Why chasing a lower CPA can reduce your profit
Optimising strictly for a lower cost per acquisition shifts delivery toward bargain hunters, lower-intent traffic and cheaper entry products. Cost per order falls, and average order value and contribution margin usually fall with it — leaving less absolute profit per order than before the optimisation, on a customer worth less over time.
The mechanism
An ad platform optimising for cost per acquisition is solving exactly the problem it was given: find the cheapest available conversion. It does that well. The trouble is that the cheapest available conversion is not a random sample of your buyers.
Cheap conversions cluster around people who were going to buy anyway, people buying your least expensive product, and people who respond to discounts. Each of those converts at a lower cost and is worth less. The platform reports success, because the metric it was optimising did improve.
Optimising for a lower cost per acquisition changes the audience being bought, not the efficiency of buying it. The cheapest conversions concentrate among discount-responsive buyers, low-intent traffic and cheaper entry products. The cost per order falls because the orders themselves are smaller, not because acquisition became more efficient.
This is why the improvement so often fails to appear in the bank account. The campaign got cheaper per order and the business got poorer per order at roughly the same rate.
CPA is not the same thing as a CAC ceiling
CPA is what you spend; a CAC ceiling is what you can afford.
CPA (Cost Per Acquisition): An observation. It’s the actual cost to acquire an order today.
CAC Ceiling: A constraint. It’s the absolute maximum you can pay per customer based on your contribution margin before you lose money on the order.
Driving CPA as low as possible isn’t always a win. A CAC ceiling is a limit, not a target. Once your CPA is comfortably below that limit, pushing it even lower often forces you to target lower-value customers—buying a worse customer just to hit a lower number.
What an ad platform reports when you ask it for a lower cost per acquisition — and the number nobody puts next to it.
Down 38%. The platform reports success — this is the metric it was given.
Down 34%. Nobody asked for this — the orders themselves got smaller.
Down 59%. The number that actually matters, and it moved furthest.
Cost per order down with order value flat is a genuine efficiency gain. Cost per order down with order value down is a change in who is being acquired — cheaper conversions cluster among discount-responsive buyers and cheaper entry products. Set the ceiling from contribution margin, then optimise for volume beneath it rather than for distance below it.
How to tell the two apart
Look at cost per acquisition and average order value together. If cost per order falls while average order value holds, that is a genuine efficiency gain. If both fall together, the audience has shifted toward cheaper buyers and contribution per order is probably falling too. The pair moving in the same direction is the signal.
Cost per acquisition on its own tells you almost nothing. Put one more metric beside it and the same movement resolves into four separate stories — two worth having, two worth stopping.
Genuine efficiency gain. Better creative, better targeting, or a better landing page — the same buyers acquired more cheaply.
You are buying cheaper customers, not buying more cheaply. Check contribution per order before celebrating.
The same problem, surfacing months afterwards. Discount-acquired buyers repeat less, so the damage shows up in the cohort rather than the campaign.
Paying more for a bigger order is frequently correct — provided it stays under the ceiling. This is the row that gets campaigns switched off by mistake.
Cost per acquisition is a measurement, not a verdict. Read it against average order value, contribution per order and repeat rate before acting on it — and hold campaigns to the ceiling your margin supports rather than to the lowest cost the platform can find.
What to optimise for instead
Don’t chase the lowest CPA—maximize volume under your margin ceiling.
Set the ceiling first: Calculate your CAC ceiling from your contribution margin. This acts as your safety limit, keeping campaigns profitable.
Focus on scale, not savings: The goal isn't to get as far below the ceiling as possible—it's to capture as many orders as you can while staying under it.
Measure order profit, not cost: Evaluate acquisition by contribution profit per order, not cost per order. This prevents you from sacrificing high-value customers just to report a lower CPA.
Set your ceiling once, enforce it as a hard boundary, and measure success by total profitable orders won beneath it.
Does this apply to automated bidding?
It applies more, not less.
Automated bidding algorithms hit their target far more efficiently—and ruthlessly—than a human ever could.
If you set a Target CPA too low, the algorithm will ruthlessly find the absolute cheapest clicks, steadily shifting your budget toward lower-quality customers.
Set your Target CPA using your contribution-margin ceiling, not last quarter’s average cost. This gives the algorithm room to bid on higher-value audiences while strictly guarding profitability.
When a falling CPA is genuinely good news
A falling cost per acquisition is a real gain when average order value, contribution per order and repeat rate hold steady while it falls. That pattern points to better creative, better targeting or a better landing page — the same buyers acquired more cheaply. The warning applies to CPA reductions bought by changing who converts, not to efficiency improvements.
What this does not tell you
It does not tell you what your ceiling is — that comes from the margin walk, and it is specific to a business's cost structure.
It also does not mean expensive acquisition is good. There is a real ceiling, and campaigns above it lose money on every order regardless of how attractive the customer looks. The argument here is against treating below the ceiling as a direction to keep travelling in, not against efficiency.
And for businesses with genuine repeat purchase, a first-order CPA understates what a customer can support. That calculation needs a measured repeat rate, not an assumed one — most shops claiming strong repeat behaviour have never checked the cohort.
