LinkedIn
Two agency proposals can promise the same CPA and leave you with completely different risks. One makes the agency answer for junk leads. The other lets it hit its target while your pipeline stays flat.
The price on page one matters. The definitions buried later matter more: what counts as a conversion or revenue, where the baseline starts, and how long poor results can be called learning. In a performance deal, those definitions decide whether the agency carries any risk. Here are 18 terms I would pin down before signing, in roughly the order a buyer meets them.
CPA
CPA is total ad cost divided by the conversions counted. It tells you what each conversion cost, not whether that conversion was worth buying. I used to report CPA as if it settled the argument. It does not. A cheaper drain-call booking that never turns into a job is still a cost, however tidy the spreadsheet looks.
In proposals, CPA gets stretched from both ends. The conversion count expands to include every form fill, short phone call or chat started. The cost shrinks to ad spend alone, with the management fee and landing-page bill left out. Ask which conversions count and which costs go into the calculation. If the agency will not list both in writing, its CPA target is decoration.
ROAS
ROAS is revenue divided by ad cost. It tells you how much attributed revenue comes back for each dollar spent. A low CPA with low order values can still lose money, while ROAS exposes that mismatch.
But a flattering ratio can hide a shrinking business. Cut spend everywhere except repeat buyers searching your brand, and ROAS may look efficient while revenue falls. Then comes the question of credit: a platform-reported sale is not necessarily a sale the ad caused. Ask whether the target uses all revenue attributed to ads or revenue the ads added. If that distinction is missing, the agency can take credit for customers already on their way to checkout.
CPA and ROAS targets
CPA and ROAS targets are two different constraints on how an account spends. Target CPA aims for conversions around a cost per action, while Target ROAS aims for conversion value around a return target. Neither metric is a substitute for the other, and an account can meet both. The trouble starts when a contract promises both without saying which takes priority when they pull in different directions.
I used to tell clients we could simply pursue both. I was wrong to make it sound that easy. A $45 CPA and 400% ROAS can encourage a narrow hunt for cheap orders and familiar buyers rather than growth. Ask which target governs when they conflict, and what happens to the fee when it misses.
Qualified pipeline vs lead
A lead is an inquiry; qualified pipeline is an opportunity sales has accepted as worth pursuing. Marketing may call a lead ready based on fit and engagement, while sales applies its own review of need, authority, timing and budget. That handoff is where a lead-based performance deal earns its name or falls apart.
A raw lead target is easy to improve by loosening the definition: count shorter calls, use broader forms, drop the budget filter. The report improves. Sales gets busier with people it cannot sell to. For the longer version of this contract rule, read what your search partner should get paid for. Ask who qualifies a lead, in which system, and within how many days.
Closed-won revenue
Closed-won revenue is the value of deals recorded as won in your CRM. It is not pipeline value, projected deal size or the value of every opportunity an ad supposedly touched. I have seen marketing celebrate sourced pipeline while finance looks at closed deals. Both numbers have a use. Only one represents a sale.
A proposal that calls pipeline revenue borrows the bigger number without delivering the thing it names. Ask whether the target uses closed-won dollars in the CRM and how long after the ad interaction a deal may count.
Baseline
Baseline is the result you compare performance against to judge what changed. If the agency gets credit for growth, the starting point can decide the size of the check before a single ad runs. For an established brand, many sales may happen without a particular ad. A comparison that ignores them makes ordinary demand look like agency-created growth.
Watch the period selected. A low starting period makes a later lift easier to claim; a period shaped by earlier advertising can confuse the comparison too. Say you spend $20k a month. A reported 40% lift tells you little until you know what it is measured against. Ask how the baseline is fixed, over what period, and who can recalculate it when seasonality shifts.
Incrementality
Incrementality is the result ads caused beyond what would have happened without them. Attribution assigns credit to touchpoints; incrementality asks whether the conversion depended on the ad. That is why a ROAS screenshot cannot settle an incrementality argument.
Take a shopper already set on buying who clicks a brand ad at checkout. The platform can credit the ad even if the shopper would have bought anyway. I used to defend brand ROAS as proof of skill. It proved, at least in part, that customers could spell the brand name. Ask whether payment follows reported conversions or incremental ones, and what comparison will test the difference.
Attribution window
An attribution window is the time after an eligible ad interaction during which a conversion can be credited to it. Google Ads allows different windows for click, engaged-view and view-through conversions. Change the window and you may change the report without changing customer behavior.
Extend a click window from 30 to 90 days, for example, and later purchases can enter the count. Reported CPA may fall even though no additional customers appeared. Calling that a measurement upgrade does not make the old and new numbers comparable. Ask for the click and view-through windows for each conversion action in writing.
Learning phase
The learning phase is the period after a significant bidding change when the system adjusts and results may swing. Its duration depends in part on conversion volume and the sales cycle. The mechanism is real. It is also a remarkably convenient explanation for a bad month.
I have used still learning, give it time after a rough week. Sometimes that was fair. Repeated changes, though, can keep the account unsettled, and an open-ended learning period gives the agency no date to answer for results. For the diagnostic I use now, see when your agency is making excuses. Ask what signals the end of learning, when you expect to reach it, and what the agency will change without an extra charge if you do not.
Retainer
A retainer is a fixed recurring payment for an agreed scope of management. Quoted monthly fees vary with the operator and the complexity of the account. Because the invoice does not rise automatically with spend, the agency has no fee-based reason to push your budget up. The tradeoff: management can mean substantial work or one monthly report and a few bid tweaks.
Say you pay $2,000 a month. A list of tasks and meetings tells you what labor to expect; a stated outcome tells you what the work is meant to achieve. You may need both. Ask what happens each week for the fee and what triggers an extra charge.
Percentage of revenue
Percentage of revenue pricing gives the agency a share of sales attributed to its work, with or without a retainer. The definition of revenue does most of the work here. Gross sales or sales after refunds? Brand-search purchases included or excluded? Which attribution window applies? Each choice changes the payment.
Because the agency earns on counted sales, it has an incentive to pursue sales the measurement system recognizes quickly. That may favor existing demand over harder-to-measure growth. If margins are thin or the sales cycle is long, inspect that mismatch especially closely. Ask which revenue counts, where it is measured, what is excluded, and who audits it.
Cost-per-acquisition pricing
Cost-per-acquisition pricing charges a set amount for each agreed conversion rather than for hours or ad spend. At $20k in monthly ad spend, a 15% spend fee is $3,000 regardless of the result. At $60 per qualified lead, 40 accepted leads cost $2,400 and 60 cost $3,600. That changes what the vendor has to deliver to get paid.
It does not solve quality by itself. If a duplicate form fill counts as an acquisition, the vendor can earn more while sales gets nothing useful. For the worked comparison, see who charges on CPA instead of ad spend. Ask which event triggers payment, who verifies it, and what happens to duplicates, cancellations and rejected leads.
Flat fee
A flat fee is one fixed monthly payment regardless of ad spend. There is no percentage increase when your budget moves from $10k to $30k. I moved clients away from percentage billing because the larger invoice had stopped reflecting the work. A fixed fee removes that particular incentive to push spend; it does not define the service for you.
The bundle is the stretch. A $1,999 fee including management, landing pages, tracking fixes and reporting is different from a $1,999 fee that charges separately for each. Ask what the exact amount includes, what it excludes, and what can raise it.
Setup fee
A setup fee is a one-time charge for work done before ongoing management begins. Builds, landing pages and tools can appear as separate costs. Good setup work can be worth paying for. So can importing an old account structure and changing the logo, if you are the one sending the invoice.
Because the fee arrives before results do, you need something more concrete than strategic foundation on a proposal slide. Ask which assets you own after setup, where they will live, and what gets corrected without another fee if tracking is wrong.
Performance guarantee
A performance guarantee is a written promise of what the agency owes you if an agreed target is missed. A refund, reduced fee and free additional work are not interchangeable. Nor does refunding a management fee return ad spend already spent.
I read guarantees from the back first: conditions, exclusions, measurement and payout. That is where the apparent risk transfer often disappears. Ask what exact miss triggers the guarantee, who measures it, and whether the remedy is cash, a fee reduction or more work. Without a number, a date and a remedy, guarantee is sales copy.
Target in writing
A target in writing is an agreed outcome with a baseline, deadline and measurement system named in the contract. Not just in the deck. I used to tell clients a target on a slide counted. I was wrong. Slides change; the signed agreement decides what either side can demand.
A target such as 40% more qualified pipeline from a fixed baseline in 90 days, measured in the CRM, gives both sides something to inspect. More leads does not. Ask to see the target sentence exactly as it will appear in the agreement. If the agency cannot write it down, you are buying hopeful language, not a performance deal.
Lock-in and account ownership
Lock-in is the difficulty of leaving; account ownership determines who controls the ad account and its history when you do. An account created through a manager account can have different ownership arrangements from an existing account linked to one. Do not assume that paying for the work gives you the access needed to keep using it.
The practical risk is a messy exit: lost access, a disputed handover or a rebuild of work you already paid for. I check ownership at the start, not when the relationship has soured. Ask whose account it is, whether you have full admin access from day one, and exactly what transfers when you part ways.
Percentage of ad spend
Percentage of ad spend is a management fee calculated as a share of your media budget. Agencies commonly quote rates around 10% to 20%, sometimes with tiers and a monthly minimum. When the budget rises, the fee rises too. That does not prove the agency gave bad advice. It does mean every recommendation to spend more also pays the person making it.
I billed this way for years. Managing $40k does not automatically require twice the work of managing $20k, yet the invoice can double. Ask what additional work a higher budget buys and what happens to the fee if efficiency drops.
This is the term I would retire. It prices the budget instead of the work or the outcome. Strike it from the contract, and the other definitions finally have to do their job.