Data Migration in E-Commerce: Those Who Wait Will Lose

Interview with the PIM Experts from ATAMYA and netcare

Those who migrate product data today rarely do so for the sake of change. Most companies only commit themselves once the paint threshold is exceeded. When Excel no longer excels. When omnichannel is suffering from a breakdown caused by fragmented data structures. When the AI investment is meaningless because the data foundation is missing.

And it is precisely here where the work of ATAMYA and netcare begins. Dominik Schröter, Head of Software Development & Quality Assurance at netcare, has been accompanying migration projects for years from the implementation perspective. David Klein, Senior Consultant at ATAMYA, knows the ins and outs of data and process problems of his clients from hundreds of projects. In a conversation with Tiffany Wiener, the two of them explain what’s really going on – and how to do it better.

Why do Companies Migrate Today and What is the Most Common Catalyst?

Dominik Schröter: In most cases, companies come to us when they hit their threshold of pain. When the ambitions of department divisions fail given the limits of dated systems. In all this, omnichannel publication pressure is often times the immediate trigger: B2B shops, international marketplaces, country-specific channels – with fragmented data structures and manual Excel tables, this house of cards will be quick to crumble.

David Klein: On top of this, time-to-market is an issue. When a new product assortment takes months because data needs to be copied and maintained manually, a valuable sales margin is lost. And then there is also the AI moment: Many companies are currently in the find-out stage of learning that they cannot even mobilize state-of-the-art technologies – such as automated product descriptions or AI-assisted data enrichment. Not because of missing technology, mind you, but because the foundation of data has glaring holes. Data must be as granular as possible to work with it on the internet, find products, and fill channels. Anybody who lacks this is only sabotaging themselves.

 

What should a Company do First Before Launching the Migration?

Dominik Schröter: Our golden rule: a successful migration starts never with code or technical mapping tables but an honest assessment of the situation. We make the beginning with a data audit – what data is there, how does its quality fare, what portion of it is no longer needed. And then there is the most crucial question: Processes before data fields. We must first gain a concept of how data comes to be within the company and who shoulders the responsibility. A PIM project is 70 percent organizational development and 30 percent IT.

David Klein: I’m happy to expand this by the channel question: What do I want to do with such data? Am I required to fill configurators with it? Do I want to be found easier by search engines? Am I expected to prepare media assets in such a way that the Google image search and AI-assisted product finders can work with it? This speaks volumes to me about what I must do with the data – and what I ought to prepare prior to the migration. Additionally: How much will I communicate with external systems? Content management, ERP, in-house systems – all this needs to be factored in from the get-go.

 

Which Past Problems do Companies Most Commonly Let Carry Over?

Dominik Schröter: “Grew historically” is, to us, more often than not a sugarcoating euphemism that stands for years’ worth of chaotic expansion. The classic case: Duplicate data, cryptic free text instead of well-structured attributes, as well as shadow IT – the Excel table that is stored on the desktop of an individual employee that is so crucial to the entire business. We conduct in-depth sample analyses at a very early stage. When half of the products are missing mandatory attributes or measures are registered in millimeters once but in centimeters another time, we hold up the mirror to our client.

David Klein: We’ve already seen more than 17 different variants of the “length” attribute in the case of one global customer. It’s hard to imagine – but such cases are far from the exception. They are the standard. The decisive question is not: Must data be perfect before migrating? But: What should I import into the PIM – and what do I exclude? A product variant that somebody used to buy 20 years ago but was not requested by anybody else ever since should not be adopted. By that I don’t mean you should delete it, but to simply not import it.

 

Where do Migration Projects Fail Based on Your Experience?

Dominik Schröter: Rarely because of the software. We repeatedly recognize three patterns. First, there’s the “IT project” misconception: When the project is purely delegated to the IT department, requirements for business, marketing, and product management are dismissed. IT delivers the infrastructure – but it’s business that must define the objectives. Secondly, missing data governance: Editors who contribute after go-live without clear processes will make data quality deteriorate in virtually no time. Thirdly and lastly, the lift-and-shift misbelief: To think that one could simply take the outdated, unstructured data and simply push it 1:1 into the new system – with the expectation that this will make all problems go away.

David Klein: There is another pitfall that I encounter again and again: Missing decision-making authority. We’ve met with clients where decision-making groups decide upon how things shall be done – just to readjust everything from scratch as early as the next review round. That’s how a project drags on forever. There has to be somebody who smashes their fist on the table and proclaims: “And that’s how things shall be done stat!” And then there’s also the process question: One has to be flexible when it comes to adjusting processes. “Never change a running system” isn’t an argument – especially not if you can save up to half of the invested worktime.

 

What do Today’s Companies Expect from a Modern PIM?

David Klein: The PIM is supposed to automate. Massively. Not just individual steps – but processes. Rules that apply without people monitoring them. New data comes in, so that a process is triggered automatically. I want to distribute more products with better quality to much more channels – simple and easy, without the need to manually intervene at every step. It should equip me with the tools to configure new channels myself and control them accordingly, without requesting a developer’s aid at every turn. And I want to communicate with AI – be it in-house or outsourced – so that I can share information and produce results that are here to stay: for text generation, quality assurance, measurement consistency checks, and translations.

Dominik Schröter: On top of this, users have long grown tired of static databases. They want graphical interfaces that can be navigated intuitively like modern SaaS tools. And the data model should be capable of integrating a new product category as early as by tomorrow – without months’ worth of develop projects.

 

How should a Modern PIM be Structured to Integrate Complex Data Structures out of ERP, PLM, or MDM?

Dominik Schröter: API-First and cloud-native are not buzzwords to us but the conditio sine qua non. Data must flow in real time – the days of batch jobs running over night are long gone. The PIM must be connected seamlessly to ERP, PLM, and DAM. Our role distribution goes like this: The ERP stays the leading system for transactional data such as prices and stock of inventory; the PIM is the undisputed single source of truth for all marketing and sales information.

David Klein: Five years ago, our company seized the chance to rethink PIM from scratch – with 30 years’ worth of proven experience in our bag and without technological compromises related to dated systems. The path was clear: MACH – microservices, API-first, cloud-native, headless. More concretely, this means: Be it ERP, PLM, or MDM, every system can send and receive data via standardized interfaces. When a microservice has a workload of 80 percent, it scales automatically. I can update individual components on this side without the fear of breaking something on the other side. Those who trust in modern technologies today don’t only build the present – but forge the very foundation upon which new developments such as AI or automation can be seamlessly integrated.

 

How does a PIM Guarantee that Data Quality is Maintained over the Long Term?

Dominik Schröter: Data quality is not a one-time project event but a continuous process. Garbage in, garbage out – this remains true now as then. We implement validation rules: A product can leave the PIM only once the data quality score has been completely met. The text is present, the image is linked, measurements are maintained. On top of that, add automated release processes and central management instead of data islands – when all teams work in and through one and the same system, people will be quick to identify and correct data waste.

David Klein: AI helps enormously here. Spellchecking, measurement checks, consistency analysis – I don’t have to monitor all this purely manually anymore. And once I’ve set up a process cleanly, I don’t have to spend anymore thoughts on whether or not the price on the datasheet is correct. I synchronize it using the API before generating the datasheet. With that, it cannot be but correct. What matters here is the human-in-the-loop.

 

What are the Benefits of Cloud Operation – and Why do Many Companies still Hesitate?

David Klein: The technological advantages are clear: automatic scaling, zero-downtime updates, simple integration of other systems. I can update microservices individually without touching the whole system. And I’m future-proof: small building blocks, automatic tests, no monolithic dependencies. Worries concerning cloud are usually related to data protection. But ATAMYA is designed so that we can simply switch cloud providers in case of issues or policy adjustments. Our datacenter is located in Frankfurt – with clear contractual regulations. While such worries are understandable, they can be resolved given modern contract structures and DSGVO-conform setups.

Dominik Schröter: We demonstrate the long-term use to clients: No more large-scale upgrade projects every two years, the system is always up-to-date, and internal maintenance expenses sink drastically. And concerning the worry that individual adjustments that have grown historically will not be mappable in a SaaS cloud – we address those via consultation and evidence. Modern APIs solve such adjustments significantly easier than monolithic custom code.

 

What Misconception about Data Migrations do You Encounter Most Frequently?

David Klein: That it will take forever. This is the greatest misconception – it only takes forever if you spent too much time talking about it instead of simply making a beginning. Over-specification is the actual problem. In the case of ATAMYA, I can export a data model, make adjustments, and import it. I can do bulk-editing. I can clean up data quickly using validated Excel exports. Errors can be corrected so fast that you may not even spend so much as an afterthought on it. We get started – and make change happen.

Dominik Schröter: Three myths we routinely disclose in our opening discussion are: “The PIM cleans up our data on its own.” – No, the system provides the tools, but the conceptual work needs to be carried out by us, together, in collaboration. “We migrate once our data is perfect.” – There will never be the perfect point in time, delays will only increase the technological debt. “It’s a pure IT project.” – A fatal approach, given that we are accompanying a transformative project that concerns the way entire departments work.

 

How Can Companies Verify After Go-Live that They Investment has Paid Off?

David Klein: Success is visible not only after go-live – you’ll notice it much earlier. Our approach: Proof of concept. We meet up with the project members, define the two biggest obstacles together, and demonstrate that we can prove them. Once this is accomplished, the rest is a matter of configuration and routine work. As part of this, the people involved get to know each other, familiarize themselves with our software, and learn how to work together. We have realized projects where members were promoted because they’ve done such great work during the PIM project. This is not a side effect – but a natural strategic result.

Dominik Schröter: Marketing can roll out campaigns within days rather than weeks. Customer support reports that they receive way less inquiries related to product properties. And the Excel ping pong between teams is a matter of history. My advice to all procrastinators: Those who don’t invest into their product database today will miss out on the transition to automation, marketplace connectivity, and AI applications in the next years. Migration isn’t a bothersome IT issue. It’s the strategic foundation for tomorrow’s competitiveness. Start small. But start now.

 

Conclusion: Those Who Wait for Perfect Data Will Wait in Vain

Most companies only begin with their migration once they have reached the pain threshold. Successful projects, however, start much earlier: With clear objectives, a realistic self-assessment, and the flexibility to question running processes.

This is the point that ATAMYA and netcare agree upon before anything else: A PIM migration is not just another technical project. It establishes the very foundation for better product data, faster processes, and future-proof product communication.

Those who centralize product data gain a better overview. With a better overview, processes can be automated. With automated processes, the time-to-market improves – and opportunities for omnichannel sales and AI naturally emerge.

 

You want to continue the conversation?

David Klein and Dominik Schröter are live on site at the K5 Conference in Berlin from 23rd to 24th of June – together at the ATAMYA stand in Halle 1, Stand 39. Anybody who wants to go more in-depth on questions revolving around the interview in person will find just the right contact person here.

Set a Date for the K5 Conference 2026