How Value Optimization works
A normal purchase campaign treats every order the same. It is told to produce purchases, so it buys the cheapest purchases it can find, and a small order counts exactly as much as a large one. Value optimisation changes the instruction: predict how much each person is likely to spend, and bid more for the ones predicted to spend more.
For that to be possible, your purchase events must carry the amount and the currency, not just the fact that a purchase happened. Meta learns from those amounts, so the numbers you send are the whole basis of the prediction. It also needs a reasonable flow of purchases with values attached before the prediction is worth anything, which is why the setting is not available or not sensible in every account.
Once it is running, the campaign will happily pay more for one customer than another, and the cost per purchase can rise while the campaign is doing its job. Revenue against spend, not cost per order, is the measure that tells you whether it worked.
Why it matters
It matters wherever order sizes differ. A clothing shop where one customer buys a single item and another fills a basket, a travel business where a day tour and a trekking package sit in the same catalogue, an electronics retailer with accessories beside laptops — in all of these, counting orders hides the customers who are actually worth having.
It also changes how creative and landing pages should be judged. If the system is chasing value, adverts that attract browsers looking for the cheapest item will be quietly discouraged, and the products you feature start to matter as much as the audience you choose.
Where it goes wrong
Sending the wrong value is the most damaging error, and it is silent. Values that include shipping and tax, or that ignore discount codes, teach the system a version of your business that is not true. Mixed currencies passed without the currency field do the same. The system will optimise faithfully towards whatever you sent.
Cash on delivery makes this sharper for advertisers in Nepal and similar markets. If an order fires a purchase event with its full value when it is placed, cancellations and refused deliveries are still counted as revenue, and the campaign learns to find more of the customers who cancel. The other frequent mistake is switching to value optimisation with barely any purchase history, then judging it over a short window while predictions are still unreliable.
How to act on it
Fix the data first. Confirm that every purchase event carries a value and a currency, that the value is the revenue you actually keep from that order, and that it is deducted or corrected when an order is cancelled or returned. Check it against your own sales records rather than trusting the platform view, since the two rarely agree exactly.
Then judge it properly: run it long enough for predictions to settle, compare revenue against spend rather than cost per order, and keep an eye on margin as well as return on ad spend, because a campaign can buy expensive high-revenue customers on products that earn you very little. If order values genuinely vary in your shop, this is one of the more worthwhile changes to make in ecommerce advertising.
