Skip to content

Search the site

This Allianz CDO built a data architecture & culture "from scratch". Here's his advice

Kafka, Azure, Snowflake underpin the insurer's data architecture...

Sudaman Mohanchandralal, Chief Data Officer (CDO) at Allianz Benelux, has spent 20+ years refining his approach to data and analytics, via a career that has spanned data roles at BNP Paribas, ING and now Allianz: a financial services behemoth with over 92 million customers, and 142,000 employees.

The CDO -- who has worked his way to the top from database administrator (DBA) -- was hired by Allianz Benelux's CEO Anthony Bradshaw in 2017 to help drive a major shake-up of the insurer's data strategy. Bradshaw, who'd newly taken on the CEO role, was intent on using data to help refine premiums in a hyper-competitive market, and Mohanchandralal jumped at the challenge. As Sudaman puts it on a call with The Stack: "My task was to make pricing as competitive as possible from a industry perspective, but as reasonable as possible from a customer perspective.”

He almost makes it sound easy. But few of the ingredients were in place -- other than raw data in various formats -- to make this happen and the journey has involved some heavy lifting. Yet three years later, he has built a finely tuned approach that makes use of a colossal 20 TB of structured data and nearly 2PB of unstructured data to inform product pricing and range of other enterprise decisions -- using a stack built around an Azure-powered data lake (fed by a Kafka engine), and data warehouse Snowflake.

Allianz Beneluz CDO: “I had to literally ignore what was there"

An early problem when he arrived: there was no real data architecture, strategy, nor culture in place. Data was fragmented and used in an ad hoc way.  Speaking to The Stack's founder Ed Targett – who suggests that it sounds like he had to build both infrastructure and culture “almost from scratch” – Sudaman is quick to respond: “It was not ‘almost from scratch’; it was from scratch! I had to literally reinvent the approach."

Lest that sound too brusque, Sudaman is keen to emphasise that if experience has taught him anything, it is that lines of business and the end-users are central to putting data to work effectively.  As he puts it: “Believe me, it's not enough that I am there. I need support from those who are the initial change agents, and to understand how product users work, what they rely on. When I came in the data was distributed, fragmented; there was no data governance in place. Obviously I had to put all of this in place.

"Yes I was hired to do data science, but for that first and foremost I need data – and I need an assembly line to put data in place, prepare data, clean data, publish data: then I can build our models on top of it and build a feedback loop also inside. That takes real engagement."

Building a data valorisation process

Building his approach started with understanding what exactly was being used to inform decisions at the coal face by the insurer's teams.

The Allianz Benelux CDO tells The Stack: “I call data management ‘data value management', because I only want to assetise data which has value. Valorisation was a principle that I had to explain to business.

"I had to say ‘not all data sets have value, but the data sets which you use for making decisions have value, so you need to explain to me what datasets you use for what decision making processes.’

Culture is king

As Sudaman has earlier noted, in building and sustaining a data culture at Allianz Benelux, his team drew on a framework developed by Charles Duhigg; a Pulitzer Prize-winning expert on habit and productivity.

Duhigg's work takes into account evidence from MIT researchers who have described a three-part neurological loop at the core of every habit: a cue, a routine and a reward; an approach that is key to what Sudaman has tried to implement at Allianz, including via dedicated data training.

As he tells us: "A data education was required. We put together an academy called the ‘Accelerate Data Academy’, which is now one of the most subscribed academies at the group level. Making it as pragmatic as possible is the key element. We teach data analytics techniques at different levels; we educate at technical, business, and executive levels.

" We do this at the bronze, silver and gold level and have nearly 80% of the group certified at the bronze level. You need to inspire habits too, so to inspire this we have individual insights hubs for sales and distribution, for marketing. We want to help users touch not data, but insights; use insights from the data to discuss projects and their progress.

"I believe soon that the dependency on us as a data team will soon go away and they will start doing themselves. My work will be done."

Whereas traditional ways of statistical modelling started from building hypotheses and confirming these with data samples, today's mission starts with a business case: building data sets that capture as many dimensions of the business challenge as possible, then generating hypotheses from this dataset. As the CDO explains: "Generating value from our data is dependent on an entire chain of events to which numerous stakeholders contribute.

"This becomes even more the case as data are scattered across the organisation. All these people need to change—to a certain or substantial degree—their behaviours. This requires a cultural change.

Proof of value first please...

So what are the key lessons he'd impart to those embarking on a bid to drive data-powered organisations, or those looking to use data to refine their existing business decisions? "The key thing I would like to tell your readers", he emphasies, as we close the call, "is this: please don't kickstart data culture initiatives or heavy data investments in terms of infrastructure until you have a serious business case -- and do it in a short loop, maximizing value.

"I have built the data office like a startup. It has a lean startup culture.

"We go to the board every year to ask for funding by showing what we have brought in, as revenue, for the company.

"So there is an attribution logic to the Data Office activities. I can proudly say ‘this is the money which our initiatives have brought in; this is the operational process. This is the growth that we have been able to bring’".