Facebook Ads for Orders, Not Reactions: How We Set Up a New Account

A client sent us a screenshot of a post that had hundreds of reactions and asked us why it wasn't selling. The post was a holiday greeting.

This isn't a dig at them — the confusion is natural because the platform shows you exactly the metrics that make you feel good. Orders are elsewhere.

First week: setup and verification

The Pixel and the Conversions API, both. The pixel alone misses some events, especially on iPhone. We manually check each event: product view, add to cart, initiate checkout, purchase. We place a test order and confirm it appears with the correct value.

The catalogue, if it's a shop. This is where projects most often get stuck: a product feed with no stock, old prices, missing images. We lost a week with one client because the feed generated once a day, at a time when stocks weren't updated. We redid the integration from scratch.

The structure we start with

One sales campaign, two or three ad sets, four or five creatives. No more. Dividing the budget into too many ad sets means none of them get out of the learning phase.

Audiences: a broad one, without interests, left to the algorithm. A remarketing one for visitors and abandoned carts. A lookalike audience, built on buyers, if we have enough.

We're using interest-based audiences less and less. Not because it's a trend, but because for our clients, the broad audience has performed comparably or better, with less maintenance work. How big the difference is depends on the niche.

The creatives

This is where the campaign is decided. Not in the settings.

We test different messages, not graphical variations of the same message. An ad showing the product being used, one answering an objection ("can be returned within 30 days"), one with the price, one filmed on a phone without editing.

The last category often surprises us. Low-quality but concrete video sometimes beats material made with lights and a storyboard. Not always. Often enough to always test one.

What we don't track

Reactions, saves, shares, reach. We see them, but we don't make decisions based on them.

We look at the cost per acquisition and the value of orders compared to spending, compared to what we know from accounting, not just from the platform. The figures in the interface are optimistic; we measure the difference monthly, overall, against real sales.

A mistake we make less often now

We stopped ad sets too quickly, after two or three days, because they seemed weak. Sometimes they settled down after a week. Now we decide at the beginning how long each ad set will run and how much we're willing to lose before making a decision — written down, so we don't change our minds under pressure.

When we don't recommend Facebook Ads

Very expensive services, with a long decision cycle and a restricted audience. Search works better there, and we say this from the first discussion, even if it means one less service sold.

If you want us to check if your pixel is sending what it should, we'll look at it together and tell you in 15 minutes.

What an account looks like for a small business, in practice

Complicated structures sound good in presentations and quickly fall apart when the budget is small. What we keep, at the start, is as minimal as possible.

What we record as a result

Only what matters to the client: the submitted inquiry, the phone call, the order. We don't set a page view or scrolling as a result, no matter how good it might look in the report. We inherited accounts where the "conversion" was a visit to the contact page — the cost per conversion seemed excellent, and no one was calling.

If the report looks good and the phone isn't ringing, the first suspicion is what you've called a conversion.

When to stop and when to be patient

We stop a creative if it has spent enough and produced absolutely nothing. We are patient if it produces, but more expensively than we'd like, because often it settles down. We don't have a magic number for "enough"; it depends on the product's cost and the daily budget.

A mistake we made twice

We changed the audience and the creative on the same day, to save time. The result changed, and we couldn't tell why. Now we change one thing at a time, even if it's slower and the client is impatient.

What depends on the niche

For impulse products, results are seen in days. For expensive services, the decision can take weeks, and monthly reporting looks unfair. In these cases, we also track inquiries that close later, with the client's help, because the platform alone doesn't see them.

If you want an opinion on your current account, give us read access and we'll tell you what we'd simplify first.