What are the real costs of bad master data: returns, errors, downtime

Why does bad master data cost more than bad processes? 

When a process is broken, you notice fast. A step is missing, a handoff fails, someone complains. You pinpoint the mistake, fix the flow, and the problem is gone. Data errors are trickier. They stay invisible until they show up as a symptom. And at first glance, that symptom points to the process, not the data.

A classic example: a customer places an order, the shipment goes out, comes right back, and no one knows why. So the investigation starts. The warehouse booked it correctly, shipping did everything right, and yet the wrong item ended up at the customer's door. The usual fixes: new checklists, more spot checks, an extra pair of eyes on every order. The real cause stays untouched: the data behind it. Because every automation rule, every shipping logic, every stock booking in the ERP runs on the master data you have on file. In this case, automation doesn't scale your efficiency. It scales your errors.

Bad master data costs you time and trust 

If data is wrong, incomplete, or inconsistent, the system will reliably produce wrong results. An item with no correct weight in the ERP gets the wrong shipping cost calculated. Without dimensions, the shop can't display a proper size chart. If an EAN is assigned twice, the warehouse books stock against the wrong item. None of these mistakes sit in the process. They all sit in the product master data. And they cost you real time.

There's more. In e-commerce, trust is built through reliable experiences. Your customers can't see these data errors, but they see the consequences. Someone who gets the wrong item, has to ask twice where their order is, or has to send back something they shouldn't have needed to return, buys somewhere else next time.

What are the effects of faulty master data? 

Type 1: Rising return rates 

Returns volume in German e-commerce is structurally high. According to a 2025 study by the EHI Retail Institute, nearly 90% of merchants report return rates of up to 50%. More than one in ten merchants surveyed report rates above 50%. In electronics, the average return rate runs 15 to 20%. Some of that is unavoidable. But the avoidable share often traces back to faulty or incomplete master data.

These are the most common return reasons that trace directly back to product data quality:

When product descriptions and attributes aren't fully maintained, marketplaces like Amazon or Kaufland can't display items correctly either. Customers order based on incomplete information and, frustrated, send back what didn't match their expectations.

Every avoidable return costs money. According to the EHI Retail Institute the average cost per returned item in e-commerce runs between €5 and €10. For high-volume or expensive items, that jumps to between €10 and €20.

Type 2: Wrong shipments 

A wrong shipment costs more than a return. You cover the cost of the return, the cost of shipping the correct item, and, if the customer doesn't wait around, the cost of the lost order. Add reputational damage on top of that, especially on marketplaces where reviews directly drive visibility.

The most common causes of wrong shipments don't sit in the warehouse. They sit in the ERP master data:

• Faulty or duplicate EAN/SKU
  ⁠When two items share the same EAN, or an EAN is stored incorrectly, the warehouse system pulls the wrong record when it's scanned. The picker grabs the right item, but the system books the wrong one.
• Missing variant assignment
  ⁠For items with multiple variants (size, color, version), assigning each variant a unique SKU is critical. If that assignment is missing or inconsistent, variants get mixed up. A customer orders size M in blue and gets L in black.
• Duplicate item entries
  ⁠When the same item is created more than once in the system (say, once manually and once via import), you end up with parallel stock bookings. Both records show inventory, but only one is physically in the warehouse. That leads to overselling and wrong shipments.

The chain reaction is the real problem: A wrong shipment creates a return, the return creates a stock booking against the wrong item, the wrong booking creates the next error. If you don't fix the root cause in the master data, you're just treating a symptom.

Type 3: Stalled processes in the ERP 

This is the type of damage that's least visible, but internally it eats up the most time. An order gets stuck in a status. No one knows why until someone checks manually. A missing required field turns out to be the culprit, someone fills it in by hand, and the order moves forward. Ten minutes of work, one time. But multiply that by 30 orders a day, five days a week, and you have a full-time role that does nothing but fix data errors.

How stalled processes come from errors in ERP master data:

• Missing required fields block automation rules
  ⁠When an automation rule in the ERP checks whether an item has a tax class assigned and that field is empty, the process halts. The order sits and waits for manual release.
• Inconsistent units create calculation errors
  ⁠When an item is stored once in units and once in cases (of 12 units), without the conversion unit defined, the ERP calculates incorrectly. Order quantities, stock levels, and supplier orders don't line up.
• Missing supplier assignments prevent automatic reordering
  ⁠If no supplier is attached to an item, the ERP can't create an automatic purchase suggestion. Purchasing has to check manually which supplier handles that item, every time the minimum stock level is hit.

The result: Your team spends time closing data gaps that should never have existed. And the more you automate, the more expensive those gaps get, because every automation rule that hits faulty data either stops or keeps running with the wrong information.

Which master data fields are the most critical sources of error? 

Not every field carries the same weight. The table below shows the fields that cause the most damage in practice, prioritized by damage type:

How does a clean ERP data model prevent the cost of bad master data? 

Data quality in your e-commerce ERP is not a one-time cleanup project. Going through every item once, filling in missing fields, and deleting duplicates solves the problem for today, but not for good. New items keep coming in, you expand variants, suppliers change their EANs. Without a structural fix in the ERP, the detective work starts all over again.

A clean ERP data model prevents errors at three levels:

1. Required fields and validation rules 

In Xentral ERP, you can define required fields for product management that have to be filled in before a new item can be saved. EAN, tax class, unit. No item goes into the system without those fields. That closes gaps at the source, so they don't surface later when the damage has already happened.

2. Variant configuration with a unique SKU per variant 

Every variant of an item, every size, every color, every version, gets its own SKU and is uniquely identifiable in the system. Xentral models variants so that warehouse, shipping, and marketplace connections always pull the right record. No mixing, no double bookings.

3. A single data structure across all channels 

If you sell across multiple channels (your own shop, Amazon, Kaufland, wholesale), every channel has to work off the same data. Xentral Connect syncs product data centrally: one change in the ERP updates every connected channel. No manual maintenance in every single system.

Clean master data as a prerequisite for AI 

The automation in Xentral's AI-native ERP can save you serious time and money. But there's something you can't overlook: AI agents only work as well as the data they run on. An agent that reads incoming order emails and creates orders automatically needs a clear SKU mapping. An agent that processes returns and corrects stock needs a clean variant assignment. An agent that reviews and approves incoming invoices needs the correct supplier attached to each item.

If those fields are missing or inconsistent, the agent can't prepare a meaningful suggestion and hands the task back to your team. That's a safety mechanism, but it means manual work for you, exactly the work you were hoping AI would take off your plate. In short: the data foundation has to be right.

In Xentral, the AI agents in the Agent Hub are built to take on operational tasks and prepare decisions for people: returns processing, delivery inquiries, invoice processing, payment matching. But they can only do that reliably if the master data behind it is right. If you don't clean up your data, you can't use AI in any meaningful way. This isn't a technical argument. It's a business one.

Cleaning up ERP master data: where should you start? 

If you're getting started on cleaning up your master data in the ERP, we recommend this order:

Step 1: Check for EAN duplicates 

Export every item with an EAN from product management (in Xentral, that's a CSV export from the item overview). Open the file in Excel or Google Sheets and use "highlight duplicates" (conditional formatting → duplicate values). Any EAN that appears more than once in the ERP is a candidate for a wrong shipment. Pay special attention to items imported from different sources, like an old shop system or a supplier file. That's where duplicates show up most often, because the same EAN gets created under two different internal item numbers. Result: the warehouse books item A, shipping pulls item B.

Step 2: Check required fields for completeness 

In the item overview, filter specifically for empty fields: tax class, base unit, and weight are the three fields that are most often missing and that block automation rules most directly. Without a tax class, no correct invoice can be generated, so the order sits in status. Without weight, no shipping label can be calculated automatically. In Xentral, you can filter the item overview by populated and empty fields and export a working list for cleanup. Start with high-volume items, because missing required fields on those cause the most damage.

Step 3: Validate variant assignments 

Export every variant item and check three things for each variant. First: does each variant have its own unique SKU? Second: is each variant assigned to a specific storage location, or does everything book against the main item? Third: is the marketplace mapping correct, meaning on Amazon the right ASIN is tied to the right variant? A typical mistake: size M and size L share the same SKU because only the main item was imported. The warehouse books correctly against the main item, but shipping pulls the wrong variant. On marketplaces, that leads straight to negative reviews and a higher return rate.

Step 4: Fill in supplier assignments 

Filter for every item without a supplier attached. Those are the items your ERP can't generate an automatic purchase suggestion for, even when they hit the minimum stock level. For each of them, add at least the supplier, the supplier's item number, and the lead time in business days. If you have multiple suppliers for the same item, define a primary supplier and set up alternates as backup. That's what makes it possible for a Xentral AI agent to prepare or trigger purchase suggestions automatically, without someone having to look up who supplies the item.

Step 5: Configure required fields in the ERP 

Go into the item configuration and define which fields have to be filled in before a new item can be saved. In Xentral, you set this up through the field configuration in product management. Recommended required fields: EAN, tax class, base unit, weight, supplier, and storage location. From that point on, anyone who doesn't fill those in can't save the item. That stops new items from entering your inventory with exactly the gaps you just cleaned up. This is the one step that turns a one-time cleanup into a permanent standard.

Final words: when you know your master data, you can trust your processes 

When something goes wrong in e-commerce, bad product data is rarely the first thing you think of. But when returns climb or orders get stuck, it's often the real cause. And unlike process errors, data errors can be fixed structurally. What you need is an ERP data model that ensures quality from the start.

Master data maintenance in your ERP is the foundation for automation. Clean master data is what lets AI agents take on real operational work and prepare decisions, without someone having to step in manually.