Completeness and quality of product master data
From the number of SKUs, filled cards, duplicates, invalid barcodes and stale records calculates four quality dimensions, an overall score by your weights and the cost of errors as a range.
At a glance
- What question does it answer?
- How ready the SKU reference book is for work and what errors cost.
- What data is needed?
- The number of SKUs, filled cards, duplicated codes, invalid codes, stale records.
- How to read the result?
- On the default data (12,000 SKUs, 10,200 filled, 120 duplicates, 60 invalid, 900 stale) completeness is 85%, uniqueness 99%, validity 99.5%, timeliness 92.5%, the overall score is 92.2% with weights 40, 20, 20, 20. Problem SKUs are from 1,800 to 2,880; the cost of errors is from 360,000 to 576,000 per period at 8% incidents and 2,500 per incident. Example on the default values: Overall quality score: 92.2%; Completeness: 85%; Uniqueness of codes: 99%; Validity of barcodes: 99.5%. The figures are illustrative: enter your own to see your result.
Why and how to reason
Errors in master data (the product register) are cheapest to fix at the moment they appear, not after they have become failures in orders, stock and marking. Quality is measured by dimensions. Completeness: how many SKUs have all mandatory attributes filled. Uniqueness: no duplicate barcodes. Validity: the barcode is correct. Timeliness: the record is updated on time. Of 12,000 SKUs: completeness 10,200 ÷ 12,000 = 85%, uniqueness 99%, validity 99.5%, timeliness 92.5%. The overall score is the average with your weights: 0.4 × 85 + 0.2 × (99 + 99.5 + 92.5) = 92.2%. The number of problem SKUs lies in a range: from the largest single cause (1,800) to the sum of causes, because one SKU can have several problems. Multiplying by the failure share and the cost of a failure gives the cost of errors: from 360 thousand.
When to use it
It is useful before introducing an accounting system, before marking and before changing a data supplier, and as a quarterly control. Do not use it as a guarantee of quality: only four dimensions out of six are tested (accuracy and consistency between systems are checked by reconciliation). Assumptions: you set the weights yourself, there is no single norm, problem SKUs are counted by the largest cause and by their sum.
How to read the result
An overall score of 92% looks good, but it hides the weak link: completeness 85%. The calculation names the weakest dimension, and that is where to begin cleaning. The range of problem SKUs (1,800 to 2,880) is more honest than one number: refine the real count by checking overlaps. Compare the cost of errors (360 to 576 thousand per period) with the cost of cleaning: if cleaning is cheaper, begin today. Set a target for the overall score and track it by period. It is worth fixing an owner and a deadline for each dimension: without an owner the register will deteriorate again and the score will return to where it was.
Common mistakes
Typical trap: counting fields with a dash or “n/a” as filled, mixing an empty barcode with an invalid one, taking weights “equal” without reason, not updating the register after a supplier change. Do not take a 100% score as proof of quality: check the accuracy of the data too, not only its form. Do not count duplicates by barcode alone: one product with two cards and different codes is also a duplicate, but only a comparison of names and attributes finds it.
What data is needed?
- The number of SKUs, filled cards, duplicated codes, invalid codes, stale records.
- The dimension weights and your target score.
- The incident share and the cost of one incident from the incident log.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| SKUs in the reference book | pcs | 12,000 | All active SKUs. The example value is illustrative. |
| SKUs with all mandatory attributes | pcs | 10,200 | Composition, color, size, category, season code, customs code and other fields you named mandatory. The example value is illustrative. |
| SKUs with a duplicated barcode | pcs | 120 | One code on two or more SKUs. The example value is illustrative. |
| SKUs with an invalid or empty barcode | pcs | 60 | Fails the check digit or is empty. The example value is illustrative. |
| SKUs not updated on time | pcs | 900 | You set the update period: it depends on the kind of record. The example value is illustrative. |
| Weight of completeness | % | 40 | A relative weight; the weights need not add up to 100. The example value is illustrative. |
| Weight of uniqueness | % | 20 | A relative weight. The example value is illustrative. |
| Weight of validity | % | 20 | A relative weight. The example value is illustrative. |
| Weight of timeliness | % | 20 | A relative weight. The example value is illustrative. |
| Your target for the overall score | % | — | Set your own target from past failures; empty: no color is assigned. There is no single standard in Russia: enter your own figure. |
| Share of problem SKUs that caused an incident in the period | % | 8 | An incident: a wrong shipment, a platform rejection, a stock mismatch. Take it from the incident log. The example value is illustrative. |
| Cost of one incident | ₽ | 2,500 | People’s time, a platform fine, a re-shipment, excl. VAT. The example value is illustrative. |
How it is calculated
| Step | Formula |
|---|---|
| Dimension | completeness = filled ÷ all; uniqueness = 1 − duplicates ÷ all; validity = 1 − invalid ÷ all; timeliness = 1 − stale ÷ all |
| Overall score | sum (weight × dimension) ÷ sum of weights |
| Problem SKUs | from the biggest cause to the sum of causes (not above all) |
| Cost of errors | problem SKUs × incident share × cost of an incident |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Overall quality score
- 92.2 %
- Completeness
- 85 %SKUs with all mandatory attributes ÷ all SKUs.
- Uniqueness of codes
- 99 %
- Validity of barcodes
- 99.5 %
- Timeliness
- 92.5 %
- Problem SKUs: from
- 1,800 · up to 2,880Lower bound: the biggest single cause; upper: the sum of causes, not above all SKUs.
- Cost of errors per period: from
- 360,000 ₽ · up to 576,000 ₽
Quality dimensions
| Dimension | Value | Weight |
|---|---|---|
| Completeness | 85 % | 40 |
| Uniqueness | 99 % | 20 |
| Validity | 99.5 % | 20 |
| Timeliness | 92.5 % | 20 |
The weakest dimension is “Completeness”: start the clean-up with it. These are four dimensions of six (accuracy and consistency between systems are the other two): they are checked by reconciliation, not by one formula.
The data-quality dimension definitions are common practice. Thresholds are not set by any Russian standard: set your own.
Nuances and limits
- The overall score hides a weak dimension: look at it too.
- Causes overlap, so the cost is a range.
- You set the thresholds: there is no single standard in Russia.
- Accuracy and consistency between systems are checked by reconciliation.