Retargeting Ads That Pay Off: How Bidease Converts Ecommerce App Users

There's nothing worse than biting into a burger piled high with your favorite toppings, only to watch them slide off the bun. In ecommerce, retargeting works the same way. You already have the ingredients: shoppers who installed your app and filled their carts. The problem is who you reach, with what, and when. Get that wrong, and the conversions slide right off.

But the retargeting playbook is already written — and getting it right has clear ROI. According to AppsFlyer, app remarketing spend grew 37% to $31.3 billion in 2025, rising from 25% to 29% of total app marketing budgets. That figure includes apps across all genres, but in ecommerce it’s looking more stark: 92% of iOS ad spend for ecom goes to remarketing, as UA gets ever-more competitive. If your priorities are shifting similarly, it’s time to dial in your retargeting strategy.

The bottom line:

• Being on your target list isn't a reason to bid. We score each user's intent and bid only when a conversion is likely.
• Different users need different plays. New installers who haven't ordered yet, lapsed shoppers, cart abandoners, and loyal buyers each get their own goal and KPI.
• Timing, creative, and frequency are decided for each user, not set once per segment.

Why most retargeting ads underperform

Here's the odd thing about retargeting. You already know these users, but another ad won't automatically get them to buy. They may have lost interest, moved to a competitor, or bought elsewhere. Your best buyers may not need a nudge at all. And with dozens of ads competing for attention every day, an unnecessary ad just adds noise.

When retargeting underperforms, reach is rarely the issue. It may just be down to a few bad calls:

• One message goes to users with very different levels of intent.
• Ads land too early, too late, or in the wrong context.
• The creative isn’t primed to capture a user in the moment..

Getting those calls right means reading behavior instead of relying on list membership: recency, frequency, purchase history, and the odds of a next step. Choose partners that have the machine learning chops to dial every element in before placing the bid, without suffering latency.

Not every retargeting campaign has the same job

In ecommerce, retargeting usually covers four jobs, and each needs its own measurement strategy:

• Turn installs into first orders. Measure cost-per-first-order, not installs.
• Win back lapsed shoppers. Measure cost-per-returning-user, and what they do once they're back.
• Recover abandoned carts. Measure order conversion and incremental revenue, not clicks.
• Grow your best customers. Measure purchase frequency, repeat-order margin, and contribution to LTV.

Across the board, you’ll want to favor conversion lift and revenue impact over install volume. If remarketing is your main play, take this a step further by measuring how your efforts influence purchases overall. Last-click attribution won’t reveal campaigns that impact in-store activities, loyalty, or deeper app engagement — these signals will boost your confidence come budget season.

Consider segmentation through the same lens as measurement. Many teams sort existing buyers into these groups with RFM segmentation (recency, frequency, monetary value). Whatever method you use, you set the goal for each group and provide unattributed data to teach the models how to behave.

A peek behind the ML curtain on retargeting

On our DSP, every decision happens at the level of a single ad request. When we receive a request, our models evaluate that impression's value across hundreds of behavioral signals — all in under 20 milliseconds. Here are some of the ways we tie signals back to your audience:

• We can ingest static lists for one-off pushes to a defined cohort, like promo participants.
• We can sync directly to audiences built in your MMP (e.g., AppsFlyer Audiences) and pass live data back as users convert or drop out.
• We can train on non-attributed data (NAD), compiling audiences from anonymized ad IDs, with no extra paid MMP features required. The caveat: they only look back as far as your event history. If you started sending data two weeks ago, we can't yet find users inactive for 30+ days.

The more in-app events you pass along (sessions, product views, add-to-carts, purchases), the more clearly we can see where each user sits in your funnel.

No cold start, thanks to a pre-trained neural network

Most programmatic retargeting starts slow. The algorithms need time to learn who's worth targeting, but you end up paying for that learning curve. We skip it by starting with a neural network model already trained on large volumes of historical data — all specific to the ecommerce vertical. The model picks up subtle patterns like visit density and shifts in activity, accurately judging when a user is likely to convert. Once your campaign has been running for a while, we add custom models trained on your own data, so the system keeps adapting to your specific goals. The fastest path to ROAS is to let the engine run, with as few manual overrides and restrictions as possible. Every bid then comes down to five decisions.

Five decisions behind every bid

1. How close is this user to your target action?

Our Intent Score estimates how close someone is to acting based on recent visits, add-to-carts, and purchases while considering visit frequency, browsing, and ad context. It doesn't decide what matters to your business. You set the goal (a first order, a repeat purchase, a large order), and the model estimates how likely each user is to get you there.

2. Is this impression worth buying?

When the odds of conversion are low, we skip the auction. A cheap bid that leads nowhere is still wasted money. When the odds are high, we bid based on what that impression is likely to be worth against your CPA or ROAS target. Every bid is priced for the individual user, not the segment, so budget flows to the users most likely to pay off.

3. When should we reach out?

Waiting too long to reach a high-intent user can cost you the conversion. A shopper who filled a cart five minutes ago is in a very different place than one who bought last month. The model weighs each signal's freshness and how fast interest fades, so you skip touches that won't make an impact. For ecommerce inactivity windows, start from your purchase cycle: shorter for everyday goods, longer for bigger purchases.

4. What should we show?

Creative often gets treated as an afterthought. In retargeting, it's part of the decision. A shopper browsing one category will usually respond better to a product they viewed than to a brand message. Format matters too: we run native, full-screen, video, rewarded, and interactive formats. A single-format campaign can't compete in auctions where a video or interactive ad would have won.

5. How often should we show the ad?

One of the easiest ways to waste retargeting budget is leaving frequency unchecked. Campaigns drift toward whoever is easiest to reach, usually people who live in apps. You end up hitting the same person with the same ad until they tune it out. We'd suggest starting with a simple cap, like a daily maximum per user. From there, let the system adjust. It keeps showing ads to people who are still responding and pulls back once another ad won't make a difference.

What this means for ecommerce growth

Someone downloads a shopping app, fills their cart, and never places the order. Retargeting is not only how you get that first order, but how you turn that shopper into a high-LTV customer. Getting there comes down to who you reach, with what, and when. That's what the machine learning is for: tying every bid to your CPA or ROAS target, and spending based on each shopper's likely value rather than a segment average. Get the complete picture of ML at Bidease, including the interplay of those models, in our recent post.

Anton Belov, Commercial Director at Bidease:

"App installation doesn't automatically convert a user into a customer. Often interaction ends at first launch, sometimes at a single purchase. The job of modern retargeting is to notice changes in behavior over time and gives users a compelling reason to return. Machine learning helps determine more accurately who, when, and with what message to target, which makes retargeting one of the most effective LTV growth tools."

Ready to turn more of your app users into loyal customers? Schedule a discovery call with our team.

‍