Syntha Lab RU

Order data quality

What mismatches in orders cost and whether fixing them pays off.

For a brand Processes and data

Open interactively →

About the topic

Every error in an order costs money, even if it was fixed quickly. Purchasing entered the wrong item, the warehouse shipped the wrong size, finance issued an invoice with a different price: each such case takes people’s time and sometimes turns into a deduction from payment, a buyer discount or a lost sale. This topic teaches you to measure the cost of such discrepancies and to answer whether fixing the process will pay back.

The reasoning rests on a simple sample. Take orders for one or two months and compute three numbers: the share of orders with a discrepancy, the average time to fix and the direct loss per order. Say of 1,000 orders a month 8% (80) have a discrepancy. Fixing takes 1.5 hours, an hour of a person’s work costs RUB 600, so RUB 900 of time. The direct loss (buyer discount, chain penalty, return, repeat delivery) averages RUB 700. In total RUB 1,600 per problem order, RUB 128,000 a month, about RUB 1.54 million a year. This is the price of “changing nothing”.

Now the fix. Suppose a regulation, a single item master and integration of the warehouse system cost RUB 600,000 and remove 60% of discrepancies: the other 40% remain because part of the errors comes from the buyer or the supplier. The saving is 0.6 × 1.54 ≈ RUB 0.92 million a year, payback 600 / 922 ≈ 0.65 years, less than eight months. Had you assumed discrepancies vanish entirely, the saving would look like RUB 1.54 million and payback 4.7 months: too optimistic.

Where discrepancies come from is hinted by the process itself: a collection calendar where each stage has a date and an owner; a single item identifier (GTIN, SKU) in all systems; one source of prices; order statuses seen by purchasing, warehouse and finance. Without a common master each system lives its own life and an error is guaranteed. The decisions of the topic: which discrepancies are the most expensive and where to start, how much to invest in the fix, how to measure the effect. Traps: taking the discrepancy share without a sample, overstating the expected reduction, counting only reconciliation time and forgetting deductions and discounts, considering the process fixed forever. Start with a sample of orders and split discrepancies by type and cost.

Key ideas

A discrepancy as a cost item

Discrepancies cost different amounts. Say, of 80 problem orders, 40 have the wrong size at RUB 1,000 each, 25 the wrong price at RUB 2,400 and 15 a missed deadline at RUB 3,000: RUB 145,000 in total. Size is the most frequent (50% of cases), but price takes the most money (RUB 60,000, 41%). You start fixing what is most expensive in money, not in count.

A sample instead of a feeling

The share of discrepancies cannot be taken by habit. Orders for 1-2 months are taken, faulty ones marked, and the share, average time and average losses computed. There is no normal value of the share for Russia; the benchmark is your own history and the goal is reduction, not matching someone else’s figure.

Not all losses are visible

Hidden losses are looked for in three places: deductions from payment under discrepancy reports, discounts and credit notes to buyers, and sales lost to short supply. For example, 30 deductions a month at RUB 1,200 each make RUB 36,000, which go through the books as “other expenses” and are not visible in the cost of the error. Ask accounting where they sit.

The unfixable remainder

Check what share of reduction pays back the investment within a year. Discrepancies cost RUB 1.54 million a year, the investment is RUB 600,000: it pays back within a year at a reduction of 600 / 1,540 ≈ 39%. At an 80% reduction the payback is about 5.8 months (600 / 1,232 × 12), at 40% about a year. If a realistic target is below 39%, the project in this form does not pay back within a year: reduce its scope.

A single master and calendar

The causes of discrepancies are most often in data: one item is named differently in purchasing and in the warehouse, prices are stored in several places. A single item master (GTIN, SKU), one source of prices and a collection calendar with dates and owners remove a whole class of errors. First order in data, then automation.

Stage readiness is all critical tasks plus a share of the rest, not an average percent

A stage is ready when all critical tasks are done and not less than 80% of the rest. The average over the route hides the blocker: 51% of all tasks and 62% of critical ones are done, yet only 3 of 8 stages are ready. Stage 3 stands exactly on the threshold: 8 of 10 tasks and all 4 critical ones. The stage “Production and the first batch” is not ready: 7 of 12 tasks (58%), 4 of 6 critical; 3 tasks are missing before the threshold, and 2 of them are critical. The transition condition itself is described in the term “Launch stage gate”, here it is shown where the blocker stands.

Formula: Ready = all critical done and not less than the threshold of the rest; missing = the larger of (needed by threshold − done; critical − critical done)

What to do: Read the route by its blocker stage: close the critical tasks first and only then count the share of the rest.

Think first: The route is 51% done and the critical tasks are 62% done. Can the launch be considered half ready?

Answer: No: 3 of 8 stages are ready, and production stopped on critical tasks. An average percent does not say which stage blocks.

How to work through it

  1. Take a sample of orders for 1-2 months and mark those with a discrepancy.
  2. Split discrepancies by type: item, size, price, date, documents, marking.
  3. For each type compute the share of orders, average fix time and direct loss.
  4. Get the cost of discrepancies per month and per year and find the most expensive types.
  5. Describe the fix: what we do, what it costs, what share it removes, what remains.
  6. Compute the saving and payback with the unfixable remainder.
  7. Assign owners, dates and the metric by which you will check the effect after 2-3 months.

Where to start

Take a sample of orders over one to two months, count the share with a mismatch, the average hours to fix one and the direct loss per order, and enter them in the model. Calculations of the topic in order: “Cost of order mismatches and payback of the fix”. See also (their primary topic is another one): “Starter brand model: first batch and cash to start”, “Label launch readiness: stages, critical tasks, the blocking stage and how many tasks are missing”.

Calculations in this topic

  • Cost of order mismatches and payback of the fix

    Estimates the yearly cost of mismatches in order data (people’s time, returns, discounts) and how fast fixing the process pays back.

    What data mismatches in orders cost a year and how fast fixing them pays back.

  • Starter brand model: first batch and cash to start

    Calculates the economics of a new brand from zero: net price and unit margin, break-even sales per month, the minimum first batch, the peak cash need and the result of the first batch by month. Samples, launch and fixed costs are entered separately so you can see what eats the start.

    What first batch is needed, how much cash the start will take and when it returns.

  • Label launch readiness: stages, critical tasks, the blocking stage and how many tasks are missing

    From a table of stages (how many tasks, how many done, how many of them are critical and how many critical are done) and a readiness threshold it counts which stages are ready, which stage blocks moving on and how many tasks are missing before it is ready. A stage is ready only when all critical tasks are done and not less than the threshold of the others. The calculation shows the order of work, not the quality of execution: “done” is marked by you.

    Which launch stages are ready, which one blocks moving on, how many tasks are missing before it is ready and how many critical tasks are left.

Self-check questions

  • What share of my orders passes with a discrepancy and on what sample was it measured?
  • How many hours and of which people does fixing one case take?
  • What direct losses does each discrepancy bring: deduction, discount, penalty, return?
  • Which types of discrepancies are the most expensive and the most frequent?
  • Do I have a single item master and one source of prices?
  • Does every calendar stage have a date and an owner?
  • What share of discrepancies will the fix not remove and why?
  • How will I know in two months that the process has improved?

Knowledge check

Multiple-choice questions. The correct answer and the explanation follow each question.

  1. There are 900 orders a month, 5% of them with discrepancies. Resolving one takes 2 hours at RUB 800 an hour, the direct loss per case is RUB 1,500. How much do the discrepancies cost per month?

    • A. RUB 72,000
    • B. RUB 67,500
    • C. RUB 139,500
    • D. RUB 2,790,000

    Correct answer: C. RUB 139,500. Discrepancies 900 × 5% = 45. One case costs 2 × 800 + 1,500 = RUB 3,100, in total 45 × 3,100 = RUB 139,500. 72,000 comes from counting only staff time (45 × 1,600), 67,500 only direct losses (45 × 1,500), 2,790,000 from multiplying the cost of a case by all 900 orders.

  2. Fixing the process costs RUB 1,432,000 one-off, support RUB 40,000 a month, losses from discrepancies RUB 139,500 a month, the fix removes 80% of the losses. How long until it pays back?

    • A. 12.8 months
    • B. 20 months
    • C. 14.4 months
    • D. 10.3 months

    Correct answer: B. 20 months. Monthly benefit = 139,500 × 0.8 − 40,000 = RUB 71,600; 1,432,000 / 71,600 = 20 months. 12.8 comes without deducting support (RUB 111,600), 14.4 from assuming the losses vanish completely (RUB 99,500 after support), 10.3 from ignoring both. In 12 months RUB 859,200 accrues, 40% less than the investment: over a one-year horizon the project does not pay back.

  3. Stage 3: 8 of 10 tasks, all 4 critical ones done, the threshold is 80%. Is the stage ready?

    • A. No: more than 80% is needed
    • B. Yes: 80% of tasks and all critical ones
    • C. No: two critical ones are not closed
    • D. It depends on stage 4

    Correct answer: B. Yes: 80% of tasks and all critical ones. A threshold of 80% of 10 is 8 tasks; 8 are done and the critical ones are closed. The condition includes a value equal to the threshold. Covered in: Stage readiness is all critical tasks plus a share of the rest, not an average percent.

  4. Production and the first batch: 7 of 12 tasks, 4 of 6 critical, the threshold is 80%. How many tasks are missing before readiness?

    • A. 5
    • B. 2
    • C. 3
    • D. 10

    Correct answer: C. 3. 80% × 12 = 10 tasks are needed; 7 are done, 3 are missing; 2 critical ones remain and are included in these tasks: the result is 3. Covered in: Stage readiness is all critical tasks plus a share of the rest, not an average percent.

  5. 51% of tasks are done, 62% of critical ones, 3 of 8 stages are ready. What blocks the route?

    • A. Stage 8: launch and first sales
    • B. A low average percent of readiness
    • C. Stage 1: idea and brand platform
    • D. Stage 4: production and the first batch, the first one not ready

    Correct answer: D. Stage 4: production and the first batch, the first one not ready. The blocker is the first stage not ready in order: stages 1–3 are ready, stage 4 is not, and the next ones depend on it. An average percent does not show the blocker. Covered in: Stage readiness is all critical tasks plus a share of the rest, not an average percent.

Common mistakes

  • Mistake: Taking the discrepancy share without an order sample

    What it costs: The process cost is guessed

    How to avoid: Measure on real orders for 1-2 months.

  • Mistake: Overstating the expected reduction

    What it costs: The investment pays back on paper but not in life

    How to avoid: Assume a conservative share.

  • Mistake: Counting only reconciliation time and forgetting deductions and buyer discounts

    What it costs: The cost is understated

    How to avoid: Add direct losses.

  • Mistake: Assuming the fixed process removes all discrepancies

    What it costs: The saving is overstated

    How to avoid: Keep a share that is unfixable.

  • Mistake: Automating chaos

    What it costs: Errors multiply faster

    How to avoid: First put order in the master and calendar, then automate.

  • Mistake: A launch is assessed by an average readiness percent

    What it costs: The blocker stage goes unnoticed and the next stages run idle Work on channels, calendar and buyers without ready production remains provisional: 31 tasks are left, 10 of them critical.

    How to avoid: Read the stages in order and look for the first one not ready Assign an owner to each critical task of the blocker and check readiness by stage once a week.

    What it looks like (illustrative figures): 51% of tasks and 62% of critical ones are done, but 3 of 8 stages are ready; stage 4 blocks, 3 tasks are missing before the threshold.

Russia and the EAEU

In Russia a significant part of discrepancies arises where operations meet documents and marking. Electronic document exchange (UPD, UKD for corrections) requires that the item, quantity, price and VAT in the document match the shipment; for marked goods Chestny ZNAK codes are added: an error in the GTIN, size or color in the card leads to refusal at acceptance, a chain penalty or a problem on withdrawal from circulation. So the item and GTIN master becomes the central element of the process.

Signing in electronic exchange under a power of attorney requires a machine-readable power of attorney (MChD), and an expired or wrong one stops the document and with it the payment. An online cash register under 54-FZ and return receipts are linked with stock accounting: a discrepancy between warehouse and register shows in marking.

Retail chains and marketplaces deduct for errors under their own rules: penalties for short delivery, label mismatch and missed acceptance deadlines are set by the offer or contract and change, take amounts from current terms. For accounting and inventory (1C and similar systems) the discipline of statuses matters: shipment, acceptance, return. A single collection calendar with dates and owners links purchasing, production, marking and sales.

In other EAEU countries discrepancies are joined by different documents and systems. A delivery between countries needs an import declaration and supporting documents (EAEU Treaty, Annex 18), indirect taxes are paid under the destination principle and invoices are issued in their own way. Electronic invoices in Belarus and Kazakhstan exist, but formats and deadlines are their own (indicative as of 14.11.2026): ask a local accountant. Light-industry marking in Kazakhstan starts on 1 December 2026 (indicative, check the order text), for the other countries the dates: indicative as of 14.11.2026, check the text of the act; for deliveries to Russia from the EAEU the Russian importer orders codes under a tolling scheme. As of 14.11.2026, VAT rates by country are on the page “EAEU and CIS countries”; check them with an accountant.

Categories and audiences

Apparel
On apparel: the longest chain of identifiers (model, color, size), so there are more errors and the cost of one discrepancy is small but frequency is high. The costliest is usually a wrong size or color in marking. A single master by color-size and an acceptance check give a quick effect.
Footwear
In footwear: the size run is longer, with half sizes and widths, and the box and the pair must match, so a size error is especially costly and leads to returns. A discrepancy between label, box and marking often arises at repacking. A control at warehouse exit pays back.
Accessories
For accessories: many items, small batches, and completeness (pouch, box, gift packaging) is often lost and brings claims. There are no sizes, and the error is more often in color, set or price. A simple completeness check at packing cuts discrepancies faster than large projects.
Kids
If it is children’s wear: besides size and color, age marks, certificates and marking matter, and a document error on children’s goods leads to refusal at acceptance more than for adults. The school and holiday season compresses deadlines, so the price of delay is higher. A calendar and owners are especially useful here.
Men
In the men’s segment orders more often run on recurring SKUs, so the catalog is put in order once and serves from season to season. The main money risk is with chains: a penalty for short-shipping a basic size. The check of basic-size availability before shipping is automated first: a rule “not in stock, stop shipment” is cheap and catches expensive errors.
Women
In the women’s segment one style is entered in several colors and sizes, and each combination gets its own SKU, so the number of cards grows faster than the number of models, and an error in one value repeats in all of them. Check cards before shipping against a template: color, size run, price. Allow more time for entering SKUs at the start of the season: discrepancies arise right there.

Questions this topic answers

  • What do mismatches in order data cost?
  • Will fixing the process pay back?
  • What share of orders has a mismatch?
  • Which mismatches are the costliest and which to fix first?

Key measures and terms

Study first

What next

Next step: “Collection lifecycle and the critical path”. Related: “Wholesale terms and credit”, “Cash and working capital”, “Systems landscape and data exchange”.

Documents on this topic

Open interactively →