Goal
Where can you take media budget out without losing growth?
Every larger account holds money that adds nothing. Finding it is work on the data: on double-counted conversions, on campaigns taking share from each other, and on frequency that nobody has ever added up.
What we do: We find double-counted conversions, campaigns that take share from each other and contact doses nobody has added up, and rebuild in steps; you take budget out without losing growth.
Why this is hard right now
Accounts that have grown over years carry decisions nobody remembers the reasons for. Exclusion lists sit unchanged for years while the search terms report shows queries that obviously do not fit and still collect clicks. Campaign types compete against each other: search campaigns lose share to a Performance Max campaign in the same account while the total stays flat.
The second source is measurement itself. The same order shows up as a purchase and as a thank-you page view, or it arrives from browser and server at once. The metrics then look better than they are, and because every automated bidding strategy bids on exactly these numbers, it bids too high. An efficiency problem that looks like success.
The third source is frequency. Three campaigns serve the same audience with their own frequency caps each; added up, the same person sees far more than is set anywhere. In programmatic delivery there is the extra trap that a cap at insertion order level constrains the caps of the line items below it, which is regularly missed when caps sit on only one level.
The questions we work through with you
Which campaigns buy demand that would have arrived anyway?
How you can tellBlended ROAS looks good because brand queries carry it. Strip them out and little new business remains. Or the share of existing customers among your conversions is high without total revenue moving.
What we doWe sort the money by purpose: brand, existing customers and new business are reported separately. Existing customers can be deliberately excluded via their own segments, so budget flows into new business. Only after this separation can anyone say what a euro actually did.
Are you paying twice for the same conversion?
How you can tellThe account holds two conversion actions for the same event, or the same order arrives from browser and server. Visible in a reconciliation with your business system: the platform reports more conversions than there are orders.
What we doWe review conversion actions, values and double counting before judging anything in the account. If the underlying number is wrong, every further statement is a guess. After a rebuild we reconcile against the business system instead of just watching a number go up.
How often does the same person really see your ads?
How you can tellThe line item has a frequency cap, the insertion order does not; across several line items the dose adds up because the outer bracket is missing. Or the time window does not match the buying cycle: with long decision journeys the cap is set per day, and the person sees you often on one day and never again.
What we doWe set the bracket on the upper level and refine below it, because the upper cap constrains the lower ones. We add up frequency across every campaign that serves the same audience and align the window with your decision cycle. Set too tight it brakes just as surely as a missing cap harms: budget then goes unspent although reach is available.
Which spend keeps running just because it was once set up?
How you can tellSeveral programmatic deals are running whose volume is never used up: the budget is committed without producing reach. Usually there is also no metric that puts the deal against comparable inventory from the open auction, so the premium cannot be justified.
What we doWe settle before buying whether volume and environment really need to be guaranteed. If the answer is no, a non-guaranteed form is usually the more flexible choice. We put comparable auction inventory next to every deal and continuously monitor whether committed volumes are actually being delivered.
What do your own rebuilds cost you?
How you can tellThe bidding strategy is switched every few weeks. Every switch resets the learning, and whoever switches often pays for learning phases permanently without ever seeing a settled result. It gets especially expensive when the learning phase lands in your strongest revenue weeks.
What we doWe schedule changes for quiet weeks and never several at once, and we change things in sequence instead of everything at the same time. What changes during a learning phase gets documented so the comparison holds up afterwards. Individual findings become a list sorted by impact with a few root causes, each with the effort next to it.
The expertise we combine for this
- SEA / Paid SearchCleans up structure, exclusions and overlap between campaign types before anyone touches the bids.
- Programmatic / Display / CTVSets frequency caps across campaigns and puts every deal against comparable inventory from the open auction.
- Measurement & AttributionMakes sure a budget shift rests on a number that means the same thing in every system.
- Tracking & AnalyticsFinds double-counted conversions before the automation bids too high on them.
And the technology to match
A deviation should surface where it happens, not in the monthly report. Rules and alerts run continuously; exceptions are decided by a team that knows the patterns from many accounts. Limits and targets are set by you.
Evidence from client work
Frequently asked questions
What does a Google Ads audit deliver, and what is in it?
It answers three questions: what are you spending money on, what is being optimized for, and which of it moves your business. The first look goes to the conversion actions and their values, because every automated strategy bids on them. The result is a work list sorted by impact, not a list of 200 defects.
Do I get a list of every single error?
You get a list sorted by impact, not by completeness. 200 individual findings are usually two or three root causes, and those sit at the top. Whoever works through symptoms one by one stays busy for a quarter without anything changing.
How high should frequency be?
That depends on audience size, budget and buying cycle; any general number would be a guess. The usable yardstick is the point at which additional contacts stop improving the result. For that we watch reach, frequency and outcome together.
What happens when insertion order and line item carry different caps?
The insertion order cap constrains the caps of the line items below it. With four impressions above and three per day below, the stricter setting applies to the line item.
Is a programmatic deal worth it compared to the open auction?
If volume or environment need to be guaranteed, yes. Otherwise you are paying for a certainty you do not need. The answer only becomes provable through a comparison with similar auction inventory. With guaranteed deals there is the added catch that targeting, budget and pacing are fixed alongside and can hardly be changed afterwards.
Is an audit worth it on a small budget too?
Especially then, because on a small budget a single error costs a larger share. Double-counted conversions hit a small account exactly like a large one.
Further reading
The special topics cover individual points in detail.
- Your Google Ads account has grown, and nobody remembers why
- The same people see your ad ten times a day
- Open auction or deal: where should your budget buy?
- Which bidding strategy fits your account?
- Your retargeting lists shrink month after month
- Performance Max is running, but you cannot see where the conversions come from
From our Insider articles: