Master data cleansing prioritization: using ABC analysis in waves
Cleanse everything or only A items? Learn how to cross ABC analysis with criticality, usage and risk to set waves and depth for master data cleansing.
Not everything at once, and not only the A items. Use the ABC value analysis as a starting point, cross it with operational criticality, recent usage and risk, and decide first what leaves the database. Then run the cleansing in waves within the domain, with a different depth for each class of item.
"Let's cleanse the whole database." "No, only the A items, that's where the money is." I have seen this conversation many times in kickoff meetings for master data and SAP projects. Both people are wrong, each in a different way.
Whoever wants to cleanse everything finds out in month three that the team is standardizing descriptions of items nobody has bought in years. Whoever wants only the A items forgets the cheap pump part: it sits in class C because it costs little and moves little, and it is the one that stops the line when it is missing or when nobody can find it in the master data. Value matters. It just does not tell the whole story of a master data record.
What is ABC analysis and where does the definition come from?
ABC analysis ranks items by relative importance. SAP's inventory controlling documentation (ABC Analysis, SAP Help Portal) states: "The ABC analysis allows you to classify materials according to their importance. The main criterion is either the usage value or the requirements value."
Usage value is consumption in the period, valuated at the current price. Requirements value is future requirements from MRP, valuated at the material master price. Class A holds the important items (high value), class B the less important ones (medium value) and class C the relatively unimportant ones (low value). The same page notes that, in general, the higher the value, the greater the chances for improvement.
The company decides where to draw the lines. The page Performing an ABC Analysis describes four strategies: percentage of value, absolute value, number of materials as a percentage and absolute number of materials. The system proposes 90 days in the past (usage) or in the future (requirements), but any period can be chosen. This documentation covers SAP S/4HANA release 1610; options and screens may differ in your release.
For inventory, that settles most of the discussion. For master data cleansing, something is missing.
Cleanse everything or only the A items: which one is right?
Neither, and for the same reason: both treat every item in a class as if they were all the same.
Cleansing the whole database at once mixes active items with dead ones. The team spends hours on records that should be blocked, the project loses pace and leadership stops seeing results before it sees the end.
Cleansing only the A items ignores three groups that usually land in class C or do not show up in the analysis at all:
- Maintenance parts that stop operations when missing, but cost little and move little.
- Safety items or items that require traceability.
- Items with sensitive tax fields (in Brazil, for example, the NCM, the Mercosur goods classification code used on Brazilian tax documents) or items used in interfaces with other systems, where a master data error causes trouble outside the warehouse.
And there is the opposite case: an item with no usage at all will not be in class A, but it should not enter the cleansing either until someone decides whether it still exists.
Is ABC analysis enough to prioritize master data?
It is not enough. It is a necessary criterion, because it shows where the money flows. For master data, it needs to be crossed with other signals. This is the matrix I use as a starting point:
| Criterion | Question it answers | Where it usually comes from |
|---|---|---|
| Value (ABC analysis) | Where does purchasing and consumption money flow? | Usage value, MRP requirements, purchase history |
| Operational criticality | If it is missing or bought wrong, does anything stop? | Maintenance, production, list of critical spare parts |
| Recent usage | Did the item have movements, orders or requisitions in the period? | Stock movements, purchase orders, requisitions |
| Risk | Does an error in this record create tax, safety or compliance exposure? | Tax, quality, occupational safety |
| Effort | How much does it cost to cleanse this group at the required depth? | Database assessment and family sizes |
Baseline figures (duplicates, completeness, validity) feed the effort criterion. If you have not measured yet, start with how to measure master data quality before cleansing.
How do you build an ABC analysis for cleansing?
The inventory analysis that planning already uses can work, but it almost always needs adjustments.
Step 1: choose the value criterion
Usage looks back, requirements look forward. For cleansing, I prefer to add purchase history, an adjustment made outside the SAP ABC analysis (the page cited uses only usage value or requirements value), because non-stock items (services, direct-use materials, immediate consumption) do not show up in stock usage and still generate orders every month.
Step 2: set the period
The 90-day default proposed by SAP is short for businesses with seasonality or an annual maintenance shutdown. Use a period that covers the business cycle.
Step 3: consolidate plants
The same material can be class C in one plant and class A in another. In master data, the record is one, so the classification has to look at the whole group. The page Performing an ABC Analysis offers this option: analyze all plants, grouping the data for each material.
Step 4: set aside items with no movement
Before classifying, take items with no movement in the period out of the calculation. They go to a decision of their own, which is wave 0. The same page lets you include materials flagged for deletion and restrict the analysis to slow-moving items, which helps build that list.
How do you organize waves within a domain?
The question here is which records and which fields come first within materials, for example. If the question is which domain to tackle first, that is covered in single-domain or multidomain MDM.
| Wave | What goes in | Depth |
|---|---|---|
| Wave 0 | Items with no movement in the period | Block or deletion decision, no cleansing |
| Wave 1 | Active class A items and every critical or risky item, in any class | Full: description, attributes, unit, tax fields, duplicates |
| Wave 2 | Remaining active items, grouped by family | Critical fields for the family and duplicates |
| Wave 3 | Long tail of class C, with no criticality | Entry rules and cleansing when the item is requested again |
Wave 0: decide what leaves the database
It is not worth cleansing what is going to be blocked. Take the list of items with no movement to the owning areas and close the decision before standardizing anything.
Wave 1: value and criticality together
Active A items go in, together with critical or risky items, even if they are class C. This is where the cheap pump part gets priority.
Wave 2: the remaining active items, by family
Working by family keeps description patterns and attributes consistent, and forces a decision on which attributes each family must have. Effort sets the order of the families: for example, small, high-impact families first.
Wave 3: the long tail, by rule
Many items, little value, little use. Instead of cleansing them one by one, put rules at the point of entry and handle each item when it is requested again.
If the cleansing is part of a conversion to S/4HANA, the triage follows a different logic, covered in material master before the S/4HANA conversion.
Do all fields need the same level of cleansing?
No. Cleansing every field of every item at the same depth is the fastest way to blow the deadline.
For wave 1 items, the full package applies. For wave 2, focus on the fields that change decisions: description, unit of measure, family attributes and tax fields. For the long tail, the minimum that ensures the item can be found and is not created again. The criteria for finding duplicates are in duplicate materials: how to identify them.
When can a wave be closed?
Use a short checklist, agreed at the start with the project sponsor:
- The items in the wave reached the quality target set in the baseline.
- The owning areas validated the samples and signed off on the result.
- The duplicates found have a recorded decision.
- There is an entry rule so the same error does not come back in that group.
- The list for the next wave is already closed, using the matrix criteria.
What are the most common mistakes?
- Using the inventory analysis without including purchases of non-stock items.
- Cleansing items that should be blocked.
- Classifying by plant and giving the same material different priorities.
- Leaving a critical item in the class C queue.
- Applying the same depth to every field.
- Closing the last wave without entry rules, and watching the database get dirty again.
Hypothetical example: how the matrix changes the order
Hypothetical example, with fictitious numbers, only to illustrate the logic.
Picture a database of 30,000 materials. The company chooses 12 months of purchases and usage and sets its own value cutoffs. Of these items, 9,000 had no movement in the period and go to wave 0. That leaves 21,000 active items.
By value, 2,000 land in class A. Maintenance flags 600 critical spare parts outside class A, 450 of them in class C. The tax team flags another 300 items, also outside class A, with tax fields to review. Wave 1 ends up with about 2,900 items, assuming the lists do not overlap (the 2,000 class A items plus the critical and risky ones that were not in it). Wave 2 handles the remaining active items by family, and the long tail goes to entry rules.
Had the company followed only the A items, the 450 critical spare parts would be at the end of the queue.
Where should you start an ABC-based cleansing?
- Measure the baseline of the chosen domain.
- Build the value analysis with purchases and usage, consolidated plants and a period that covers the business cycle.
- Set aside items with no movement and take the block decision to the owning areas.
- Ask the areas for the list of critical and risky items.
- Build the matrix, define the waves and the depth of each one.
- Agree on the wave closing checklist before you start. The full cleansing method is on the data cleansing page.
Frequently asked questions
Is the ABC analysis for cleansing the same as the inventory ABC analysis?
Not necessarily. The inventory analysis looks at usage or requirements. For cleansing, it is worth including purchase value, because non-stock items generate orders and do not show up in stock usage.
Should an item with no movement be cleansed?
First decide whether it stays in the database. If it is blocked, it does not need cleansing. If it stays, it goes into the wave that matches its risk and criticality.
What period should the analysis use?
SAP proposes 90 days as a default, but the company can choose any period. My recommendation is a period that covers the business cycle, including seasonality and maintenance shutdowns.
Can you run an ABC analysis on suppliers and customers?
Yes: suppliers by spend and customers by revenue. But master data risk does not follow value. A low-spend supplier may lack qualification, have an expired document or have a registration status that does not match the Brazilian federal tax authority records.
What should you do with the class C long tail?
Put rules at the point of entry and cleanse each item when it is requested or reactivated. That way the effort follows actual use.
How many waves should a cleansing project have?
There is no right number. What moves you to the next wave is the closing criterion agreed at the start, not the calendar.
How 4MDG can help
Prioritizing requires knowing the size of the problem. The 4MDG assessment runs on a sample of the database, without installing anything, and measures duplicates, inconsistencies, empty critical fields, invalid records and hidden tax risks. From there, data cleansing can start with a single domain and end with ongoing governance (entry rules, approval workflow and indicators), so bad data does not come back. If you want to build your prioritization matrix with us, start with consulting and assessment or reach us through the contact page.
Sources
- SAP Help Portal, "ABC Analysis" (SAP S/4HANA 1610, Inventory Controlling), https://help.sap.com/doc/9313c453f57eb44ce10000000a174cb4/1610%20002/en-US/7310c453f57eb44ce10000000a174cb4.html, accessed October 15, 2026.
- SAP Help Portal, "Performing an ABC Analysis" (SAP S/4HANA 1610, Inventory Controlling), https://help.sap.com/doc/9313c453f57eb44ce10000000a174cb4/1610%20002/en-US/7610c453f57eb44ce10000000a174cb4.html, accessed October 15, 2026.