Do’s and Don’ts of Data Mesh

Kineret Kimhi
Published in
4 min readSep 6, 2022


Are you considering data-mesh for your organisation? Here are some tips on how to do it right — and what to avoid at all costs .

The data community is buzzing around the Data Mesh concept. If you are a data enthusiast, there is little chance you haven’t heard of it. The famous article which started it all was How to Move Beyond a Monolithic Data Lake to a Distributed Data Mesh written by ThoughtWork’s Zhamak Dehghani. Deghani suggests taking inspiration from the last two decades’ software engineering movement from monolithic systems to micro-services, bringing it to the data world. Data mesh is a mindset change, a shift in perspective from a one-team owning multiple deliverables to multiple data teams each owning a deliverable. Data Mesh, according to Deghani, is key to democratising data and fulfilling the dream of every organisation — being truly data-driven.

At BlaBlaCar, Data Mesh resonated with our challenges: navigating through growing pains of a startup transitioning to a mid-sized enterprise. Our two main challenges were scale and data quality: while growing in data talents, we couldn’t seem to improve velocity in our Data teams, which were organised per profession. We kept solving workload problems with hiring, but still the quality of our data was poor — we found issues too late, sometimes even after strategic decisions were made.

And so, we decided to go for it — adopt the data mesh in our organisation. This is BlaBlaCar’s before and after Data Mesh organisation:

Eighteen months later, we have learned first-hand the do’s and the don’ts of a data mesh. Here are some of them.


Data Governance comes first

A Data Governance framework is a set of rules and tools set in place to ensure clarity on data ownership, security and quality. Your Data Governance Framework should be ideally set in place before re-organising. Why? Imagine giving a single team full autonomy with zero standards nor processes. You'll find yourself cleaning up a huge and expensive mess, one year down the line. We want to ensure that the different domains communicate in a similar language in case of dependencies between domains, creation of new domains and movements of talent from one domain to another.

Communication Communication Communication!

Adopting data mesh is a nice term for a re-organisation, which means change and uncertainty. Why so much uncertainty? Those who’ve adopted data mesh are still pioneers, and there’s not enough knowledge on the long-term effects on data teams. This is why we want to involve our greatest minds on the team, early in the decision process, and refrain from top-down decisions. We recommend educating teams on the data mesh concept before it becomes a solid plan: have them read about it, find use cases, create an environment of challenging the idea. At BlaBlaCar, we held weekly updates on the plan, as it slowly unfolded. Periodically, we held open Q&A sessions as well.


Go by the book

Deghani’s fascinating article is the basis for a data mesh org, but every organisation is unique. Feel free to accommodate the data mesh to your specific company, and don’t stick to the instructions word-by-word. Why? First, the transition phase to a 100% Mesh could take you too long and be too costly. Second, You will find little practical advice in the literature, and could end up cooking a 4-layer cake without a recipe. At BlaBlaCar, we were willing to trade key aspects of Data Mesh to make it fit our organisation, such as keeping ingestion a central function for the sake of little duplication.

Start all at once

Data Mesh is a mindset shift more than a re-org. Mindset shifts need maturity, and that doesn’t happen overnight. Despite the Mesh Rush of data orgs around the tech world, we would not recommend fully transitioning the org. Rather, implement in small increments — and start with a Proof of Concept. At BlaBlaCar, we created one single data domain while keeping the rest constant. It allowed us to collect periodic feedback from the team along the way, and improve through these feedback loops.

In summary, push for a strong Data Governance framework, with a clearly communicated reorg plan, instead of rushing towards a reorg that can not accommodate your specificities.

Hopefully, our lessons from the trenches will help you kickstart your data mesh. Good luck!

Thanks to Emmanuel Martin-Chave for your help on writing this article!



Kineret Kimhi

Over a decade of experience in Big Data Analytics and Data Management