The label “Facebook ads for online stores” actually covers ads across the whole Meta family: Facebook and Instagram, with their feeds, stories and short videos. For clothes, footwear, make-up or skincare, this is the natural place to be: people do not search for the product, they see it, like it and only then find out they want it.
That changes the work. In search, the winner is whoever answers a question best; here, the winner is the image that stops the thumb mid-scroll. The account settings matter less than people think, and the creative and the measurement matter more.
A good-looking store, with ads that look like everyone else’s
The starting point is much the same from one store to the next. There are product photos on a white background, good for the website, and a few posts boosted now and then. The ads use the same photos, with a line about discounts, and are shown to audiences chosen by interests such as “fashion” or “beauty”.
The result is hard to read. The platform reports one set of orders, the store sees another. Nobody knows which ad brings anything in, because they all start at once and are changed before they have gathered any data.
Facebook ads for online stores: the catalogue and product sets
The catalogue is the store’s product list, sent to Meta through an automatically updated feed. It is the source of dynamic ads: the platform chooses for itself which product to show each person, based on what they looked at on the site or on what people similar to them buy.
In fashion, the trap is stock by size. A style that has sold out in the sizes people want keeps being promoted if the feed lists it as “in stock”. In cosmetics, shades and pack sizes have to appear as variants of the same product.
- Product sets: groupings from the catalogue (new collection, bestsellers, clearance), each with its own ad and budget.
- Exclusion rules: products left only in rare sizes, or too cheap to bear the cost of advertising, are taken out of promotion.
- Up-to-date prices: discounts on the site reach the feed immediately, otherwise the ad shows a different price from the page.
Your own photos and footage, tested one at a time
Stock imagery does not sell clothes. What works is what looks as if a person made it: the dress worn and filmed in motion, the texture of a cream spread on a hand, a shopper showing how she combines the pieces (with her written consent). Short vertical videos, with the product visible from the first second, suit the way people watch on a phone.
Testing has one simple rule: one change at a time. The same clip with three different openings, or the same image with three texts. Each variant gets enough budget and enough time for the difference not to be down to chance.
The raw material comes from you. We agree a realistic rhythm for shoots, because a good ad wears out: after a while, the same people have seen it too often.
Broad audiences or interest-based audiences
For a long time, the craft meant choosing the right interests. Today the algorithm finds buyers mainly from signals: who stops at the ad, who adds to basket, who pays. A broad audience gives it room to look, and the ad itself does the sorting: an evening dress draws its own audience.
Interests remain useful in a few situations: a new account with no order history; a small budget; a niche product. We usually start with both options side by side and keep whichever brings orders at a better cost. Separately, we keep a campaign for people who have already visited the store, excluding recent buyers.
Pixel, Conversions API and the visitor’s consent
Without correct measurement, the platform optimises blind. The pixel is the code on the site that tells Meta when someone views a product, adds it to the basket or buys. On its own it is no longer enough: some browsers and ad blockers stop it, and some of the orders disappear from the reports.
Conversions API sends the same events directly from the store’s server. The two sources are paired through a shared identifier, so that an order is not counted twice.
- Complete events: product viewed, add to basket, checkout started, order with value and currency.
- Consent: events for advertising purposes are sent only for visitors who have accepted; sending from the server does not get round a refusal, and the details are settled with your data protection officer.
- Comparison with reality: the figures in Meta are set beside the orders in the store and beside returns, which the platform does not see.
- Product claims: promises of results and “before and after” images can lead to rejections, especially for products that promise changes to the skin or the figure, and Meta’s rules are checked at the time of the campaign; what you are allowed to claim about a cosmetic product is checked with a regulatory specialist.
Frequently asked questions
What budget do I need for ads on Facebook and Instagram?
It depends on the average order value, the margin and how many ad variants you want to test in parallel. The algorithm learns from orders, so the budget has to allow for a reasonable number of orders in each campaign.
Why don’t the orders in Meta match the ones in my store?
The platform also claims orders from people who clicked on the ad a few days earlier or only saw it, and the store counts differently. On top of that come the visitors who refused cookies. What matters is to follow the same source over time.
Can we use the photos we get from the supplier?
They can be used in the catalogue, if you have the right to use them, but the same photos probably appear at other stores too. The ads that stand out are the ones with material you made yourself, however simple.
This is a typical project description: it shows how we usually approach this kind of work and does not present a project carried out for a particular client. Every real project starts from your company’s situation, and the stages, timescales and price are agreed after the initial discussion.