How much to spend on Meta ads as a new Shopify store

The honest answer is €50 to €100 per day for a minimum of 14 days — so €700 to €1,400 as a learning budget. Anything less and you will not get enough data to tell signal from noise. Anything more, as a brand new store, is you paying for the privilege of learning that your product-market fit is not there yet.

Why €50/day is the floor

Meta's optimization model needs about 50 conversion events per ad set per week to stabilise. If your conversion rate is two percent and your CPC is €1, that is 2,500 clicks, which at €1 each is €357 per ad set per week — call it €50/day. Below that threshold, Meta is essentially guessing, and so are you.

Why 14 days is the floor

Meta's learning phase lasts about 7 days. For the first week the numbers are noisy and unreliable — you will see a €5 CAC on day three and a €40 CAC on day five and neither is real. Week two is when the math stabilises. If you read week-one numbers as gospel, you will shut off a winning ad or double down on a losing one.

The three-phase budget model

Phase 1: validation (€700 to €1,400 total)

Spend €50 to €100/day for 14 days. Run two ad sets, each with three creative variants. Goal: answer a single yes/no question — can you get to a CAC under your break-even number, at any volume? If the answer is no after 14 days, the problem is either the product, the price, or the creative — not the budget.

Phase 2: scaling (€150 to €500/day)

Once you have one winning creative and a sub-break-even CAC, increase the budget by 20 percent every 3 days. Any faster and you re-trigger Meta's learning phase. This phase ends when you hit the point where CAC starts to rise with each budget increase.

Phase 3: maintenance (whatever is profitable)

You now have a profitable CAC ceiling. Spend up to it. Every new creative test is a small experiment on top of the baseline, funded by the margin on the ads that are already working.

The most common mistake

Founders test with €20/day for a week, see a €12 CAC with two purchases, and conclude the product does not work. With two data points you have conclusively proved nothing. The only real answer is "not enough data — keep going." If you cannot afford the full €700 learning budget, wait until you can. Meta will not be a cheaper lesson six months from now.

What nordenagent automates here

The product watches for the learning-phase boundary automatically and will not recommend a major budget change during the noisy first week. It also catches the "false CAC" problem — reporting a €3 CAC that is actually €15 once attribution settles — so you do not accidentally scale a phantom winner.