Advantage+ Sales, Andromeda, the new attribution and AI-generated creative. What it means for your webshop, how to structure your account and a budget calculator.

If you started running Facebook and Instagram ads five years ago, you probably remember detailed interests, 1% lookalike audiences, dozens of ad sets and A/B tests on every segment. Almost none of that matters in 2026, and in 2027 it will matter even less. Meta has moved nearly every decision into the algorithm: who sees the ad, where, with what budget and, more and more often, with which creative.
That doesn't mean your role disappears. It changes. In this article I show you what has already changed in Meta Ads, what's coming next, what a webshop selling in Denmark should do differently and how to structure your account based on your budget. There is also an interactive calculator that tells you what structure your budget allows. Examples use DKK.
To understand where Meta is heading, it's worth looking at what happened over the last two years first. Many webshops still run structures built for rules that no longer exist.
Advantage+ Shopping campaigns were replaced by Advantage+ Sales, and the old ASC campaigns can no longer be created. In practice, every sales campaign now starts with automatic targeting, automatic placements and a campaign-level budget. You can still restrict some settings, but every restriction takes optimisation room away from the algorithm.
Meta's recommendation system, called Andromeda, chooses from millions of ads the right one for each person, based on what the ad contains. A video of a parent dressing their child in the morning reaches different people than a video about the technical fabric of the same jacket. The interests you choose matter less and less. What you show in the ad matters more and more.
Since March 2026, Meta has changed how it reports conversions. Click-through attribution now counts only link clicks. Interactions like likes, shares and saves are reported separately in a new engage-through column with a one-day window. In short: if you only look at the main column, many accounts appear to have fewer conversions than before, even though real sales haven't changed.
Campaign-level budget, Advantage+ campaign budget, is the default. Even when you exclude placements, Meta may spend a small share of the budget, up to 5%, on excluded placements if it expects better results. Some placements have disappeared entirely, like Instagram Explore Feed and Messenger Stories, and Threads has become a standard placement.
Ads Manager can generate backgrounds, expand images to other formats, write text variations and create short videos from static images. Some of these options are switched on by default, so it's worth checking what you publish, especially if your brand has a strict visual style.
Meta has been clear in its investor communications: the goal is for an advertiser to provide an objective, a budget and some product information, and for AI to do the rest, from creative to targeting. The press reported on plans to fully automate ad creation and targeting by the end of 2026. We don't know exactly what will launch and when, but the direction has been the same for three years: fewer buttons, more automation.
For a webshop, this means three things will matter that the algorithm can't do on its own:
Meta's algorithm needs around 50 conversions per week per ad set to exit the learning phase and deliver steadily. Every separate campaign or ad set splits the budget and the conversions. With a small budget, splitting into several campaigns means none of them learns.
Let's take a concrete example: a skincare webshop in Copenhagen spending 25,000 kr. a month on Meta with a cost per purchase of 200 kr.
Skincare webshop, 25,000 kr. a month DKK Purchases per month (25,000 / 200) 125 Purchases per week about 29 Old structure: 4 campaigns (prospecting, retargeting, lookalike, broad) about 7 per week each Recommended structure: 1 Advantage+ Sales campaign about 29 per weekWith four campaigns, each gets around seven conversions a week, so none exits the learning phase. With one, the algorithm has four times as much data. It still doesn't reach 50, but it's much closer. Results become steadier, and cost per purchase usually drops after a few weeks.
Try it with your own numbers:
If the algorithm chooses the audience based on the ad, then variety in creative becomes your main lever. I don't mean ten versions of the same photo with different text. I mean different concepts that speak to different people.
Here's what three concepts could look like for the same product, a moisturiser for sensitive skin:
1. The problem Who it speaks to Message "Does your skin go red after every new cream?" Format UGC video, a customer talking straight to camera Likely audience People with sensitive skin who have tried other products 2. The ingredient Who it speaks to Message "5 ingredients. No perfume. Dermatologically tested." Format Static image, product and ingredient list Likely audience People who read labels 3. The routine Who it speaks to Message "My 3-minute morning routine" Format Reel showing the steps of a real routine Likely audience People looking for simple routinesEach concept reaches a different type of person. If you only have the first concept in ten variations, Meta keeps showing it to the same type of audience, and cost rises as that audience gets saturated.
A simple system for a webshop with a mid-sized budget:
Videos of real people filmed on a phone still work very well for smaller webshops. You don't need expensive production. You need variety and clear messages in the first three seconds. For Danish audiences, Danish-language voice-over or subtitles almost always beat English.
When automation does almost everything, the one thing the algorithm can't guess is what a good customer means to you. If you only tell it "someone bought", it will look for cheap buyers, including those who return or only buy on sale.
What you can do concretely:
I wrote about the difference between ROAS and profit in the article contribution margin vs ROAS. In 2027 that difference matters even more, because the algorithm optimises for exactly the value you give it.
The 2026 attribution change showed how much Ads Manager figures depend on rules you don't control. Overnight, the same account seems to have 20% fewer conversions, even though sales are the same. For budget decisions you need several sources.
How to check whether Meta brings extra sales When to use it Simple comparison: total sales vs Meta spend, weekly Always, takes 10 minutes Geo test: pause ads in 1-2 regions for 3-4 weeks When you spend over 30,000 kr. a month Conversion Lift, Meta's official test with a control group When you have enough conversions, ask your Meta rep "Where did you hear about us?" after checkout Always, complements the other methodsA real geo test example: a webshop pauses Meta ads in Region Nordjylland for four weeks and compares sales there with the rest of Denmark against the previous period. If sales drop 12% where ads were paused, you know roughly how much extra Meta brings, whatever Ads Manager says.
When every competitor uses the same automated tools, the difference no longer comes from settings. It comes from what you sell and how. A good product, a fair price, fast delivery to a parcel shop, easy returns, MobilePay at checkout and real reviews convert better in any campaign. The algorithm only amplifies what you already have.
What I check in every webshop before increasing Meta budget:
For details on each campaign type, read Meta Ads campaign structure for webshops and the article on the Meta Ads features I use the most.
In 2027, Meta Ads will be even more automated. You no longer win through clever settings but through clean data, varied creative and honest measurement of results. Webshops that adapt first will have a clear advantage over those still splitting their budget across ten campaigns. If you'd like to go through your account and structure together, you'll find details on the Meta Ads page, or you can write to me on the contact page.