Short answer: a high CPM is a price, not a diagnosis. If purchases are arriving at a cost per purchase you can live with, the auction is finding buyers and the CPM is what reaching them costs. If purchases are not arriving, the CPM is a symptom of a setup that boxed you into an expensive corner. Forcing it down by loosening targeting or switching objectives usually buys cheaper people who do not buy, so read cost per purchase first and CPM second.
Two advertisers we talked with recently had the same complaint from opposite directions. One ran purchase-optimized campaigns to a huge audience and watched CPM climb; their instinct was to loosen targeting or switch objectives to bring it down. The other paid seventy to eighty dollars per thousand impressions and assumed their creative was the problem. Both were treating CPM as a dial to turn. It is not — it is a thermometer, and the reading only means something once you know what is being measured.
CPM is the price of the specific people Meta is currently bidding on for you. Optimize for purchases and the auction hunts for likely buyers — a small, expensive, heavily contested slice. Optimize for clicks or views and it hunts for people who click and watch, who are cheap precisely because everyone can have them. A high CPM on a purchase campaign is often the system doing its job; a beautifully low CPM is often the receipt for an audience that will never buy. The number that pays your bills is cost per purchase, and the two frequently move in opposite directions.
The question that sorts it: expensive people, or an expensive corner?
There are two honest reasons a CPM reads high, and they need opposite responses. Reason one: you are buying valuable people. Purchase optimization, competitive vertical, strong signal — the auction charges more per thousand because each thousand contains more buyers. If cost per purchase is acceptable, this CPM is not a problem, and every lever that lowers it will quietly swap your buyers for spectators. Reason two: your setup boxed the auction into a small, overpriced pool — an audience narrowed to a sliver, a brand-new dataset asked to find purchasers with no purchase history to learn from, or delivery concentrated in the most expensive placement and country. That one is worth fixing, not because the number offends you, but because the pool is genuinely too small to price efficiently.
Split it before you judge it
The account-level CPM is an average that hides the diagnosis. Break it down twice in Ads Manager: by placement, and by country. A blended $75 CPM often turns out to be one expensive placement or one premium market dragging the average, with the rest priced normally — which changes the conversation from “my ads are expensive” to “this one slice is expensive, is it earning its price?” Check each slice's cost per result before cutting: the priciest placement is sometimes also the one producing the purchases.
The fixes that lower CPM without buying tourists
If the breakdown points at a boxed-in setup rather than valuable people: widen the audience and let the delivery system do the narrowing — a purchase-optimized campaign with a broad audience usually finds cheaper paths to the same buyers than your hand-drawn sliver allows. Give a new dataset an event it can actually learn from before asking for purchases, so it is not bidding blind in the most contested corner of the auction. And let placements breathe before pruning them. What not to do: switch the objective to traffic or engagement to enjoy the lower CPM. The reading on the thermometer will drop, the people arriving will click generously, and the purchases column will explain — slowly and expensively — what those impressions were actually worth.