Most SMEs that run remarketing make the same mistake: they show the same generic ad to everyone who visited their site, regardless of what those visitors actually did there. That's not remarketing — that's spending your budget twice on the same person with the same message.
We see it constantly in new clients who come to us. They have the pixel installed, the audience configured, and remarketing still doesn't convert any better than the cold campaign. The problem is almost never the platform. It's the lack of segmentation by intent.
💡 Remarketing isn't "showing the same ad again." It's talking differently to each person depending on where they stopped in their decision. That's the difference between spending your budget and turning it into sales.
From 2% to 9%: the recovery rate that changed the ROAS
For one of our clients, an online store based in Los Lagos, remarketing existed but ran like any other campaign: same ad, same offer, for anyone who had set foot on the site in the last 30 days. The recovery rate — people who came back to buy after seeing the remarketing ad — sat at just 2%.
We restructured the segmentation into three groups based on real purchase intent. In six weeks, the recovery rate climbed to 9% and the account's ROAS went from 2.1x to 5.8x. We didn't increase the budget. We changed who we were talking to and what we were telling them.
The three segments that made the difference
There's no magic here. There's structure. We split the remarketing audience into three groups, each with a different message:
1. Visitors who browsed but never added to cart
This person is still evaluating. Don't show them a discount yet — show them social proof and differentiators. Reviews, comparisons, reasons why your product is the best option. The goal is to move the decision forward, not close it right away.
2. Abandoned carts
This is the hottest audience — and the one most wasted with generic messaging. Real urgency works well here: limited stock, a time-limited free shipping offer, or a simple reminder that the cart is still there. The copy needs to feel like a nudge, not an aggressive discount pitch from day one.
3. Past buyers
The most common mistake is to stop investing in people who already bought. This group has the lowest acquisition cost in the entire account. Cross-sell and upsell with complementary products generate repeat sales at a fraction of the cost of winning a new customer.
🎯 The lesson: Remarketing isn't about spending more budget on the same people. It's about speaking differently to each segment based on where they stopped in their purchase decision.
The system we use to segment without wasting time
You don't need a complicated stack. With the right setup from day one, this becomes almost entirely automated:
- GA4 audiences: segmented by event (view_item, add_to_cart, begin_checkout, purchase) to know exactly which stage each visitor stopped at.
- Meta Pixel custom audiences: with different time windows — 1-3 days for abandoned carts, 30 days for general visitors.
- Exclusion lists: recent buyers are excluded from acquisition campaigns and moved into the cross-sell campaign only.
- Frequency caps: weekly impression limits so you don't burn out the audience or generate backlash against the brand.
- Rotating creative: at least 3 variants per segment, reviewed every 2 weeks to avoid ad fatigue.
The warning sign
If your remarketing campaign has run for months with the same ad and the same undivided audience, you're burning budget on autopilot. The platform will keep delivering impressions — but at an ever-higher cost per result, because the same audience has already seen that ad dozens of times.
This applies whether you sell from Puerto Varas to all of Chile or your store ships to Latin America and Spain. Market size isn't the variable that matters — segmentation by intent is.
🔍 If your remarketing has been stalled for more than a month, before raising the budget check whether you're actually segmenting by intent or still showing the same thing to everyone. The answer usually explains why ROAS isn't climbing.
