How to Set Amazon Display Remarketing Windows

Name the historical action the campaign is allowed to use. Views remarketing concerns people who viewed a product or related product, while purchases remarketing concerns people who bought a product, complementary product, brand, or category within the selected historical period. These are different human states. Someone considering a durable product, replenishing a consumable, or cross-shopping an accessory should not be placed into one audience merely because the console offers a convenient shared window.

Write the campaign job in one sentence: recover recent product consideration, prompt expected replenishment, cross-sell after a known purchase, or introduce a product to buyers of a relevant category. Then list exclusions, product eligibility, geography, inventory requirement, and the customer action that ends the audience's usefulness. If the business cannot explain why a person remains relevant on day seven, thirty, or ninety, it does not yet have a defensible lookback window.

Use the brand's own order intervals, product-view timing, repeat-purchase distribution, and category consideration pattern to propose the first window. Work with medians and distributions rather than a single average that can be distorted by a small number of long gaps. Separate first purchase, replenishment, replacement, and cross-sell behavior. The chosen window should cover the period in which the intended action remains plausible without keeping stale shoppers eligible simply to make the audience larger.

Document data coverage and gaps. A new product may lack enough history, and a seasonal product may have a buying cycle that changes across the year. In those cases, use a conservative canary based on the closest verified product family and label the assumption. Do not convert Amazon's examples into a forecast for a different catalog. The examples establish that windows are customizable; they do not prove the correct interval for an individual advertiser.

Create distinct audience cells when the product set, historical action, window, bid, optimization, or exclusion logic differs. A recent-view recovery cell can use a shorter consideration window, while a replenishment cell can begin later and align with an observed reorder interval. A complementary-product cross-sell should name both the prior purchase family and the advertised product family. Clear names make the account auditable and prevent one blended audience from hiding which customer state received spend.

Keep audience overlap visible. If one person can qualify for views remarketing, purchases remarketing, and an Amazon audience at the same time, record the intended priority and how reporting will be interpreted. Do not assume platform delivery isolates the causal effect of each cell. Use campaign and targeting reports to examine delivery, and preserve the definitions used at launch. If an audience definition changes, version the name or annotate the effective date rather than silently overwriting the meaning.

Amazon documents different optimization choices for conversion, page-visit, and viewable-impression goals. Select one that matches the approved campaign purpose, then set an exposure limit the product economics can support. Replenishment creative should not pretend the customer needs a product on an unsupported schedule, and cross-sell creative should explain the complementary value without making claims the product detail page does not support. Confirm the advertised product is available and the landing experience matches the message.

Use separate creative when the human question changes. A shopper who viewed but did not buy may need product clarification, while a prior buyer may need a reminder about a compatible or replenishable item. Avoid generic copy that treats every eligible audience as equally familiar with the brand. Review product titles, imagery, price, promotion, and delivery promise before launch because display ads can inherit retail information. An audience test is not interpretable when the retail offer changes without being recorded.

Run the smallest set of windows that can answer the decision. If the account has enough traffic, compare adjacent intervals while holding product, creative, bid strategy, and observation dates stable. When that is not practical, test one conservative window against the prior operating baseline. Record eligible audience size, delivery, spend, clicks or page visits, attributed orders or sales, repeat-purchase indicators, and new-to-brand measures only where each metric fits the audience purpose.

A longer lookback window may increase eligibility while weakening recency, and a shorter one may preserve recency while limiting delivery. Neither result is automatically good or bad. Judge the window against the approved customer state, cost guardrail, inventory plan, and creative promise. Label low-volume results as inconclusive. Do not expand the window just to spend the budget, and do not claim incrementality from attributed performance alone when no controlled comparison isolates the exposure.

Approve the audience action, product family, lookback window, exclusions, bid strategy, and maximum exposure as one versioned definition. Write the approved purpose, accountable owner, input evidence, exclusions, and review date before changing production. Keep the prior configuration or report export with the decision record so the team can distinguish a deliberate change from an unexplained drift. A reversible canary is more useful than a broad rollout because it exposes mismatched data, eligibility, status, or workflow assumptions while the affected set is still small.

Monitor audience overlap, product availability, delivery, spend, and the metrics appropriate to consideration, replenishment, or cross-sell. Review the first complete operating period against the documented baseline, not against a desired outcome. Record exceptions separately from normal cases, and do not assign causal impact when the available evidence only shows association. Pause or restore the prior window when relevance, cost, inventory, or retail-readiness guardrails fail. The durable deliverable is a traceable decision with a named owner, comparable evidence, and a clear next review, whether the team keeps, revises, or removes the change.