Voice of Customer: return and complaint reasons → actions
VoC gathers the reasons for returns, complaints, reviews and surveys into one table, estimates their cost and turns them into actions for design, production, content and logistics. It decides which reason to remove first because it is the most expensive, and who is responsible for the fix.
Why this document
Returns and complaints arrive from different places and look like noise until they are brought together and valued in roubles. The product head, the technologist, the content manager and the logistician read this document: it shows each of them their share of the losses, for example size grid, photos or packaging. It is needed to spend effort on the reason that costs most rather than on the loudest one. Without it returns are counted as losses and not as a signal, and the same problem repeats every season.
At a glance
- What is this document and why is it needed?
- VoC gathers the reasons for returns, complaints, reviews and surveys into one table, estimates their cost and turns them into actions for design, production, content and logistics. It decides which reason to remove first because it is the most expensive, and who is responsible for the fix.
- Who owns the document and how often is it rebuilt?
- Brand: analyst or head of service, a weekly roll-up and a monthly review with design and production; the commercial director decides on actions. Retailer: category analytics and service, a review with suppliers by brand and category.
- What mistakes are the most common?
- The return reason is free text → it cannot be counted or aggregated → a choice list and definitions. Priority by return count without cost → cheap reasons are solved and expensive ones remain → count loss in roubles. Data does not reach design and production → the same defect in the new season → assign an owner and a term to each action.
For whom
- Brand
- The brand can fix the cause: change patterns and size chart, color in shoots, fabric, packaging. So the “reason → owner → action → term → effect check” table is the main output. The brand matches reasons by styles and batches and sees where the problem is construction and where it is delivery.
- Retail
- The retailer influences selection, presentation and service and passes data to suppliers by brand. For a retailer the reason “expectation did not match the card” (its own card) and “size did not fit” (by brand) matter. A supplier summary becomes an argument in negotiations: part of the losses on defects and sizes is claimed from the brand.
What it consists of
Reason catalog
- A single list: size, color and photo, quality, did not like, delivery, multi-size fitting, a lower price after purchase, a warehouse error. No more than 10-12 reasons, each with a definition.
- Brand: the company-side cause more often lies in the product (patterns, fabric, photos, size description), so the action closes with a model change, not a reply script.
- Retailer: the company-side cause more often lies in the process (warehouse, delivery, price after purchase, description of a third-party brand), so the action closes with a protocol and data to the brand.
- Two levels: the customer reason (what they chose) and the company cause (what went wrong); matched in a separate column.
Reason → action table
- Columns: reason, number of returns, share, cumulative share (Pareto), cost of one return, loss in roubles, hypothesis, action, owner, term, control indicator.
- Brand: slices by model, size, batch, factory and channel (own D2C versus wholesale); the channel matters because the wholesale customer returns under the contract, not for the shopper reason.
- Retailer: slices by brand, supplier, store, region, channel (store versus online) and delivery method; brands of one category are compared by return rate to choose whom to expand.
- Cuts: style, size, batch, channel, region; they show where the reason is local.
Handover and control
- Brand: data goes to design and patterns, production and quality, a claim to the factory and a revision of the size grid for the next season; the aim is to remove the cause in the product.
- Retailer: data goes to the category manager and buying: no repeat, review of the brand role in the category, a claim to the brand and return terms under the contract; the retailer does not change the product, it chooses and reports.
- To whom: design (patterns and size chart), production (quality and batches), content (photos and copy), logistics (terms), service (scripts).
- Control: change in the reason share after 1-2 sales cycles; an action without effect measurement is not closed.
How to read the result and which decisions follow
- The first cause “Size did not fit” gives 37.6% of returns, the first two 58.4%; total loss RUB 816,630 at a 28.1% return rate. Decision: start with causes at the head of the chart, where one rule removes a large share of losses.
- Brand: the most expensive categories by loss per order: “Footwear” RUB 263.8, the cheapest “Accessories” RUB 64.3. Decision: for footwear and women invest in size data and the last, for accessories in hardware control, for kids in height instead of age.
- Retailer: some causes are yours (late delivery, fitting several sizes), some the brand (defect, not as in photo). Decision: keep a cause directory with an owner and return confirmed defect returns to the brand with documents.
- Brand: the “did not like” cause is not cured by service, only by the range; models with a high share of such returns should not be repeated.
- Retailer: fitting several sizes inflates returns artificially; limit the number of sizes per order or charge return delivery only for serial returns, checking against consumer rights.
Control checks
- Check: the sum of causes of 1010 returns gives 28.1% of 3600 orders; the cumulative share of the last cause equals 100% (code).
- Brand: count defects and unsellability from the inspection of the return, not from the cause in the client request.
- Retailer: clients confuse “size did not fit” and “did not like”; introduce a mandatory choice of cause when a return is filed.
- Brand: add lost sales and reputation separately to the cause losses; the table has direct costs only.
- Retailer: compare return data by order cohorts, not by return date, otherwise seasonal peaks distort the rate.
Where the data comes from
| Data source | Fields |
|---|---|
| Returns records | Order, SKU, size, reason, date, channel, amount, item condition. |
| Reviews and questions on platforms | Text, rating, style; keywords of size and quality complaints. |
| Surveys and service | NPS and CSAT answers, requests by type, store comments. |
How to build it step by step
- Step 1. Agree the reason catalog and definitions; one reason is mandatory at return (a list choice, not free text).
- Brand: Step 2. Link reasons to style, size, batch and channel; without it a local defect cannot be found.
- Step 3. Collect reviews and questions; classify by the catalog (manually by sample or by keywords).
- Retailer: Step 4. Compute the cost of one return by reason: logistics + handling + loss of value of the unsellable part.
- Step 5. Build a Pareto table: reasons by descending count and loss; mark the 2-3 reasons giving most of the loss.
- Step 6. For each reason write a hypothesis and an action with an owner and a term.
- Brand: Step 7. Find styles and sizes with a return share above the category median; add them to the pattern and quality check list.
- Retailer: Step 8. Pass data to buying and suppliers: a list of defect and size claims with photos and batches.
- Step 9. After 1-2 sales cycles measure the change in the reason share and close or revise the action.
Formulas
| Measure | Formulas | Note |
|---|---|---|
| Return rate | Return % = Returned orders ÷ Bought-out and returned orders × 100 | The definition must be the same for all channels; marketplaces count buyout their own way. |
| Cost of a return | Return = Logistics + Handling + Cost × Unsellable share | The unsellable share is high for defects and underwear, low for size. |
| Loss by reason | Loss = Number of returns by reason × Cost of a return | Excludes lost revenue; it can be estimated separately. |
| Cumulative share (Pareto) | Cumulative = Sum of reason shares up to and including this one | The “80/20” split is not a law; look at the real curve. |
Formulas in notation
- Loss
Loss = Returns × c
The loss of the period by cause or category.
Worked example
Exercise: monthly returns across five categories and their cost (conditional assumptions)
| Category | Orders | Return rate, % | Returns | Size, % | Defects, % | Cost per return, RUB | Loss, RUB | Loss per order, RUB | Main action |
|---|---|---|---|---|---|---|---|---|---|
| Women | 6,000 | 22 | 1,320 | 41 | 5 | 755 | 996,138 | 166 | Garment measurements and fit |
| Men | 2,400 | 14 | 336 | 33 | 6 | 756 | 254,050 | 105.9 | Pattern and batch check |
| Kids | 1,600 | 12 | 192 | 36 | 7 | 741 | 142,186 | 88.9 | Height instead of age, safety |
| Footwear | 2,000 | 26 | 520 | 52 | 9 | 1,015 | 527,634 | 263.8 | Last and half sizes |
| Accessories | 1,000 | 8 | 80 | 3 | 12 | 804 | 64,301 | 64.3 | Hardware control |
| Total | 13,000 | 18.8 | 2,448 | 811 | 1,984,308 | 152.6 |
Of 13000 orders 2448 came back (18.8%), loss RUB 1984308. Cost per return = logistics RUB 420 + handling RUB 90 + the price loss on the unsellable part (a defect is lost entirely, others by the unsellable share). Footwear loses RUB 263.8 per order against RUB 64.3 for accessories: a high return rate and an expensive pair. Priority is set in roubles, not only by count. All figures are exercise assumptions.
Figures
Data as a table
| Measure | Loss |
|---|---|
| Size did not fit | 230k |
| Did not like | 147k |
| Not as in photo, color | 106k |
| Fitting several sizes | 64.5k |
| Quality, defect | 241k |
| Late delivery | 28.4k |
Return losses by cause. Causes in descending order of returns; losses are computed per return including the unsellable share.
Data as a table
| Measure | Cumulative share |
|---|---|
| 1 | 37.6 |
| 2 | 58.4 |
| 3 | 74.3 |
| 4 | 85.1 |
| 5 | 95 |
| 6 | 100 |
Cumulative share of returns (Pareto). How many causes explain most returns: the first two causes give more than half. Horizontal axis: cause number in descending order.
Data as a table
| Measure | Loss per order |
|---|---|
| Women | 166 |
| Men | 106 |
| Kids | 88.9 |
| Footwear | 264 |
| Accessories | 64.3 |
Loss per order by category. How much returns cost per order of a category: footwear is the most expensive at a 26% return rate.
Specifics by category
- Apparel
- Main reasons: size and fit, color, “did not like”, fitting duplicates (ordering two or three sizes). Fitting duplicates are not an error but behavior: they are counted as a separate reason and subtracted from the target for reducing size returns. The Pareto usually shows 2-3 reasons giving most of the loss; actions: garment measurements in the card, fit on models of different sizes, a “runs small” hint. Separate core and fashion: for core the reason is stable, for fashion it changes every week.
- Footwear
- Size and last give more than half of returns (assumption 52%), and the cause is often at model level (a narrow last) rather than size. Create separate codes “narrow”, “wide”, “tight at the instep”, “too big” and compare by insole length in mm and the EU, UK, US, Mondopoint systems. Keep the Pareto by model and last, not only by size number. Cost per return is higher: the pair is more expensive and wear marks make it unsellable. Sole and adhesive defects go as a claim to the factory.
- Accessories
- There are no sizes (except belts and headwear), so size returns are almost absent, and leading reasons are “not as in the photo” (scale, color, metal glare), hardware defects and gift returns “did not like”. The Pareto is kept by hardware and supplier: one batch of clasps can produce the whole defect share. Jewelry has a separate procedure: fineness, hallmark and the list of goods without a right of exchange are checked under the current rules (unverified). After holidays returns grow seasonally.
- Kids
- The decision has two sides: the parent buys, the child wears, so the customer reason is recorded from the parent and “the child grew” is logged separately (not a defect). Size is set by height (6 cm step), not age. Safety complaints (rash, odor, small parts, composition) are kept outside the Pareto: any signal is escalated regardless of share, and for items for children under one year compliance with TR CU 007/2011 is checked. The school peak distorts shares: compare season with season.
- Men
- The return rate is lower (14% against 22% for women in the example), and reasons concentrate on fit (sleeve, length, shoulder) and fabric quality. Models repeat from season to season, so batch and factory are chosen as the slice: growth of returns on an unchanged model means a batch defect. Compare the model rate with the category median, not the company total: the total hides outsiders in men. The size curve is narrower, fewer fitting duplicates.
- Women
- Returns are more numerous because of vanity sizing, many silhouettes and the habit of ordering several sizes. The best slice is size × model: if size S consistently “runs big”, it is a signal to fix patterns or the hint, not growth of claims against the customer. Fashion loses price quickly, so a return at season end costs more: the cost of a return includes the loss from markdown. Color and “not as in the photo” take a noticeable share, so check retouching and color calibration.
Common mistakes
- The return reason is free text → it cannot be counted or aggregated → a choice list and definitions.
- Priority by return count without cost → cheap reasons are solved and expensive ones remain → count loss in roubles.
- Data does not reach design and production → the same defect in the new season → assign an owner and a term to each action.
- The action effect is not measured → unclear what helped → a control indicator and review term in the table.
- Reviews are read selectively, the loudest first → a distorted picture → classify the whole flow or a random sample.
Russian specifics
- Collecting reviews and surveys tied to a customer is personal data processing: consent, purpose and retention period; use anonymized data for aggregated analysis. Rule: Law 152-FZ of 27.07.2006: Federal Law of 27.07.2006 No. 152-FZ on personal data.
- Retailer: The “defect” reason must match accounting under consumer rights (examination, term): assign the “confirmed defect” status only after a check. Rule: Law 2300-1 of 07.02.1992: Law of 07.02.1992 No. 2300-1 on consumer rights protection.
- Brand: For kids goods, safety complaint data may need to be recorded under the technical regulation requirements; check details with the certifier. Rule: CU Commission decision 797 of 23.09.2011: TR CU 007/2011 on safety of products for children and adolescents (Customs Union Commission Decision of 23.09.2011 No. 797) (rule check not completed).