in two places. Well-meaning intervention keeps resetting the platform's model, so it never reaches the stable state where automated bidding actually outperforms. The larger cost is that hours spent on console operations are hours not spent on the two inputs the platform cannot supply itself: the creative it has to serve and the conversion signal it has to optimise toward.

We wrote in "Playable ads are becoming the operating system of mobile UA" that creative had become the buyer's primary lever. The corollary is that a team's time allocation should look like its lever allocation. For most teams it does not.

Where the hours should go

Sort the working week into four buckets, then check the proportions against what the platform can and cannot do for you.

• Signal design: the events sent back to the network, their timing, their values, and what the model is being told "good" looks like. On SKAN and AAK this is conversion schema design. On Android and in-network postbacks it is event selection and value passing. A wrong signal makes every automated bid wrong in the same direction.
• Creative supply: concepts, variants, playables, refresh cadence, and the read-out of which concepts the model favours and why. This is where marginal effort has the highest return, because the model cannot invent a new hook.
• Cohort economics: retention, monetisation and payback curves by source, checked against what the platform reports. This is the audit function. It catches the case where the model is hitting its target on a metric that no longer predicts revenue.
• Console operations: budgets, targets, launches, pauses. Necessary, but it moves very little on any platform with a mature model.

For a team of four on automated networks, an illustrative allocation might be 20 percent signal, 40 percent creative, 30 percent cohort economics, 10 percent console. A team we would worry about runs 10, 20, 10, 60. That is roughly what a 2021 media-buying team looked like, and plenty still do.

The decision rule for touching a campaign

Console time matters less than it did. Less is not none, and the rule we suggest is that a buyer changes a live campaign only when one of three conditions holds.

Signal has moved: the events or values going back to the network are different now, so the model needs a new target. Or creative supply has moved, meaning a concept has fatigued or a new one is ready and the campaign structure has to accommodate it. The last one is economics, where the cohort audit shows the platform's optimisation metric has decoupled from actual payback, which makes the target itself wrong.

None of those? Leave it alone. Restlessness is not a condition, and nor is a bad Tuesday. Automated models need consistent inputs to converge, and the most common failure we see in reviews is a campaign that never converged because someone kept helping.

What this means for hiring and for vendors

The team that follows this allocation looks different from the team that ran the 2021 console. It has more analysts than buyers, its creative strategist sits in the weekly performance review rather than reporting to it, and its senior buyer's job is mostly to know when a platform's model is lying.

It also has a different relationship with its networks. A team that spends little time in the console is less impressed by console features and far more interested in what the network will disclose about how its model treats creative and signal. That is a better negotiating position than the one most buyers currently hold. It is available to any team willing to close the dashboard for a few hours a week and look at what the dashboard cannot show them.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Automation did not remove judgement, it moved it
• What Big-Budget Launch Campaigns Teach Mobile UA Teams