Which Meta campaign objective should an online store use?

This is the whole principle, and it explains most disappointing campaigns. Delivery is a prediction engine pointed at one outcome. Point it at link clicks and it finds people who habitually click, which is a real population and not the one that buys. Point it at purchases and it finds people who habitually buy.

The temptation to choose something else usually comes from cost. Clicks look cheap, purchases look expensive, and a campaign optimised for traffic reports a very low cost per click. What it does not report is that those clicks convert at a fraction of the rate, which is only visible once you divide spend by orders.

So the default is the sales objective with purchase as the optimisation event. It is also the setup that makes the campaign's own reporting comparable to your store's, since both are counting orders.

The exception is volume. If your store produces too few purchases for the system to learn from, purchase optimisation will deliver slowly and erratically, and the answer to that is a temporary compromise rather than a different objective.

The objective determines which formats, placements and buying options are available. The optimisation event determines who the campaign tries to find. Choosing the sales objective and then optimising for a page view produces a campaign that behaves like a traffic campaign wearing a different label.

When purchase volume is too thin to learn from, step one event earlier: optimise for initiate checkout, or add to cart if checkouts are also scarce. Those events happen more often, so the system gets more examples, and they still correlate with buying intent.

Treat that as temporary. An event earlier in the funnel is a proxy, and proxies drift: the campaign will find people who add to cart readily and buy rarely if you leave it there indefinitely. Move back to purchase optimisation once the volume supports it.

Also keep the event definitions clean. If add to cart fires on a quick-add from a collection page and again on the product page, the count is inflated and the optimisation is aiming at noise. Audit what each event actually means before you optimise towards it.

Awareness earns its keep during a real launch, a new market entry, or a category where nobody yet knows the product exists. It buys reach and frequency rather than response, and it should be measured on branded search demand and on the performance of your other campaigns rather than on its own return.

Leads fits stores that quote rather than sell online: made-to-measure goods, trade accounts, large equipment, anything with a consultation in the middle. Here the form is the conversion and the store's checkout is not part of the journey.

Catalogue-driven retargeting sits inside the sales objective rather than as a separate choice, and it remains one of the most efficient things a store with a connected catalogue can run, provided the audience is capped so it does not simply re-buy people already returning.

Engagement and traffic have almost no place in a store's account. Both optimise for behaviour that does not correlate with buying, and both produce reporting that looks encouraging while the orders sit flat.

Split acquisition from retention before you split anything else. A campaign that can spend freely on people who already buy from you will do so, because they are the cheapest conversions available, and the account will report efficiency while the customer base stops growing.

Use whatever controls the campaign type gives you: an existing customer budget cap, exclusions built from your customer list, or separate campaigns. Then report on new customers separately, using your own order data rather than the platform's classification alone.

Keep retargeting proportionate. A large share of budget aimed at people who visited last week produces a flattering return and very little growth, and it is the single most common way a store's advertising quietly becomes a subsidy for orders it was already getting.

Then let the creative do the segmentation the targeting no longer does. Different concepts for different objections will reach different people through delivery, which is closer to how the system works now than building audiences by hand.