You have probably heard of “data mesh” – the data-architecture idea everyone has been quoting since 2019. Here is the part nobody mentions: boTec has been building exactly this since 1998, two decades before it had a name.
In Brief
- Data mesh was introduced in 2019 and subsequently formalised around four principles: domain ownership, data as a product, a self-serve data platform and federated computational governance.
- boTec has practised those same principles in production since 1998 – two decades before the term existed.
- The DataBridge is not an implementation of data mesh; it is the proven, industrial original.
- It differs from data lakes, generic middleware and textbook data mesh alike - links to live sources, never copies.
Part of the series , Data mesh & the boTec DataBridge
- Part 1 - Data mesh, since 1998 - 20 years before it had a name – you are here
- Part 2 - Link, don’t copy – why referential integration beats copying
- Part 3 - Harmonise, don’t standardise – Business Integration Models
- Part 4 - From data maturity to AI – getting the sequence right
The term was coined by Zhamak Dehghani at ThoughtWorks in 2019. It said something simple and correct: stop dragging all your data into one central pile. Instead, let the domains that know the data own it, serve it as a dependable product, give them a shared platform to do so, and govern it together rather than through one central bottleneck. Sound familiar? It should – it reflects the principles that the boTec DataBridge has applied in refineries and factories since the late 1990s.
The idea arrived in 2019. The practice arrived in 1998.
In 1998, two refineries merged into a single Bavarian group spread across three sites, each with its own tank, process and laboratory systems, its own terminology, and its own overlapping equipment names. boTec integrated that tangle into one consistent view – and did it exactly the way data mesh would later prescribe: the data was integrated, not duplicated. It stayed in its original source systems, related through a shared object model, with one consistent calculation method so the numbers finally reconciled. The project was documented publicly in Computerwoche in 1999, twenty years before anyone wrote the words “data mesh”.
We didn’t adopt data mesh. We’ve run its principles in production since 1998.
Principle by principle
What data mesh named - and how boTec already did it
The four principles the industry labelled in 2019 map one-to-one onto how the boTec DataBridge has always worked. Nothing here was retrofitted to fit a trend – the trend caught up to the architecture.
- Domain ownership → owners keep their systems. Product structures stay in PLM, inventory in ERP, process values in the historian. Each process owner keeps their own system and model; nothing is migrated or taken away.
- Data as a product → a single source of truth. A shared translation layer lines up units and meaning so every department reads the same number - one definition of the apple, not four.
- Self-serve platform → a platform that already exists. Data mesh calls for a self-serve platform; the boTec DataBridge provides one – with 100+ connectors and the ability to connect systems incrementally.
- Federated governance → Business Integration Models. Units of measure, time and mapping logic are harmonised once, centrally - while each system keeps its own wording.
How we’re different
Everyone claims the principles now. Few actually deliver them.
If these ideas are suddenly popular, why not just buy the nearest tool that claims them? Because most approaches borrow the label but not the substance. Here is where the boTec DataBridge is genuinely different.
- Not a data lake or warehouse. The central-pile model copies everything into one store where it immediately begins to age, the very thing both data mesh and boTec reject.
- Not generalist middleware. Point-to-point and iPaaS tools still copy data and multiply interfaces (n×m). boTec is referential and system-agnostic: each system connects once and remains the source of truth.
- Not data mesh as a theory. The textbook version asks you to reorganise into “data-product” teams and build a platform that, by its own author’s admission, doesn’t fully exist yet. boTec is a proven platform – 25+ years, 100+ connectors, built for process and production industries.
Others sell you the label. boTec has shipped the substance for more than 25 years.
→ Curious where your own landscape stands? Book a free 30-minute data check
Why “data as a product” matters - one apple, four systems
| ERP (material master) | APPLE GALA 70/75 BIO CAT1 |
| PLM / spec | Gala, cal. 70-75, organic, class I |
| Warehouse (WMS) | Pomme Gala BIO 13 kg colis |
| Quality lab (LIMS) | calibre 70-75 mm , 180 g avg |
| Variety / Calibre / Category | no shared field |
Four true descriptions of one apple, and no system can join them. Whether you call the fix “data as a product” or simply a shared definition, until it exists you cannot compute margin per variety if “variety” isn’t a field anywhere.
The payoff
Same data for everyone, resilience, and real AI readiness
- Everyone works from the same data – one definition, and reports that agree across departments.
- Resilience against system changes – replace or upgrade any single system without disrupting the business on top.
- Genuine AI readiness – clean, harmonised data is the readiness AI actually needs; most “AI-readiness” gaps are data-maturity gaps.