Teardown: The Performance Guarantee That Lets an Agency Keep Its Fee

A Google Ads contract can promise performance and still let the agency keep every dollar when performance misses. I chose this composite performance guarantee because its five lines show how: charge on spend, count leads, choose a soft baseline, exclude client-side changes, then offer an exit instead of a remedy.

None of those lines looks outrageous on its own. Together, they put the risk where it already was: on the business paying the ad bill. I’ll read the clause in the order a founder would encounter it, give each line a verdict, and show what I would ask to change before signing.

The artefact: five lines that sound like a guarantee

This is a composite clause, not a quotation from one contract. Its parts capture the structure of a familiar pay-for-performance pitch:

Fee: 15% of ad spend. We guarantee growth in leads versus prior-period performance. Excludes results affected by client-side changes outside our control, including landing pages, offers, budgets and tracking. If targets are not met, client may terminate on 30 days notice.

The problem is not that an agency needs a fee, a measurement period, or sensible limits on what it controls. It does. The problem is what happens when you read those terms together. The fee can rise while efficiency falls. The target can pass without producing pipeline. The baseline and exclusions give the vendor room to dispute a miss. And if a miss survives all that, the remedy is your right to leave.

Start with the money.

Line 1: “Fee: 15% of ad spend”

This line decides whether the rest of the promise has teeth. Percentage-of-spend pricing commonly runs 10 to 20% of monthly budget, with higher percentages on smaller budgets. Say you spend $20,000 a month. At 15%, the management fee is $3,000 a month, before ad spend, creative, or landing-page tools. I ran the broader comparison against CPA and flat-fee pricing in the $20k Google Ads test. The incentive problem is the same here: the fee grows when spend grows, whether efficiency improves or not.

Raise the budget by $10,000 and the 15% fee rises by $1,500. Recommend a budget cut because search demand softened or the landing page cannot handle more volume, and the fee falls. I used to tell clients this structure aligned incentives because everyone wanted growth. I was wrong. It rewards more spend directly; lower cost per acquisition only helps the vendor indirectly. As one contract guide notes, an agency can get paid more in a bad month if spend scales while CPL rises 40%.

That does not prove every agency on a percentage fee will waste your budget. It tells you what this clause pays them to do. Verdict: a performance promise should not sit beneath a fee that increases automatically when the vendor asks you to spend more. Change the fee basis, or make the guaranteed outcome strong enough to counter that incentive.

Line 2: “We guarantee growth in leads”

A lead target sounds useful until you ask what qualifies. Does a form fill count? A call? A chat from someone who cannot buy? If the contract does not say, the agency can satisfy the guarantee by increasing activity at the top of a report while your sales team gets no closer to a deal.

Performance deals can price qualified home-services leads at $50 to $150 or take 10 to 30% of sale value on pay-per-sale. Those structures need an upfront definition and CRM or call-tracking proof. Without one, a per-lead incentive favors quantity over quality and can favor bottom-funnel branded search over testing that builds new demand. The report gets prettier. The business may not.

For B2B, the gap between leads and pipeline can be brutal. One guide puts MQL-to-SQL conversion at 9.8%. At that rate, roughly 90 of 100 MQLs do not become SQLs. A guarantee on the original 100 says little about the sales conversations the business needs. Weekly lead count climbs; closers work through poor-fit enquiries; pipeline reads zero while the guarantee reads met.

I would tie the obligation to MROI, or to MQLs and SQLs with mutually agreed qualification criteria. Closed revenue, qualified pipeline, CPA, and ROAS can also be useful, depending on what the business can reliably count. Reach, clicks, and raw leads still matter as diagnostics. They just should not become the prize the contract pays someone to chase.

Verdict: define the outcome before anyone starts optimizing toward it. If the contract guarantees leads, make it say which leads count and where that count comes from.

Line 3: “Versus prior-period performance”

Prior period has the reassuring sound of a spreadsheet cell. Which period? Measured in which system? Over what window? Compared with what happens when tracking changes? The clause answers none of those questions, so the baseline can do work the campaign did not.

Seasonality makes the weakness obvious. Compare a home-services account’s March with February and ordinary seasonal demand can flatter the result. Compare January with December and the same problem can appear in another direction. Neither comparison is automatically dishonest. Neither establishes improvement simply because one number is higher.

A baseline worth signing names the starting number, the source of truth, the judgment window, and the treatment of tracking changes. Contract examples include 30 days’ notice if blended CPL runs 35% above an agreed baseline for two straight months or ROAS stays below 1.5x for two months. Those are examples of specific triggers, not targets I would paste into every account. The important part is that both sides know what will be measured before the result arrives.

The window matters too. Performance deals may involve 60 to 90 days of optimization while you pay full price. A short window during that ramp can mislead in either direction; an undefined window lets either party argue over when the test began. Write down the dates, the baseline, and the counting system before payment.

Verdict: prior period is not a baseline. Until the contract pins it down, the target is movable.

Line 4: “Excludes results affected by client-side changes”

This is the line I understand and still would not sign as written. Landing pages, offers, budgets, and tracking affect results. An agency should not have to guarantee an outcome after the client changes the offer or breaks measurement. But an exclusion this broad can swallow the promise.

A changed headline, a delayed page approval, or a week with a paused budget could become grounds to disregard the result altogether. One review of agency-contract red flags describes the version where client deadlines are binding and a missed approval becomes the agency’s excuse. The wording here leaves the same opening: it does not say how large a change must be, whether it caused the miss, or which part of the measurement period is affected.

Landing pages really can tank accounts. I have watched Quality Score fall apart after a page changed without warning. That is not a reason to pretend the page does not matter. It is a reason to write a narrower exclusion. Name the change, record when it happened, and specify how it affects the guarantee. Then name what the vendor still owns: the tracking setup it manages, page-speed issues it can fix, the negative list, and the bid logic. If every factor that touches conversion sits outside its responsibility, what exactly is left inside the guarantee?

Verdict: keep exclusions for genuine changes outside the vendor’s control, but do not let client-side change serve as a universal escape hatch. A promise with a doorway built into every wall is a hallway.

Line 5: “Client may terminate on 30 days notice”

Read that line and ask who pays for the miss. You paid the fees and ad budget. The target did not land. The vendor keeps what you have already paid, and you receive permission to start again with someone else. An exit is useful, but it is not a vendor-paid remedy.

There is a fair version of an exit clause: 30 days’ notice, no wind-down fee, and clear client ownership of the account, tracking, and creative. That protects your ability to leave. It matters because contracts can add a wind-down fee of one to three months, voided setup fees, non-competes lasting up to 24 months, or transition fees for access to your own credentials. None of that belongs in a clean handoff.

Ownership belongs here for another reason. Search-term history, negative lists, tracking, and creative tests represent work the business paid to build. If they sit only in an agency-controlled account or drive, leaving can mean losing the learning along with the vendor. I have learned to ask for the ownership language before the engagement starts, not during the awkward final week.

But even a frictionless exit does not make the performance guarantee real. If the target is missed, the remedy should cost the vendor: for example, continued execution without another management fee until the agreed number is reached. Verdict: keep the right to leave and add a separate consequence for missing the target. If failure costs the vendor nothing but your business, it did not guarantee the result.

The replacement clause: five things to get in writing

After those five lines, I would send the contract back with five corresponding changes. This is not about winning a negotiation over adjectives. It is about making the promised result measurable and the consequence of missing it unmistakable.

1. Target before payment. Put the exact lift, CPA ceiling, or pipeline number in the signed agreement before the invoice. Do not leave the target for a kickoff call.
2. A metric tied to the business. Use qualified pipeline, closed-won revenue, CPA, or ROAS where those can be counted reliably. If SQLs are the target, agree on the qualification criteria line by line. Do not let raw leads stand in for money.
3. A fixed baseline and window. Name the starting number, source of truth, dates, treatment of tracking changes, and seasonality call-outs. Set the judgment period in writing rather than picking one lucky month after the fact.
4. A vendor-paid remedy. If the target is missed, the vendor continues the agreed work without another fee until it hits. Cancellation rights are separate.
5. Ownership of the build. Keep the ad account, tracking, negatives, and creative in your hands when the engagement ends. The broader version of this review is in the 15 contract checks to run before you sign.

That list is a test of what the vendor is willing to be accountable for. It also applies to ours.

Holding the groas sprint to the same test

I would not run another agency’s clause through five checks and wave ours past them. The groas paid-ads and SEO sprints charge one flat fee, paid once up front and quoted on the fit call. The stated target is 40% more sales from ads or 40% more visibility in search within 90 days. It goes in writing before payment; the fee does not rise with ad spend; and a miss means free work until the number hits. Those terms address the fee and remedy problems in the composite clause directly.

There is still a distinction worth keeping in view: 40% more search visibility is not 40% more revenue. If you are considering an SEO sprint, ask how visibility is counted and whether that is the result you want to buy. If you are considering paid ads, establish the sales baseline and counting method before paying. A written target is only as useful as the metric underneath it.

This model will not suit everyone. If you have no conversion tracking, tiny search volume, or a two-week deadline, a fixed 90-day target will frustrate you. It needs enough traffic to learn from and enough calendar for bids, budgets, and pages to do their work. Do not mistake a stronger contract for a shortcut around those constraints.

Overall verdict: keep the promise, replace the escape routes

The composite clause gets one thing right: it puts a performance promise in writing. I would keep that instinct, plus a clear right to leave. I would replace almost everything that determines whether the promise can be missed: the spend-linked fee, the raw-lead target, the floating baseline, the blanket exclusion, and the exit presented as compensation.

My final test is plain. Send the contract back with the target, counting method, baseline, exclusions, remedy, and ownership terms typed in. Ask the vendor to agree to them in writing. If the answer is a call to explain the nuance rather than an edited contract, you have learned something before paying for it. A guarantee worth signing says what counts, when it counts, and what the vendor owes if it misses.