How to Start an Ecommerce Business With Facebook Ads

Starting an ecommerce business through Facebook Ads is a common, proven route — whether you run Cash on Delivery, prepaid, or both — but without the right structure, budget disappears fast with nothing to show for it. Here’s the exact process we follow for every new campaign, built from running it across Asia; the structure holds regardless of where you’re targeting, and getting the budget size and creative mix right at launch matters just as much as the technical setup steps below.

Step 1: Campaign Structure

Always build a new campaign in PAUSED state — campaign first, then ad set, then ads. Activate in sequence, and check performance after each step before moving to the next.

• Campaign objective: Purchase (if your pixel is already collecting events) or Conversions
• Budget: Start with CBO (Campaign Budget Optimization) so Facebook can decide which ad set deserves more spend

Step 2: Targeting

Broad targeting consistently delivers the best results early on, before your pixel has enough data to work with:

• Location: Go broad — a whole region or continent rather than one city or country, unless you have a strong reason to restrict it. In our own campaigns (run across Asia) that’s meant targeting the whole continent rather than a single country; the same principle applies wherever you’re running from — widen to your own broad region instead of narrowing too early.
• Age: 25–55
• Interests: Never use invented or placeholder interest IDs — if you need a specific interest, select it from Facebook’s official targeting search only

Step 3: Pixel Setup

Install your Facebook Pixel on every page of your site, especially:

• The product/landing page (PageView event)
• After the order form is submitted (Purchase or Lead event)

Without a pixel, Facebook has no way of knowing which audience is actually converting, and optimization simply won’t work. If you’re running Google Ads alongside Facebook, the same principle applies there too — see our guide to Google Ads conversion tracking setup.

Once the base pixel is live, add Conversions API alongside it — browser tracking prevention alone drops a real share of pixel-only events, which quietly inflates CPR without an obvious cause.

Step 4: Placement — An Important Warning

Always exclude Audience Network placement. We discovered this in our own campaigns — orders coming from Audience Network have a much higher ratio of fake or wrong-number entries, which is especially costly if you’re running COD (wasted courier time) but hurts data quality either way.

In your placement settings, keep only:

• Facebook
• Instagram

And omit the audience_network_positions field entirely — don’t include it at all.

Step 5: Activation

Once everything’s reviewed:

1. Activate the campaign
2. Activate the ad set
3. Activate individual ads

After each step, wait 24–48 hours and check CPR (Cost Per Result) before taking the next step.

How Much Budget to Actually Start With

A first campaign needs enough daily budget for Facebook’s algorithm to gather real signal, not the smallest number that feels safe — a budget so small it can’t produce even a handful of results per day just stretches out the learning phase without actually saving money overall. A starting point of roughly 5-10x the target cost-per-result, per ad set, gives the algorithm enough room to find real conversions within the first few days rather than guessing indefinitely on too little data.

Creative Testing Cadence for a First Campaign

Launch with 3-4 genuinely different creative concepts per ad set rather than minor variations of the same one (different colors of the same image isn’t real variation) — testing different angles (a direct product shot, a before/after, a short testimonial-style video) reveals which approach the audience actually responds to, information a single creative can’t provide on its own. Once one clearly outperforms the others, that’s the point to narrow down and start scaling the winner rather than continuing to split budget evenly.

What a Realistic First Week Actually Looks Like

Expect a rocky first few days — costs are often higher and less consistent than they’ll settle into once the algorithm has real data, and a few days of above-target CPR isn’t a signal to panic or immediately change everything. The genuine signal worth reacting to is a clear trend after the full learning phase (roughly 50 optimization events or about a week), not the first 24-48 hours of a brand-new campaign still finding its footing.

Once This Manual Structure Is Working

The manual, paused-state build above is the right starting point, but it isn’t the only campaign type worth knowing about — once a pixel has real purchase history from campaigns built this way, our Advantage+ Shopping Campaigns guide covers Meta’s automated campaign type and when it’s actually worth adding alongside a manual campaign like this one, not before it.

What Comes After Facebook

Once your Facebook campaign is stable, two things are worth trying next: a structured use of AI for writing ad copy, and Google Search Ads for intent-based traffic — together these two channels cover both interruption-based and intent-based traffic, giving a store two genuinely different ways for a customer to discover it rather than relying on a single channel’s algorithm and cost trends indefinitely.

FAQ

Why build the campaign in paused state?

So nothing spends before every layer is checked. Build campaign, then ad set, then ads, activate in that sequence, and check cost per result after each step — a mistake caught while paused costs nothing.

Why exclude Audience Network placement specifically?

Because in our own campaigns, orders from Audience Network carried a much higher share of fake or wrong-number entries. On COD that’s wasted courier time on top of wasted spend; on any model it poisons the data the algorithm learns from.

How much daily budget should a first campaign start with?

Roughly 5–10x your target cost per result, per ad set. Less than that doesn’t save money — it stretches the learning phase out without producing enough results for the algorithm to learn from.

When should early results actually worry me?

Not in the first 24–48 hours. Costs run high and inconsistent while the algorithm has no data. The real signal comes after the full learning phase — roughly 50 optimization events or about a week — not from day-one numbers.