Meta ads for a Shopify store: the first 90 days
A week-by-week plan for your first three months on Facebook and Instagram ads: what to set up, what to spend, what to read and when to decide.
Three months is roughly how long it takes to find out whether paid social works for your store. Less than that and you are reading noise; much more than that without a decision and you are funding a habit.
This is a plan for those ninety days: what to do in each phase, what each phase is supposed to produce, and the decision that closes it.
Before day one: the setup that cannot be added later
Three things are worth more than any amount of early optimisation, and all of them are free.
The pixel, with real events. Not just page views — view content, add to cart, initiate checkout and purchase. Every day of spending without them is a day of data you cannot recover. Shopify’s Facebook channel does most of this for you; verify it in Events Manager rather than assuming.
The Conversions API. Browser tracking loses a meaningful share of events to ad blockers and privacy settings. Server-side recovers a good part of it, and it is close to a checkbox on Shopify.
An email capture. Even with nothing to send yet. A list of real purchasers is the seed for the best lookalike you will ever build, and 200 buyers beat 20,000 site visitors as a seed.
Then do the arithmetic before you spend anything. Work out your gross margin, your break-even ROAS (1 / margin) and the smallest budget that can reach 50 optimisation events a week. Forecast the campaign in a pessimistic, central and optimistic scenario. If the pessimistic case is unsurvivable for three weeks, fix that before launching, not after.
Days 1–21: buy one answer
The first phase is not trying to be profitable. It is trying to find out whether the creative connects and whether your funnel converts cold traffic at all.
Structure. One campaign, conversion objective. One ad set, broad targeting, optimising for the cheapest event you can get 50 of — usually add-to-cart, not purchase. Three to five ads with genuinely different angles, not four crops of the same photo. The structure arithmetic explains why one ad set beats two at this budget.
Budget. The number your forecast produced. Commit the whole three weeks up front and treat it as the price of finding out — because stopping on day five is how you pay the full cost of the learning phase and throw away the result.
What to do during it. Very little. Week one is learning and the numbers are not predictive. Do not change budgets, do not add ad sets, do not swap creative. The discipline here is the hard part.
What this phase produces. Answers to three questions, in order: does anyone click, do the clickers add to cart, do the add-to-carts convert. A campaign with 1.4% CTR, healthy add-to-cart and no purchases has told you something valuable — the ads work and the problem is downstream.
Days 22–45: fix the weakest link
Now you read the export properly. Pull it at ad set level, then at ad level, and work through the columns rather than staring at the dashboard.
Compare each rate against a category benchmark to find which link in the chain is broken:
| Symptom | Where the problem is | What you change |
|---|---|---|
| CTR under 0.6% | Creative or audience | New creative angles |
| CTR fine, add-to-cart under 3% | Product page or price | The page, not the ads |
| Add-to-cart fine, purchases low | Checkout | Shipping, payment, account friction |
| Everything in band, ROAS under break-even | Unit economics | Price, AOV or margin |
Change one thing, wait three days, read. Two changes at once means you learn nothing from either.
By day 45 you should know your real CTR, your real cold-traffic conversion rate and your real CPA. Those three numbers are the actual asset produced by the first six weeks — they make every subsequent forecast meaningful.
Days 46–75: make it repeatable
If the fixes worked, this phase turns a result into a process.
Switch the optimisation event to purchase once you can sustain 50 purchases a week. If you cannot, stay on add-to-cart; there is no prize for switching early.
Add the second ad set. A lookalike from purchasers if you have enough of them, or a broad interest cluster. Split on audience type, not on micro-interests that overlap each other.
Start a creative queue. Two ads in reserve at all times, written calmly. Copy and images fatigue together, and the moment performance drops is the worst time to write.
Begin scaling, slowly. Around 20% every three or four days, and only when ROAS is comfortably above break-even and stable across days rather than in aggregate. The conditions worth waiting for matter more than the percentage.
Expect at least one week where the numbers fall for no visible reason. Before rebuilding anything, work through the causes in order — two thirds of the time it is creative fatigue or a change someone made and forgot.
Days 76–90: decide
The last two weeks are for a decision, and it should be made against numbers you wrote down in advance.
Keep and scale, if CPA is below your contribution margin per order and has been stable for three weeks. Move to the rhythm of the kill-keep-scale rules and start testing Advantage+ alongside the manual campaign now that you have the pixel history to feed it.
Keep at current spend, if you are near break-even and improving. More budget will not fix a marginal campaign; better creative might.
Stop, if after three weeks of clean data at an adequate budget the pessimistic case is what you got. This is a legitimate outcome and it is cheap at ninety days. What is expensive is the version where you never set a decision date and find out at month nine.
The three habits that separate stores that make it work
Write down the decision rule before you look at the numbers. Judgement applied while watching money leave is not judgement.
Change one thing at a time. It feels slow and it is the fastest route to knowledge that transfers.
Compare against your own margin, never against someone else’s ROAS. A 2.4 ROAS is excellent at 70% margin and loss-making at 35%. Most advice on the internet does not know which one you are — nor what you sell, which changes the numbers just as much.
AdPilot covers the three parts of this plan that are mechanical: the generator builds the campaign structure and ad angles from your Shopify catalogue, the simulator runs the forecast before you commit a budget, and the optimizer turns each export into a per-ad-set verdict with your own margin in the arithmetic.