Retargeting shows ads to people who visited your store but didn't buy. It's popular because the reported numbers look great. The catch is that many of those buyers would have come back anyway. Good retargeting is set up by intent, capped so it doesn't annoy, and measured by the sales it actually adds.
Build audiences by intent
Not every visitor is equally close to buying. Typical tiers, from warmest to coldest:
- Started checkout in the last few days.
- Added to cart in the last week or two.
- Viewed a product in the last two to four weeks.
- Visited any page in the last month.
Warmer audiences are smaller and convert better. Budget and bids should reflect that. You'll need your ad platform's pixel (and ideally its server-side conversions API) installed correctly; see Snap Pixel and Meta ads for the tracking side.
Match the message to the drop-off
- Abandoned checkout: a reminder of the exact product, and whatever usually stops people at that stage (shipping cost, delivery time, returns).
- Added to cart: the product plus reviews or a guarantee.
- Viewed product: the product and close alternatives, or a best-seller in the same category.
- General visitors: your brand story and strongest products.
Discounts work, but train customers to abandon carts on purpose. Test a non-discount reminder first.
Cap frequency and exclude buyers
Seeing the same ad fifteen times in a week annoys people. Set frequency caps where the platform allows, rotate creative, and exclude anyone who has just purchased. Nothing looks worse than an ad for the shoes someone bought yesterday.
Don't forget the free channel
Shopify's abandoned checkout emails reach people who entered an email address, at almost no cost. Make sure they're switched on and well written before spending on ads for the same people.
Measure incremental lift with a holdout
Platform dashboards count any purchase after an ad view or click as a conversion. For retargeting, that overstates the impact because the audience was already likely to buy.
The fix is a holdout: a random share of the retargeting audience that doesn't see the ads. Compare purchase rates between the exposed and held-out groups. Several ad platforms offer lift studies that do this for you; otherwise you can split audiences yourself.
Example. 40,000 cart abandoners are split 90/10. The exposed group converts at 6.2% within two weeks; the holdout converts at 5.4%. The incremental lift is 0.8 points, about 15% relative, far less than the dashboard's attributed conversions would suggest. Enter the counts in the significance calculator to check that the gap is real, then compare the incremental orders with what you spent.
When retargeting isn't worth it
If a holdout test shows little incremental lift, put the budget into reaching new customers or fixing whatever made people leave in the first place. Often the cheapest "retargeting" is a product page that answers the question that sent them away. The conversion rate guide covers those fixes.
Common questions
Do retargeting ads actually work?
They often get credit for sales that would have happened anyway, because the people they target were already likely to buy. They can still be profitable, but the only way to know is to measure incremental sales against a holdout group that doesn't see the ads.
How long should a retargeting window be?
Match it to how long customers usually take to decide. For low-price impulse products, a few days to two weeks; for considered purchases, several weeks. Shorter windows usually perform better per impression.
Should I retarget people who already bought?
Exclude recent buyers from ads for the product they just bought. Retargeting past customers with complementary products or replenishment reminders can work well, as a separate campaign.