How retargeting recovers baskets

Split by intent#

Separate viewed, basket and cross-sell into distinct ad sets with distinct bids and creative pressure. Viewers need reminding, basket holders need reassurance on delivery and returns, past buyers need complements not repeats. Blended retargeting overpays browsers and under-serves baskets. Catalogue ads set the matching base.

Exclude recent purchasers of the viewed item from its own recovery set; nothing burns goodwill like chasing the converted. Start from the Meta Ads hub for the family view. Meta own catalogue ads guide covers the audience options.

Build the lists from Pixel and app events with matching IDs; mismatched variants silently empty the most valuable set. Custom audiences build the source lists.

Windows that respect cycles#

Size lookback windows to buying cycles: short for replenishment and fashion, longer for considered purchases. Three day baskets, fourteen day viewers and thirty day cross-sell suit many shops, but validate against own checkout intervals. Stale windows chase settled decisions and invite irritation.

Layer engaged viewers and video watchers where Pixel pools run thin, but keep them separate from product intent sets. Intent quality differs and budgets should know it. Prospecting splits govern the divide.

Shorten windows when frequency climbs and pools shrink; small audiences fatigue fast. Fatigue guide shows the decay curve.

Creative that reassures#

Show the viewed item first, then complements, with delivery, returns and stock honesty beside the price. Basket creative answers the top doubt: delivery time, returns ease, payment choice. Social proof near the action lifts recovery without discounting the brand. Dynamic overlays stamp honest prices and badges automatically.

Rotate creative monthly on small pools, sooner where frequency runs hot. Familiarity erodes even perfect reminders. Meta carousel specs suit multi-item recovery stories.

Keep prospecting covers distinct from recovery cards so warm viewers feel recognised, not re-acquired. Catalogue creative separates the jobs.

Frequency that never stalks#

Cap frequency and honour recency: refreshing reminders beat relentless ones. A thread that feels stalked closes forever, while a well timed nudge reopens. Exclude converters daily, suppress chronic non-responders after a defined window, and pause recovery the moment stock or price breaks. Selling absence to warm shoppers poisons every later ad.

Coordinate catalogue retargeting with message follow-up and email recovery so channels sequence rather than pile on. One reminder per trigger, not three at once. Message follow-up needs the same caps.

Watch comments and hides as closely as ROAS; irritation precedes the numbers. Lower pressure the week complaints rise.

Cross-sell and replenishment#

Build cross-sell sets from complements and consumables tied to past purchases: accessories, refills, seasonal companions. Time replenishment to usage intervals, not calendars. Generic bestsellers sent to everyone train buyers to ignore the brand. Product sets slice the complements.

Fund cross-sell from retention budgets with separate targets; different money, different maths. Judge on repeat rate and margin, not on first-order CAC. Ecommerce playbook weights the mix.

Retire cross-sell angles that never attach; attachment rates differ by category and honesty beats persistence.

Prove incrementality#

Hold out retargeting periodically to prove it creates orders rather than claiming them; retargeting always looks brilliant until tested. Judge recovery rate and contribution with holds, not platform ROAS alone. Review quarterly without exception. Conversion Lift settles the argument.

Compare recovery against prospecting on new-customer share so acquisition stays funded. Last-touch reporting starves the prospecting that starts everything. Attribution windows shape the read.

Bring doubts to people who test for a living. Our team designs holdouts that survive finance scrutiny and repeat them on a fixed quarterly cadence.

Frequently asked questions#