A performance-based search deal can charge you for a miss and still call itself performance-based. The trick is usually in four blanks: the starting number, what counts as a win, which system judges it, and what happens when the target is missed. I have watched those blanks turn pay for results into a retainer with nicer wording. Keep the proposal open and put a written answer beside every item below. If the answer lives only on a sales call, it does not belong in your decision.

A. The target: make a win countable

  1. Is the target in the order form before you pay? Write down the percentage lift, CPA, ROAS or cost per qualified lead the vendor is promising. Do not accept a target in the pitch deck and a different description of the work in the contract. If payment does not depend on a stated result, call the fee what it is: a retainer. The useful question is not whether the vendor sounds confident. It is whether you can point to the same number on signing day and on the day performance is judged.
  2. What is the baseline period, and who pulled the number? A target tied to qualified leads, sales or lift over a baseline needs dates, spend and a source attached. Record the starting count before work begins. Otherwise, a good month can become the vendor’s baseline when they want a harder-looking win, while a bad month becomes the baseline when they want an easier one. You and the vendor should be able to recalculate the starting number from the same records without asking anyone what they meant.
  3. Does the metric measure sales and qualified pipeline, or clicks and form fills? Raw fills are cheap to inflate and painful for sales to work. If the fee depends on leads, define what makes one qualified and where that status is recorded. A vendor paid per lead has a reason to favor high-volume fills over buyers your sales team can close. This is where a tidy-looking performance deal can go wrong without anyone disputing the arithmetic: the count rises, the sales team gets busier, and the business has little to show for it.
  4. Which system is the referee: Google Ads, GA4 or your CRM? Name one system for payment, then state exactly what it will count. Google Ads data helps optimize campaigns, GA4 compares channels, and the CRM tracks pipeline and revenue; they will not necessarily agree. Attribution, lookback windows, counting rules and later qualification all change the total. GA4 can report fewer conversions than Google Ads for the same period, with the CRM different again. For a revenue deal, I would judge payment on the named CRM count, even if that number is lower. One referee beats three convenient scoreboards.

B. The fee: follow it when spend changes

  1. Is the formula a flat fee, a percentage of spend or a percentage of revenue? Put the calculation in the contract, not just the model’s name. Revenue-share deals can sit at 5–15% of attributable revenue, while hybrid arrangements put 10–30% of pay at risk. Those labels still leave the hard part unanswered: which revenue is attributable, which fee is fixed, and which amount changes when the target is hit or missed. Work one month through the formula using your own numbers. If two people get different invoices, the formula is not finished.
  2. What does a spend increase do to the vendor’s fee? At 15% of spend, the fee is $1,500 on $10k and $15,000 on $100k. It grows with the budget whether efficiency holds or not. A flat fee within agreed spend bands removes that particular push. Ask what happens when spend moves up, moves down or crosses a band: you should know the fee before approving the change, not after receiving the invoice. I do not object to a vendor earning more. I object to calling higher spend a performance result.
  3. What sits outside the headline price? Setup can run $1,000–$5,000, with landing pages, creative, tracking and CRM work often billed separately. Put each potential charge beside a price or a zero. This is where I have seen a performance proposal change shape: the statement arrives with a management fee, page builds and tracking hours. Run the quote against these seven pricing red flags. If the baseline and qualification rules in Section A are loose, a clean fee formula will not save you; it will only calculate the wrong result neatly.

C. The miss: give day 90 a consequence

  1. What exactly happens on day 90 if the target is missed? Choose the answer in writing: free work, a fee credit, more time at a stated cost, or nothing. Name the date, the amount and the condition that triggers the remedy. A promise to “keep optimizing” is not a remedy; it describes the work you thought you had already bought. If the vendor will not put a dated consequence beside a dated target, the target is sales language, not a payment term. This is the line I read before any guarantee headline.
  2. Are “free” months billed later or banked against future fees? Write one sentence showing the invoice after a miss and the invoice after a later hit. A credit that can only be used by staying for another term is not the same as money back. Neither is extra work if another charge appears beside it. You do not need a clever clause here. You need arithmetic a person outside the sales meeting can follow. If the vendor cannot show what you owe after missing the target, do not let the word free do the accounting.
  3. Which external events void the guarantee? List the specific events, such as seasonality, budget caps, offer changes, tracking breaks and approval delays, and say how each affects the target or deadline. Do not let a broad exception swallow the promise. The familiar mismatch is a proposal promising 40% more while the terms require 12 months of fees regardless of outcome. A performance escape clause around six months when targets are missed is a clearer alternative to a lock-in with no consequence. This is also why traditional pricing models can misalign incentives: the vendor gets paid on schedule while your result remains negotiable.

D. The exit: keep the work and the evidence

  1. Who owns the ad account, pixel data and landing pages? Specify ownership and access for each. Linking an agency’s manager account to a client-owned ad account is not the same as letting the agency control the account itself. You want the former, with access you retain when the relationship ends. The account-ownership trap is documented plainly; one owner who lost direct access also lost campaign history, keyword research and proof of performance at transfer. Check the setup rather than accepting “it’s your account” as an answer. Ownership is most expensive to negotiate after you have left.
  2. What happens to the term after the first commitment? Month to month is the answer I want. If the contract renews into another fixed term, write down when you can leave, how you give notice and whether missing the target changes either rule. A long lock-in with a vague exit belongs on the same warning list as guaranteed positions, refused admin access and impression reports with no cost-per-conversion line. You can like the team and still insist on an exit. The clause exists for the month when the relationship stops working, not the month when everybody signs.
  3. Can you see every change and the reason for it, weekly? Google’s Change History records who changed what and when for two years. It does not, by itself, explain why a change was made. Ask for a weekly log that connects actions to reasons and lets you compare the work with the target. A polished monthly summary is not a substitute: it may show the result while hiding the decisions that produced it. When a number slips, the log tells you whether the vendor changed course or merely changed the slide title.

Ink cartoon of a marketer examining a contract while money slips from the back page

  1. Who is the named human you call when the number slips? Get a name, a direct channel and a communication cadence that does not depend on you chasing a team inbox. I used to tell clients a shared inbox was fine if the work was good. I was wrong. On a miss, someone needs to explain the referee count, the spend, the change log and the next move on a call. A named owner does not replace the written remedy in item 8. They make it harder for that remedy to disappear into a queue.
  2. How soon does live work start? Put the start date in days, not in an onboarding promise with no date. Live work inside the first week separates an operator ready to begin from a seller with a backlog; a start four weeks out lets the baseline drift before the first change. That matters especially when the proposal measures a 90-day lift. Ask when the clock starts and what happens between signature and that date. You should not discover after paying that the measurement window started while access requests were still sitting in an inbox.

Three answers that should end the call

  1. “We guarantee position one, or 50 leads a month.” Nobody owns the auction, competing bids or your close rate. A fixed lead count without a qualified-lead definition is a way to buy form fills sales will not call. Do not spend the next meeting debating how impressive the number sounds. Ask for the qualification rule and the payment metric in writing. If they keep selling the count instead, thank them and leave.
  2. “You do not need admin access; we will send screenshots.” That is an ownership choice dressed as a process choice. Screenshots cannot give you your own access to Change History or an independent way to check what happens before renewal. Return to items 11 and 13: who controls the account, and how will you see and understand the work? If the answer still depends on their PDF, you will be negotiating your exit with evidence they chose for you.
  3. “We report on clicks and impressions; cost per conversion is in there somewhere.” If cost per qualified lead is the target and that line is missing, the missing line is the problem. Add a long lock-in and a vague exit, and the guarantee is weaker still. You need the same metric in the order form, the report and the miss clause. Otherwise the vendor can celebrate activity while you pay to find out whether it produced anything useful.

Put the answers beside the quote

  1. Fill the blanks in writing. Print the proposal or mark up the order form and answer all 15 items beside their numbers. I give a quote 20 minutes: 10 to fill the blanks and 10 to decide whether the blanks are the offer. Email beats a call here. A vendor who will not write down the baseline, referee and miss remedy can say all three on Zoom and forget them by invoicing day. If three or more answers come back as “we can discuss,” have that discussion before money moves.
  2. Run one miss through the contract. Say you spend $20k a month and the pitch is 40% more qualified pipeline in 90 days. The order form should state the baseline number, its date and source, the CRM count of sales-accepted leads used as referee, and what happens on day 90 if the lift is 12% instead. Then check the fee, your ownership of the account, pixel and pages, the term after the sprint, the weekly change log and the named person responsible. A clean answer gives the remedy a date and a dollar, not “we keep optimizing until you are happy.”

Paper checklist with the referee system circled beside three mismatched conversion counts

  1. Compare the quote with a clean offer, not another vague one. groas puts its Paid Ads and SEO sprint targets in writing before you pay: 40% more sales from ads or 40% more visibility in search inside 90 days, at one price paid up front. Use that specificity as the comparison point. Then apply this checklist to the terms in front of you, including the referee and miss remedy. A performance label is no substitute for those lines, whether it appears on a traditional agency proposal or a sprint.
  2. Check whether a lift target is sensible yet. If you close fewer than thirty qualified conversions a month, or tracking broke last quarter and nobody trusts the baseline, do not sign a lift-based deal yet. The sample is too small for a percentage to tell you much, and the starting number is too shaky to settle a dispute. Bank sixty to ninety days of clean CRM counts, then run the list again. Bring that baseline to the table yourself: the vendor should have to respond to your number rather than supply one after the fact.
  3. Last, return to No. 4: name the referee and the cost of leaving it blank. This is the line vendors skip most. “We will look at all platforms” lets them claim the win wherever it appears first: an Ads conversion, a GA4 attribution or a qualified person in the CRM. Put one sentence in the order form naming the CRM object used for payment, the date it will be pulled and the baseline it will be compared with. If the vendor will not sign that sentence, the other 14 answers do not matter. The cost is a bonus on a reported lift while sales works the same pile of leads. Definitions first, price second.