The Real Cost of a Bad PPC Agency
The invoice is not the cost. For a £50k/month advertiser, an average agency quietly costs about £240,000 in cash and another £60,000 in growth left on the table. Here is where it goes.
Founders ask what a good agency costs. Almost nobody asks what a bad one costs. The bad one costs more. It just costs it in places nobody bothers to measure.
This is the arithmetic. One brand, twelve months, £600,000 of paid spend. Where a "reasonable ROAS, everything looks fine" agency quietly leaves £240k on the floor and the same amount again in growth you never booked.
The invoice is not the cost
A £3k/month retainer looks like a £36k/year expense. That is what appears on the P&L. It is also the smallest number in this article. Because the agency does not just cost you their fee. They cost you the shape of every pound you spend after that.
On £50k/month of ad spend, the retainer is 6% of the paid media budget. The other 94% is where the damage happens. And that damage compounds monthly, not annually, because bidding decisions made in January are still misfiring in June.
So the useful question is not "am I paying too much for the retainer". It is "what shape does £600k of spend take when it is managed by someone reporting on ROAS instead of contribution margin". The answer, in our audit data, is depressingly consistent.
A worked example: £50k/month for 12 months
One brand. £600,000 total ad spend over 12 months. Blended target ROAS of 4:1, so £2.4m in reported paid revenue. Contribution margin of 30% at the SKU level once COGS, returns, shipping and payment fees are stripped out. Retainer £3k/month, sitting on top.
On paper this looks healthy. ROAS above target, revenue growing month on month, monthly report neat. Now decompose the £600k.
12%
of spend
Spend on unprofitable SKUs
£72,000
Roughly 20-35% of Shopping spend typically sits on SKUs that lose money at contribution margin. We are modelling the conservative end.
6%
of spend
Learning-phase resets
£36,000
Four to eight weeks a year of avoidable resets from ill-timed restructures, tag changes, PMax rebuilds. Bidding runs inefficient throughout.
8%
of spend
ROAS-chasing margin erosion
£48,000
Bidding to a blended ROAS target rewards low-margin, high-conversion SKUs. Cash comes in. Contribution margin walks out.
4%
of spend
Brand tax dressed as growth
£24,000
Branded search and remarketing conversions that would have happened anyway, counted as paid wins. Inflates ROAS, hides incremental performance.
10%
of spend
Opportunity cost
£60,000
The Scale SKUs that were underfunded because budget was tied up elsewhere. This is the growth you did not book, not the money you spent.
Cash cost of a "fine" year
£240,000
Plus £60,000 of growth you never booked. The retainer, at £36k, is the cheapest line in this equation.
Leak 1: Spend on the wrong SKUs — £72,000
Every catalogue contains SKUs that lose money at contribution margin. Low-margin bundles, high-return sizes, freight-heavy items, discontinued lines still stuck in Performance Max. In an average account they consume 20 to 35% of Shopping spend. We are modelling the conservative end at 12%.
The reason it persists is simple. If nobody has assigned SKUs to roles — Scale, Profit, Protect, Recovery, Gateway — then the algorithm optimises for the metric it was given. That metric is usually revenue or a blended ROAS. So it keeps buying clicks for the SKUs that convert cheapest, regardless of whether those SKUs make money. Read Every SKU Has a Job for the framework.
£72,000 a year of pure waste. Not underperforming spend. Loss-making spend.
Leak 2: Learning-phase resets — £36,000
Every time an agency restructures campaigns badly, retags conversions carelessly, or rebuilds Performance Max asset groups without a plan, Smart Bidding re-enters learning. During learning it bids inefficiently and spends more to gather data. In our audit sample, four to eight weeks of avoidable learning happens in an average year.
Six weeks of £11.5k/month spend running at, say, 25% worse efficiency is £17k of pure inefficiency. Do that twice a year and it is £34k. Round to £36k.
Nobody sees it because the report shows "we restructured" as a positive. In practice, learning periods are the most expensive weeks of the year, and unforced ones are the definition of avoidable cost.
Leak 3: ROAS-chasing eating margin — £48,000
Bidding to a blended ROAS target rewards whichever SKUs convert cheapest, not whichever SKUs make money. A 5:1 ROAS on a 15% margin bundle loses cash. A 3:1 ROAS on a 55% margin hero product is profitable. Blended targets cannot tell them apart.
So over a year, mix drifts. High-margin SKUs get outbid, low-margin SKUs consume more budget, the reported ROAS holds up, and contribution margin quietly drops 3 to 5 points. On £2.4m of paid revenue, that is £48k of margin the P&L never sees, because the P&L only sees the aggregate.
The fix is not clever. It is arithmetic. Bid to a break-even ROAS per SKU role, not a blended target. See POAS vs MER vs ROAS.
Leak 4: Brand tax dressed as growth — £24,000
Branded search and remarketing convert cheaply and inflate the report. A large portion of it would have converted anyway. When an agency counts those conversions as paid wins with no incrementality test, the account looks 10 to 20% better than reality.
The economic damage is not in overspending on brand terms — it is small — but in the false confidence it creates about the rest of the account. Executives greenlight spend increases based on a ROAS that is not being earned on non-brand traffic. Modelled cost of that misallocation on a £600k budget: about £24k.
Leak 5: The opportunity cost of the year you lost — £60,000
This one is the biggest, and it never appears in any invoice. If 12% of your budget is on loss-making SKUs, the same 12% is not on the Scale SKUs that would have compounded. Those SKUs did not grow. Their audiences did not accumulate. Their algorithmic learning did not deepen. Next year's baseline is lower than it should have been.
Model it conservatively at 10% of contribution margin foregone on the redirected spend and it is £60k. Model it aggressively — including the second-year compounding — and it is materially more. This is the number that turns a bad agency year into a two-year hole.
The total: what a bad year actually costs
£276,000 against a retainer of £36,000. The invoice was 13% of the real cost. The other 87% was invisible, because it lives inside the ad account and inside the mix, not on the ledger.
Why none of this shows up in the monthly report
Because the monthly report is designed by the agency being reported on. It contains ROAS, spend, conversions, impression share, click-through rate. It does not contain contribution margin by SKU role. It does not contain break-even ROAS. It does not contain a learning-days ledger. It does not contain incrementality on branded search.
Those five omissions are exactly where the £240k lives. It is not that the agency is hiding anything. It is that they built a scoreboard that cannot measure the leaks.
A report that reconciles to your P&L will find these leaks in a week.
A report that stops at ROAS will hide them for a year. That is the actual difference between a good agency and a bad one — not talent, not price, not headcount. Reporting architecture.
Our BOI® method is what a P&L-reconciled account looks like from the inside. SKU roles, contribution-margin bidding, and a break-even ROAS that means something.
Want to know which of these leaks are live in your account?
Our 9-Point Google Ads Profit Audit quantifies the five leaks above against your real spend and margin file. Free, capped at four a month. We ask for your cost file before the call so it is not a sales pitch dressed up as an audit.
Request a Free AuditThe verdict
A bad agency is not defined by incompetence. It is defined by reporting architecture. If the scoreboard cannot see contribution margin, SKU roles or incrementality, the account will bleed in exactly the five places above. It will bleed quietly. It will bleed predictably. And it will bleed at roughly seven times the retainer fee, every year, for as long as the arrangement holds.
The retainer is never the number worth arguing about. The 94% of spend it shapes is.
FAQ
Is £50k/month a realistic example?
It sits in the middle of our ICP. The pattern scales linearly. At £15k/month the numbers are smaller but the percentage leaks are identical. At £150k/month they are three times larger and the compounding is worse because more of the spend sits inside Performance Max.
Are these numbers made up?
The model is illustrative, but the leak percentages come from patterns we see in audits: roughly 20 to 35% of Shopping spend on SKUs that lose money at contribution margin, 4 to 8 weeks a year in avoidable learning resets, and 10 to 20% of blended ROAS driven by branded search that would have converted anyway. Your account will differ. The categories will not.
How is 'bad agency' being defined here?
Not incompetent. Not lazy. Simply an agency reporting on ROAS and spend rather than contribution margin, treating every SKU the same, and never asking about COGS, returns or product-level economics. The default setup of most retainers in this market.
Does switching agency guarantee I recover this?
No. Switching to another blended-ROAS agency just resets the clock. What recovers this is a commercial operating model: SKU roles, contribution margin bidding, a real break-even ROAS per product, and reporting that reconciles to the P&L. Any agency that can do that will recover it. Most cannot.
What is the fastest way to see if this applies to me?
Two questions to your current agency. One: what is my break-even ROAS by SKU role. Two: how much of last quarter's spend went to SKUs that lose money at CM3. If either answer is a shrug, the leaks in this article are almost certainly live in your account.
Is this an ad for your audit?
Partly. The audit is how we find these leaks in a specific account. But the article stands on its own: if you take nothing else away, ask the two questions in the FAQ above. You do not need us to run them.