If your account sells more than one thing, it needs more than one CPA target. Set a single fallback target at the account level, then override it on each campaign with a number built from that campaign's own margin. Group campaigns by role, and judge each group on the KPI that fits its job.
That is the whole idea behind CPA targets per campaign. The rest of this post is how to do it without turning your account into a pile of numbers nobody trusts.
Why one CPA for the whole account breaks
Say an account runs two campaigns. One sells a high-ticket product. The other sells an accessory. Both report conversions. A blended CPA of, say, $120 looks fine on a dashboard.
It is wrong for both campaigns. (Every number in this post is made up for illustration.)
- The high-ticket campaign could afford a CPA of several hundred dollars. At $120 it is being starved, and you are leaving profitable volume on the table.
- The accessory campaign might break even at $50. At $120 it is losing money on every sale, and the blend hides it.
A blended target pulls both campaigns toward a number that fits neither. Smart bidding does exactly what you tell it. If you tell it $120 everywhere, it will chase $120 everywhere.
The fix is not complicated. Each campaign gets the target its economics support.
Step 1: Find break-even CPA for each campaign
Break-even CPA is the most you can pay for a conversion and still not lose money. Everything else is built on it.
For sales
Take revenue per order, multiply by gross margin, and you have profit per order before ad cost.
Example (illustrative): Say a campaign sells a $1,000 product at a 50% margin. Profit per order is $500. Break-even CPA is $500.
Another example (illustrative): Say a second campaign sells a $100 accessory at the same 50% margin. Profit per order is $50. Break-even CPA is $50.
Same account, same margin percentage, a tenfold gap in what a conversion is worth. That is the case for CPA targets per campaign in one line.
For leads
A lead is worth profit per sale times the share of leads that become sales.
Example (illustrative): Say a closed sale earns $1,000 in profit and one lead in four closes. A lead is worth $250. Break-even CPA is $250.
If the client has several lead types, such as a quote request and a newsletter signup, they should not share a target. Count the ones that matter to revenue, and give the rest a looser goal or leave them out of the bid signal.
Step 2: Set the target below break-even
Break-even is a ceiling, not a goal. The target should sit under it, and the gap is the profit.
Example (illustrative): For the $500 break-even campaign, a target CPA of $350 leaves $150 of profit per order. For the $50 break-even campaign, a target of $35 leaves $15.
How much room to leave depends on what the client needs. Some want maximum volume at thin profit. Others want profit first. This is a business decision, so ask for it instead of guessing. A target you cannot explain to the client is a target you will end up defending at the worst moment.
Two checks before you commit:
- Is there enough conversion volume? A campaign with only a handful of conversions a month gives bidding very little to learn from. A tight target on thin data tends to throttle delivery. Start looser, or group the campaign with similar ones.
- Is the conversion the right one? If the campaign optimizes toward a low-value action, no target will rescue it. Fix the signal first.
Step 3: Account level is the fallback, campaign level is the rule
The question comes up constantly: should target CPA live at the account or the campaign? Both, in layers.
- Account level: one fallback target. It exists so a new campaign never launches with no goal. It should be a number you are comfortable with for an average campaign, and it should be conservative.
- Campaign level: an override wherever a campaign's economics differ from the average. Most campaigns that matter will end up here.
- Group level: a KPI for a set of campaigns that do the same job.
The order of precedence is simple. The campaign number wins over the group, and the group wins over the account.
Keep a short sheet for each account listing the fallback, every override, and one line on why each number is what it is. When a client asks why a campaign is set at $35, the answer should take ten seconds to find.
Step 4: Group campaigns by role
Not every campaign should be judged on CPA. Roles differ, and the KPI should follow the role.
Prospecting
Prospecting finds new people. Many of them will not convert on the first visit. Judging these campaigns on last-click CPA often makes them look bad even when they are filling the funnel for the campaigns below.
A fairer yardstick is something earlier in the chain. For TikTok prospecting, for example, you might judge every campaign on link cost per click, because that is the signal the platform actually delivers at that stage. Set one rule for the whole group instead of one target per campaign.
Retargeting
Retargeting talks to people who already know the brand. It should convert cheaply, and a CPA target here can be tight. It also has a ceiling on volume, so do not read cheap CPA as proof you should push more budget in. There is only so much of an audience to retarget.
Brand search
Brand search is usually the cheapest conversion in the account. It is also the easiest place to claim credit that belongs to other campaigns. Give it its own target and its own reading, and keep it out of the blended number you show the client.
Non-brand and shopping
These tend to carry the real profit conversation. This is where the per-product margins from Step 1 matter most, and where overrides pay off.
Writing the role down next to each campaign is worth the five minutes. It turns "why is this campaign at that number" into a question with a written answer.
Step 5: Keep targets honest as margins move
A target is only correct on the day you set it. After that, it drifts.
Things that quietly make a target wrong:
- Margin changes. A supplier raises prices or a discount becomes permanent. Break-even drops and the target does not.
- Offers start and end. A promotion lifts conversion rate for two weeks. If you set the target in the middle of it, you set it on a number that will not last.
- Close rate moves. Sales changes how it follows up, lead quality shifts, and the value of a lead moves with it.
- Product mix changes. The cheap product starts outselling the expensive one inside the same campaign.
Review on a schedule, and also whenever one of those events happens. Change one thing at a time, note the date, and give it enough days to read before you judge it. A target change made in reaction to three bad days mostly teaches the algorithm to flinch.
A quick checklist you can use tomorrow
- List every campaign in the account and write what it sells or what lead it chases.
- Work out break-even CPA for each, using margin and, for leads, close rate.
- Set a target below break-even for each, sized to the profit the client needs.
- Set a conservative account-level fallback.
- Tag campaigns by role and pick a KPI for each role.
- Write one line of reasoning beside every override.
- Put a review date on the calendar, plus a trigger list: margin, offer, close rate, mix.
Frequently asked questions
How do you set different CPA targets for different campaigns?
Work out break-even CPA for the product or lead type each campaign sells, then set the target below it with room for profit. Enter that number on the campaign, not the account. Anything without its own number falls back to the account target.
Should target CPA be set at the account or campaign level?
Both, with different jobs. The account target is a fallback so no campaign runs with no goal at all. Any campaign whose margin, lead quality or funnel role differs from the average should carry its own target that overrides the fallback.
How do you work out a target CPA?
Start with profit per conversion: revenue per sale times margin, or for leads, profit per sale times close rate. That is your break-even CPA. Set the target at a share of it that leaves the profit the client needs.
Can top-of-funnel campaigns have a different KPI?
Yes. Prospecting often should not be judged on CPA at all. Link cost per click, cost per landing page view or a looser CPA is a fairer test, as long as you review how those campaigns feed the ones closer to the sale.
How often should CPA targets change?
Review them when margin, pricing, offers, close rate or product mix change, and on a fixed schedule even if nothing seems different. Do not change a target in reaction to a few days of data. Make one change at a time so you can see what it did.
Where Ads Arsenal fits
Ads Arsenal is built around this layered way of working. You set a target for the account as a fallback, then a CPA or ROAS for any campaign where it matters, and you can set rules for groups of campaigns, such as judging every TikTok campaign on link cost per click. You also write how you run the account, and what the system must not touch, in plain English, and it reads every account through that lens. It runs across Google, Meta and TikTok from one login.
It works from your numbers, and it works with your approval. Nothing changes in an account until someone approves it. If it suggests something that does not fit how you run the account, dismiss it with a sentence saying why, and that becomes a rule for that account. You can take the rule back later. If you manage accounts for clients, see Ads Arsenal for agencies.
Try it on a real account
Ads Arsenal is in a free design-partner beta for agencies and in-house teams running real spend on Google, Meta and TikTok. If you want to see how per-campaign targets and group rules work on your own accounts, Become a design partner.