On this page本页目录
How do you calculate SKU return rate?如何计算 SKU 退货率？
Divide returned units for one stable sellable SKU by eligible fulfilled units of that same SKU in a mature cohort, then multiply by 100. Use the same event rules and return opportunity for every comparison. Publish the numerator, denominator, observation window, as-of date, confidence or uncertainty, reason coverage, returned value, and contribution impact. Never rank tiny or reused SKUs from percentages alone.
用一个稳定可销售 SKU 的退回件数除以同一 SKU 在成熟群组中的合格履约件数，再乘以 100。所有比较使用相同事件规则与退货机会，同时发布分子、分母、观察窗口、截至日期、置信或不确定性、原因覆盖、退回金额与贡献影响。不要仅凭百分比对小样本或复用 SKU 排名。
A SKU should identify one sellable variant consistently across catalog, order, inventory, fulfillment, return, and finance records. A parent product or item-group ID groups variants and is not the same analytical unit. If the business has reused or changed SKUs, create a versioned surrogate key rather than silently joining unlike items.
SKU 应在目录、订单、库存、履约、退货与财务记录中稳定识别一个可销售变体。父商品或商品组 ID 用于聚合变体，不是同一分析单位。如果业务复用或更改过 SKU，应创建版本化代理键，而不是静默关联不同商品。
Preserve six levels of product identity and evidence保留六层商品身份与证据
An accurate rate depends on identity before calculation. Keep the business SKU visible, but join through a stable variant key and effective dates. Preserve parent grouping and attributes so analysts can roll up or drill down without changing the base unit.
准确退货率先依赖身份，再依赖计算。保留业务 SKU 展示，但通过稳定变体键与生效日期关联；保留父级分组和属性，使分析员能汇总或下钻而不改变基础单位。
Immutable internal key plus effective dates for each sellable configuration.每个可销售配置的不变内部键与生效日期。
SKU, barcode/GTIN where applicable, platform variant ID, supplier code, and legacy aliases.SKU、适用时的条码/GTIN、平台变体 ID、供应商代码与历史别名。
Parent/style, item group, category, brand, model, launch, lifecycle, and bundle relationship.父级/款式、商品组、品类、品牌、型号、上市、生命周期与套装关系。
Size, color, material, capacity, compatibility, pack quantity, condition, and locale.尺码、颜色、材质、容量、兼容性、包装数量、状态与语言。
Eligible fulfilled units, purchase/fulfillment cohort, return events, exchanges, cancellations, and window maturity.合格履约件数、购买/履约群组、退货事件、换货、取消与窗口成熟度。
Raw/mapped reason, inspection, disposition, refund, cost, recovery, contribution, and cause status.原始/映射原因、质检、处置、退款、成本、回收、贡献与原因状态。
Shopify distinguishes SKUs from barcodes and recommends unique SKUs per variant for effective tracking and reporting. Google distinguishes unique item IDs from shared item-group IDs. Apply the principle to your own governed catalog; neither source validates your data.
Shopify 区分 SKU 与条码，并建议每个变体使用唯一 SKU 以便有效追踪和报告。Google 区分唯一商品 ID 与共享商品组 ID。应把原则应用到自己的受治理目录；两者都不会验证你的数据。
Define the SKU metric contract before ranking排名前先定义 SKU 指标契约
Write one metric contract and version it. A return request, received item, approved return, completed refund, exchange, and cancellation are different events. The chosen event can be useful, but it must remain stable and be named in the result.
编写并版本化一份指标契约。退货申请、收到退回商品、批准退货、完成退款、换货与取消是不同事件。所选事件可以有用，但必须稳定并在结果中命名。
| Decision决策 | Recommended record建议记录 | Risk if omitted省略风险 |
|---|---|---|
| Observation unit观察单位 | Returned units and eligible fulfilled units by stable variant key按稳定变体键记录退回件与合格履约件 | Order-level partial returns are distorted订单级部分退货失真 |
| Cohort anchor群组锚点 | Purchase or fulfillment date with one defined observation window购买或履约日期加定义观察窗口 | Event month and cohort month are mixed事件月与群组月混淆 |
| Return event退货事件 | Request, receipt, approval, or another explicitly named state申请、收货、批准或其他明确命名状态 | Operational lag looks like performance change运营滞后被误认为表现变化 |
| SKU identitySKU 身份 | Immutable variant key plus SKU alias and effective dates不变变体键加 SKU 别名与生效日期 | Reused codes merge different products复用代码合并不同商品 |
| Exchange logic换货逻辑 | Return leg and replacement leg stored separately退回与替换环节分开保存 | Units are double counted or hidden件数被重复计算或隐藏 |
Create a trustworthy SKU return table建立可信 SKU 退货表
- Build the variant dimension建立变体维表
Assign an immutable variant key, product parent, attributes, source IDs, aliases, and effective dates.分配不变变体键、商品父级、属性、来源 ID、别名与生效日期。 - Create eligible fulfillment lines创建合格履约行
Use one row per unit or order line with quantity, event time, store, channel, currency, promotion, and eligibility.按件或订单行记录数量、事件时间、商店、渠道、币种、促销与资格。 - Join return lines关联退货行
Link case and returned quantity to the original fulfillment line; quarantine ambiguous and unmatched records.把案例与退回数量关联原履约行，并隔离模糊和未匹配记录。 - Apply one event definition应用统一事件定义
Select and version the qualifying return state, exclusions, exchange treatment, and duplicate rules.选择并版本化合格退货状态、排除、换货处理与重复规则。 - Mark cohort maturity标记群组成熟度
Calculate return opportunity from the window and as-of date; separate incomplete cohorts.根据退货窗口与截至日期计算退货机会，并分离未完成群组。 - Attach outcomes关联结果
Join reasons, inspection, disposition, refund, processing cost, recovery, and contribution before/after return.关联原因、质检、处置、退款、处理成本、回收与退货前后贡献。 - Validate and publish coverage验证并发布覆盖
Reconcile totals and report missing SKU, duplicate SKU, unmatched returns, unknown reasons, and uninspected units.核对总数，并报告 SKU 缺失、SKU 重复、未匹配退货、未知原因与未质检件。
The CSV includes stable variant identity, parent grouping, attributes, eligible fulfillment, cohort maturity, return events, reasons, inspection, value, cost, recovery, contribution, and evidence status.
CSV 包含稳定变体身份、父级分组、属性、合格履约、群组成熟度、退货事件、原因、质检、金额、成本、回收、贡献与证据状态。
Download CSV template下载 CSV 模板 ↓Calculate rate, uncertainty, and materiality together同时计算退货率、不确定性与重要性
Pair the rate with returned units, fulfilled units, returned value, contribution loss, reason/inspection coverage, confidence interval or shrinkage estimate, benchmark definition, and as-of date. For low-volume SKUs, show uncertainty or roll up before ranking.
退货率应与退回件数、履约件数、退回金额、贡献损失、原因/质检覆盖、置信区间或收缩估计、基准定义及截至日期一起展示。低销量 SKU 应显示不确定性或先汇总再排名。
| Output输出 | Required context所需背景 | What it can support可支持内容 |
|---|---|---|
| SKU return rateSKU 退货率 | Stable SKU identity, mature eligible units, qualifying return event稳定 SKU 身份、成熟合格件、合格退货事件 | Comparable variant-level frequency可比变体级频率 |
| Rate interval / shrinkage退货率区间/收缩 | Returned and fulfilled counts, chosen statistical method退回与履约数量、所选统计方法 | Avoid overreacting to small samples避免对小样本过度反应 |
| Excess returns vs reference相对参考的超额退货 | Explicit matched reference and expected rate明确匹配参考与预期退货率 | Estimate operational opportunity估计运营机会 |
| Contribution after returns退货后贡献 | Net revenue, COGS, shipping, processing, recovery, fees净收入、商品成本、运输、处理、回收、费用 | Prioritize economic consequence确定经济后果优先级 |
| Reason/condition profile原因/状态画像 | Capture coverage, versioned codes, inspection, multi-reason采集覆盖、版本化代码、质检、多原因 | Generate hypotheses only仅生成假设 |
Worked example: rate alone changes the wrong priority示例：只看退货率会排错优先级
A synthetic mature cohort compares three SKUs. SKU A has 8 returns from 40 fulfilled units. SKU B has 180 returns from 2,000 units. SKU C has 70 returns from 500 units. Returned value and after-return contribution are also available.
一个模拟成熟群组比较三个 SKU。SKU A 在 40 件履约中退回 8 件，SKU B 在 2,000 件中退回 180 件，SKU C 在 500 件中退回 70 件，并可获得退回金额与退货后贡献。
| Synthetic SKU模拟 SKU | Rate退货率 | Count数量 | Decision note决策说明 |
|---|---|---|---|
| A | 20.0% | 8 / 40 | Highest point rate, widest uncertainty点估计最高，不确定性最宽 |
| B | 9.0% | 180 / 2,000 | Largest return volume and dollar exposure退货量与金额暴露最大 |
| C | 14.0% | 70 / 500 | Material rate and analyzable cohort退货率重要且群组可分析 |
A should not automatically rank first: eight events provide limited precision. B may create the largest total loss despite a lower rate. C may be the best diagnostic starting point if its interval, contribution loss, and one reason/condition cohort remain material. Publish the chosen decision rule instead of presenting one universal ranking.
A 不应自动排第一，因为 8 个事件精度有限。B 的退货率较低，却可能产生最大总损失。如果 C 的区间、贡献损失及某个原因/状态群组仍然重要，它可能是更好的诊断起点。应发布所选决策规则，而不是给出一个通用排名。
All store figures and records in this example are synthetic. They illustrate the method and do not represent InfiniSynapse customer results or industry benchmarks.本示例中的商店数字与记录均为模拟，仅用于说明方法，不代表 InfiniSynapse 客户结果或行业基准。
Read SKU rate through volume, value, and evidence结合数量、金额与证据解读 SKU 退货率
Use rate to compare frequency, counts to understand workload, value and contribution to understand economics, and reason/inspection evidence to form hypotheses. Segment by variant attributes only after confirming the SKU-to-attribute mapping and enough exposure. A high SKU rate can arise from product, content, fulfillment, customer mix, policy, promotion, season, or chance.
用退货率比较频率，用数量理解工作量，用金额和贡献理解经济影响，用原因/质检证据形成假设。只有确认 SKU—属性映射且暴露足够后，才按变体属性细分。高 SKU 退货率可能来自商品、内容、履约、客户结构、政策、促销、季节或随机波动。
Show uncertainty, review severe cases, and avoid broad action from a volatile estimate.展示不确定性，复核严重案例，避免基于波动估计采取广泛行动。
Prioritize when sales exposure and unrecovered contribution make total impact material.当销量暴露与未回收贡献使总影响重要时提高优先级。
Compare matched variants and verify size, color, material, compatibility, or supplier evidence.比较匹配变体并核验尺码、颜色、材质、兼容性或供应商证据。
Stop analysis when duplicate, missing, recycled, or cross-channel SKU mappings are unresolved.重复、缺失、复用或跨渠道 SKU 映射未解析时停止分析。
The best first target is often not the highest percentage. Choose a cohort with material preventable loss, adequate sample and evidence, a plausible mechanism, and a reversible test.
最佳首要目标通常不是最高百分比。应选择可预防损失重要、样本与证据充分、机制合理且可逆测试的群组。
Run nine controls before comparing SKUs比较 SKU 前完成九项控制
- Uniqueness: every active sellable variant has one stable key and no unresolved duplicate SKU.唯一性：每个活跃可销售变体都有一个稳定键，且没有未解析重复 SKU。
- Effective dates: SKU aliases, product hierarchy, attributes, costs, and prices are time-valid.生效日期：SKU 别名、商品层级、属性、成本与价格均按时间有效。
- Unit reconciliation: ordered, fulfilled, canceled, returned, exchanged, and refunded quantities balance.件数核对：下单、履约、取消、退回、换货与退款数量平衡。
- Mature cohorts: compared SKUs have equal return opportunity.成熟群组：被比较 SKU 具有相同退货机会。
- Event consistency: one qualifying return event and duplicate rule apply to all SKUs.事件一致：所有 SKU 应用统一合格退货事件与重复规则。
- Coverage: missing identities, unmatched lines, unknown reasons, and uninspected items are published.覆盖：发布身份缺失、退货行未匹配、原因未知与未质检商品。
- Uncertainty: counts and interval or shrinkage method accompany percentages.不确定性：百分比同时展示数量及区间或收缩方法。
- Economic context: value, cost, recovery, contribution, and workload are kept separate from rate.经济背景：金额、成本、回收、贡献与工作量同退货率分开。
- Change log: launches, price, content, promotion, policy, supplier, and fulfillment changes are dated.变更日志：上市、价格、内容、促销、政策、供应商与履约变化均标日期。
Avoid nine SKU return-rate mistakes避免九个 SKU 退货率错误
- Treating a customer-selected reason as a verified root cause.把客户选择的原因当作已核验根因。
- Combining customer reason, observed condition, disposition, and refund outcome in one field.把客户原因、观察状态、处置与退款结果混在一个字段。
- Changing code labels without versioning or remapping historical records.更改代码标签却不进行版本化或映射历史记录。
- Ranking percentages without counts, eligible denominators, value, or uncertainty.只按百分比排序，不展示数量、合格分母、金额或不确定性。
- Comparing products, channels, or periods with different question wording and missingness.比较问题措辞与缺失程度不同的商品、渠道或期间。
- Discarding “other,” free text, multi-reason, changed, or unknown responses.丢弃“其他”、自由文本、多原因、已更改或未知回答。
- Acting on correlation before reviewing cases and testing a mechanism.在检查案例并测试机制前就依据相关性行动。
- Treating a parent product or item-group ID as if it were one sellable SKU.把父商品或商品组 ID 当作一个可销售 SKU。
- Sorting small samples by point estimate without counts or uncertainty.不展示数量或不确定性，按点估计排序小样本。
Do not repair identity problems with a dashboard filter. Quarantine ambiguous joins, fix the catalog history, and rerun the metric before making a merchandising, supplier, or product decision.
不要用看板筛选器修补身份问题。应隔离模糊关联，修复目录历史，再重新计算指标，然后才做商品运营、供应商或商品决策。
Prioritize one material SKU-reason cohort优先处理一个重要 SKU—原因群组
Filter to mature SKUs with adequate fulfilled units, then rank with a declared combination of excess returns, contribution loss, customer severity, evidence coverage, and actionability. Review representative cases, state the mechanism and alternatives, and run one reversible change with matched exposure and guardrails.
筛选履约件数充分的成熟 SKU，再按声明的超额退货、贡献损失、客户严重度、证据覆盖与可行动性组合排序。复核代表性案例，陈述机制与替代解释，并用匹配曝光与护栏运行一项可逆变更。
| Signal信号 | Evidence to check待检查证据 | Safe next step安全下一步 |
|---|---|---|
| One SKU exceeds matched variants一个 SKU 高于匹配变体 | Identity, maturity, attribute mapping, reason, inspection, content exposure身份、成熟度、属性映射、原因、质检、内容曝光 | Test one attribute-specific mechanism测试一个属性相关机制 |
| Many SKUs rise together多个 SKU 同时上升 | Metric/pipeline version, policy, promotion, channel, fulfillment, season指标/管道版本、政策、促销、渠道、履约、季节 | Diagnose shared change before SKU edits修改 SKU 前诊断共同变化 |
| High-rate SKU has negligible exposure高退货率 SKU 暴露极小 | Counts, interval, severity, value, strategic role数量、区间、严重度、金额、战略作用 | Monitor or aggregate unless harm is severe除非伤害严重，否则监控或汇总 |
Prepare a SKU-level return evidence file准备 SKU 级退货证据文件
Export stable variant key; current and legacy SKU; platform variant, parent, and item-group IDs; effective dates; size, color, material, and other attributes; eligible fulfilled quantity; cohort and maturity; return line and quantity; reason; inspection; disposition; refund; revenue; cost; recovery; contribution; currency; promotion; policy; channel; and change events. Return Compass can compare governed SKU cohorts; it cannot repair ambiguous identity or infer cause from a rate.
导出稳定变体键；当前与历史 SKU；平台变体、父级与商品组 ID；生效日期；尺码、颜色、材质及其他属性；合格履约数量；群组与成熟度；退货行及数量；原因；质检；处置；退款；收入；成本；回收；贡献；币种；促销；政策；渠道与变更事件。逆向罗盘可比较受治理 SKU 群组，但不能修复模糊身份，也不能从退货率推断因果。
Open Return Compass打开逆向罗盘 →SKU Return Rate FAQSKU 退货率常见问题
It is qualifying returned units for one stable sellable SKU divided by eligible fulfilled units for that SKU in the same mature cohort, expressed as a percentage.它是同一成熟群组中，一个稳定可销售 SKU 的合格退回件数除以该 SKU 的合格履约件数，并表示为百分比。
Units or order lines usually preserve partial returns at sellable-variant level. If an order-based metric is used, name it separately and do not compare it directly with a unit rate.件数或订单行通常能保留可销售变体层的部分退货。如果使用订单指标，应单独命名，不能与件数退货率直接比较。
Show returned and fulfilled counts plus a confidence interval or shrinkage estimate, require minimum exposure, or roll up to a stable parent/attribute cohort before ranking.展示退回与履约数量及置信区间或收缩估计，要求最小暴露，或先汇总到稳定父级/属性群组再排名。
Some systems may permit it, but duplicate or reused SKUs create ambiguous inventory and return joins. Use a unique stable variant key and govern aliases and effective dates.某些系统可能允许，但重复或复用 SKU 会造成库存与退货关联模糊。应使用唯一稳定变体键，并治理别名与生效日期。
Include counts, mature denominator, uncertainty, returned value, contribution loss, reason and inspection coverage, product hierarchy, variant attributes, data-quality flags, and dated changes.应包含数量、成熟分母、不确定性、退回金额、贡献损失、原因与质检覆盖、商品层级、变体属性、数据质量标记与带日期变更。
Sources, evidence labels, and limitations来源、证据标签与限制
- Shopify Help Center: Using SKUs to manage inventory — Official guidance that SKUs are internal inventory and reporting identifiers and should be unique for each product variant.Shopify 帮助中心：使用 SKU 管理库存——官方指南说明 SKU 是内部库存与报告标识，每个商品变体应使用唯一 SKU。
- Shopify Help Center: Variants — Official documentation that option-value combinations such as size and color form product variants with variant-level inventory.Shopify 帮助中心：商品变体——官方文档说明尺码、颜色等选项值组合构成商品变体，并可在变体层管理库存。
- Google Merchant Center Help: Item group ID — Official guidance distinguishing a unique product ID from a shared item-group ID used to group variants and align them with landing-page choices.Google Merchant Center 帮助：商品组 ID——官方指南区分唯一商品 ID 与用于聚合变体并对齐落地页选择的共享商品组 ID。
- Microsoft Learn: Sales returns in Dynamics 365 Supply Chain Management — Official documentation for customer-selected reason codes, reason groups, return-line references, disposition actions, and inventory/credit implications.Microsoft Learn：Dynamics 365 供应链销售退货——客户选择原因码、原因组、退货行关联、处置动作以及库存/贷项影响的官方文档。
- Microsoft Learn: Return reason codes and disposition codes — Official distinction between why a customer requests a return and the condition/action assigned after physical inspection.Microsoft Learn：退货原因码与处置码——官方区分客户申请退货的原因与实体质检后分配的状态/动作。
Evidence statement: Official commerce and product-data documentation supports stable SKU and variant identity, return-workflow stages, and evidence separation. Shopify and Google documentation is used only to establish identifier and variant principles, not to validate a retailer’s catalog or metric. Examples are synthetic. No customer result, universal threshold, causal claim, or guaranteed improvement is asserted. Sources were reviewed September 15, 2026; named subject-matter review is required before publication.证据声明：官方商业与商品数据文档支持稳定的 SKU 和变体身份、退货工作流阶段与证据分离。Shopify 与 Google 文档仅用于确立标识符和变体原则，不验证零售商目录或指标。示例为模拟。本文不声称客户结果、通用阈值、因果结论或保证改善。来源核验于 2026 年 9 月 15 日完成；发布前需要具名领域审核。
Make SKU identity trustworthy before ranking rates对退货率排名前先让 SKU 身份可信
Calculate SKU return rate on one stable sellable-variant identity with a mature fulfilled-unit denominator and explicit event rules. Publish counts, uncertainty, coverage, value, and contribution beside the rate. Use parent and attribute rollups for context, then investigate one material SKU-reason cohort with case evidence and a reversible test. A precise percentage on ambiguous identity is not a reliable decision signal.
在一个稳定可销售变体身份上，用成熟履约件数分母与明确事件规则计算 SKU 退货率。退货率旁同时发布数量、不确定性、覆盖、金额与贡献。使用父级和属性汇总作为背景，再用案例证据与可逆测试调查一个重要 SKU—原因群组。基于模糊身份的精确百分比并非可靠决策信号。