Most PPC agencies charge a percentage of your ad spend, and most founders sign it without thinking. That is a mistake. A percentage-of-spend model quietly pays the agency more to recommend the thing that suits them, and less to recommend the thing that suits you. A flat fee is not a magic bullet, but for the majority of accounts it is a more honest structure. This post explains why, when the reverse is true, and what to check before you sign either one.
The structural conflict hiding in your agency contract
Percentage-of-spend is the default in PPC because it is easy to sell and easy to scale internally. It also creates a conflict of interest that no amount of good intent can fully neutralise.
The mechanics are simple. Your agency earns 15% (or 20%, or 25%) of whatever you spend on Google and Meta. If they recommend cutting a wasted campaign, their invoice shrinks. If they recommend pushing more budget into a channel that is doing fine, their invoice grows. Efficiency work, the exact thing you are paying an expert for, directly reduces their revenue.
When the recommendation to raise budget is also the recommendation that raises the agency's own invoice, that advice is no longer neutral, and the client is rarely in a position to audit it.
Nobody has to be dishonest for this to distort decisions. It shapes which experiments get proposed, which under-performing campaigns get quietly left alive, and how aggressively spend caps get challenged. A flat fee removes the mechanism. You still need good reporting and a competent team, but the incentive to keep spend high for its own sake is gone.
Why agency workload and ad spend diverge
The unspoken assumption behind percentage pricing is that a bigger budget means proportionally more work. It does not.
Campaign management labour is driven by the number of campaigns, the creative testing cadence, feed complexity in shopping accounts, conversion tracking architecture, and how noisy the signal is. Not by the size of the media budget. A £5,000-a-month account with twelve campaigns, a broken product feed and shaky conversion tracking is genuinely more work than a £30,000-a-month account with three well-structured Performance Max campaigns and stable signal.
A larger budget does require more daily monitoring. It does not require ten times the manual labour. Even agencies that defend percentage pricing acknowledge this: large accounts routinely negotiate tiered percentages or hybrid retainers precisely because labour does not scale in a straight line with spend. That concession is the tell. If workload tracked spend, tiering would not be necessary.
Pricing based on spend alone is a bad proxy for the thing you are actually buying, which is expert time and judgement.
What the numbers look like over twelve months
The compounding effect of percentage pricing is where founders get caught out. It looks reasonable in month one and expensive in month twelve.
Take a mid-sized ecommerce account that starts the year at moderate spend and grows as campaigns prove themselves. Under a percentage-of-spend model, every increase in media budget triggers a proportional increase in the management fee. If your monthly spend doubles because you found a winning campaign, the agency's fee doubles too, without them necessarily doing twice the work. Scaling a winning campaign should be the moment your unit economics improve. Under percentage-of-spend, part of that upside is taxed by your agency.
Under a flat fee, the management cost is fixed regardless of whether the account is running hot or cold that month. In a slow quarter you are not stuck paying an inflated fee based on last quarter's numbers. In a growth quarter you keep the full benefit of the scaling. Your finance team can also actually plan, because the management line does not move with the media line.
A practical test
Ask your prospective agency to model your management fee at three spend levels: current, half current, and double current. If the fee moves proportionally, you are buying spend, not work. If it stays flat or moves in bands, you are buying the thing you actually want.
The percentage model is defensible when spend is high enough that workload genuinely scales with it. For accounts below that threshold, and there are more of them than the industry admits, a flat fee is simply better maths for the client.
What a flat fee must include to be worth anything
A flat fee is only better if the scope is defined properly. A cheap-looking flat fee with a narrow scope will hit you with add-on invoices for every task that falls outside it, and by month six you are worse off than under a transparent percentage arrangement.
Before signing any flat-fee retainer, get the scope in writing at a task level, not a category level. "Reporting" is not a scope item. "One monthly performance report covering channel-level ROAS, top campaigns by profit, tracking health flags, and next-month recommendations" is a scope item.
Scope items clients routinely assume are included
- Conversion tracking QA on an ongoing basis, not just at setup
- Feed review and diagnostics for shopping and Performance Max
- Landing page feedback with specific recommendations, not just "the page is slow"
- Call tracking configuration and lead quality feedback loops
- Creative testing cadence: how many new ads per month, across which channels
- Attribution and consent mode maintenance as platforms change
If any of those are excluded, that is fine, but you need to know before you sign. This connects to a wider point about signal quality. If your tracking is broken you are paying for optimisation against noise, which is why we treat fixing conversion tracking with server-side GA4 as scope-in for most retainers rather than a paid extra.
The minimum-spend floor problem
Percentage pricing gets worse at small budgets because of the minimum fee. If the agency's floor is £1,000 a month and you are spending £2,500 on media, your effective management rate is 40%. The percentage on the proposal is not the percentage you actually pay.
Flat fees have the opposite version of the same problem. On a very small budget, a flat management fee can cost more than the media itself, which is uneconomical. Below roughly £1,500 in monthly ad spend, running campaigns yourself, or with a freelancer, is usually the rational choice regardless of which model the agency prefers. An honest agency will tell you that rather than take the retainer.
Get plain-English guides like this in your inbox.
One short email a month. WordPress, Shopify, SEO, no fluff. Unsubscribe in one click.
We never share your email.
When percentage-of-spend is actually the right call
Being useful means naming when the argument breaks down. Percentage-of-spend is genuinely the right structure in a few cases.
Genuinely large accounts. Above a threshold, workload does scale with spend. More campaigns, more creative iterations, more segment testing, more daily attention to auction dynamics. A low, tiered percentage on a large account can be fairer to both sides than a flat retainer that undervalues the actual labour.
Highly variable spend. Seasonal businesses, launch-driven brands and marketplaces with unpredictable inventory sometimes need spend to swing hard month to month. A flat fee priced against peak activity is unfair to the client in quiet months. A percentage that flexes with the account can be a cleaner fit, provided the ceiling is capped so a good month does not become a windfall for the agency.
Very early-stage accounts. If nobody yet knows whether the channel will work, a percentage model shares some of the risk. You pay less while testing at low spend, and more once things are proven. Get the terms of the transition documented so you are not locked into a percentage indefinitely.
Performance-heavy hybrids. A modest flat retainer plus a genuine performance bonus tied to revenue or profit, not clicks or impression share, can work. The detail that matters is which metric the bonus tracks. Anything the agency can influence without moving your business forward is a metric that will be gamed. Tie it to profit contribution or nothing.
For everything else, which is most SMB and mid-market accounts, a flat fee with a clearly documented scope is the structure that aligns incentives with your interests. This matters even more as ad platforms consolidate around AI-led campaign types, a shift we cover in Google Ads in 2026: Performance Max and AI Max explained, because with fewer manual levers to pull, the value you are paying for is judgement, not lever-pulling volume.
How to evaluate any agency pricing model before you sign
Ignore the pricing label and ask five questions instead.
- What exactly is in scope, at task level? If the answer is vague, the scope is vague. Vague scopes become invoices.
- How does the fee move if my spend halves or doubles? This exposes the incentive structure faster than any explanation the agency can give you.
- What is your minimum fee, and what is my effective management rate at my current spend? For percentage models especially, this is the number that matters.
- What happens when you recommend I cut spend? Listen for whether the answer treats efficiency as a normal part of the work or as an exception.
- What does the reporting actually contain, and how often? Monthly reports framed around impressions are a signal the agency is optimising for something other than your bottom line.
Our own digital marketing services sit on flat-fee retainers for the reasons in this post. We think it produces better decisions on both sides, and it makes conversations about cutting spend as easy as conversations about increasing it. If you are still shortlisting agencies, we go deeper on the selection process in how to choose a digital marketing agency in London.
The pricing model is not just a commercial detail. It is a structural choice about what your agency is incentivised to recommend. Pick the one whose incentives match yours, get the scope in writing, and renegotiate as the account changes shape.
If you are stuck with a percentage-of-spend contract and unsure whether the recommendations you are getting are neutral, see our work or bring us the account for a second read.
