Skip to content

5 Areas for Analytics Leaders to Obsess About – Fast & Flexible Architecture

You Are Not A Unicorn, You Exist in Reality

Look across The 5 Differentiators of companies that use analytics for competitive advantage and you’ll see one of the great challenges for analytics leaders – the sheer diversity of areas of knowledge that analytics leaders need to understand well in order to lead advancements in analytical maturity. Analytics leaders need to (naturally) lead stakeholders across the organization, move from project to product, build systems of innovation and data mastery; and in addition, they need to stay up to date on fast moving tech. It’s the rare individual who possess expertise in all areas. Furthermore, the analytics leaders who might currently possess expertise in all areas will likely lose the expertise in tech fairly quickly if they aren’t dedicated to that area. Tech is simply moving too fast, in too many directions.

In most large organizations where data itself is not the product, the technical infrastructure is too big and too complex (and likely too dated) to be fast or flexible. Analytics leader should meet this challenge by enabling the change, which should be led by technical leaders.

Below are 4 areas to focus on the play the leadership role of enablement.

1. Get a friend and be a better friend

When I was at IKEA, I was so fortunate to have brilliant colleagues with hard core tech chops that I could rely on as we made decisions to unwind a massive technical debt. I got to know one so well that he could convey without speaking when he thought we were leaning towards a bad decision. If he was tilting his head to the side like a dog that heard a high-pitched noise, it was time to take a pause and reconsider. Find someone you can trust and know them that well. In fact, that friend 1 provided critical input into this blog. So did a few other friends.

Good friendship is symbiotic and as an analytics leader there several things you can do to empower your tech friends – to your own benefit. Focus on the needs, so they can focus on the solutions. As an analytics leader you should know what your team needs to deliver to your strategies, for example the corporate strategy, your data strategy* (client access only), etc. Focus on clarifying those needs, including making statements about future needs based on future business needs that are, in turn, driving future analytics needs. Be clear where/how much uncertainty exists.

Invest your time, influence and money in enabling standardization. Specifically:

Invest in a co-created tech & data direction, focused on enabling the org with what tech stack and which data standards you will start from and build onto (Friend 1).

Invest in a data catalog, potentially a separate API catalog etc (Friend 1).

Incentivize teams to publish data through the standard tech components and patterns1

As a benchmark of how hard you should push standardization consider how hard Bezos pressed the “API Mandate” back in 2002.

2. Embrace modular thinking

Advances in technology and technique require new ways of thinking, dispensing with a “one-size fits all” approach. This result only ever really resulted in mediocrity for most, pain for a few. Conceptually the big shift is toward thinking about flexibility in architecture – a shift from monolithic systems to modular systems. And while analytics leaders have long known that monolithic systems are well past their prime, the move to modular has not been strong enough.

Modular systems enable firms to leverage the binary nature of modern business data (pun intended) – structured and unstructured, batch and real-time, proprietary and public. These features of this data (to name a few) are too different for the data to be treated in the same way. This has always been the case and we never liked it, we just accepted it. The difference is that modular architecture can differentiate, monolithic architecture could only ever sub-optimize.

Modular systems enable data teams to work in fit for purpose ways with tools that are suited to the need. For example; modular systems “enable people to have possibilities to deploy neural networks and random forests into GPU scaled execution engines and simple decision trees into CPU [and] you get tremendous effects there.” (Friend 2) And best in breed tools like Snowflake are built to be maximized in these environments.

Scalability is another key benefit of modular architecture. As friend number 2 notes it’s about “scalability on multiple layers, data being one but also in the compute power on the analysis execution.” (Friend 2)

3. Challenge legacy ways of working as much as legacy tech

Data practices are changing at a slower rate than tech. Tools, like cloud, micro services, and graph databases to a name a few are under leveraged because organizations won’t change how they work.

Two examples of new approaches that have less traction at larger legacy bound enterprises are ELT (as opposed to ETL) and domain-based architecture. Cloud technologies, including embedded services are able to support the volume and velocity of data without transformation (T) happening before the loading (L). Yet, almost out of habit, few embrace the new approach. Domain-based architecture challenges the notion that a set of data must have rigid attributes regardless of use and instead uses data virtualization to make curated data sets specific to intended use without excess risk of redundant or contradictory data elements, or challenging the sacred cow of a “single source of truth”.

The biggest challenge in overcoming these types of obstacles is inertia and analytics leaders can lend a hand to push data practices in more progressive direction.

4. Explore organizational changes

In some sense, in analytics we are behind the software field in how much energy we put into conceptually thinking about data (how it flows, the transient nature of its features, etc.). So, as the analytics filed matures, the future technical changes are likely to be inspired by software, and software inspired technical changes will drive organizational changes.

A thought leader and member of IIA’s expert network, Casey Rosenthal has written and spoken about software approaches to analytics products* (client access only), including organizational approaches. Specifically encouraging the concept of the “highly aligned, loosely coupled”, writing “If you can structure your organization in such a way that there are small, loosely coupled teams highly aligned on the goal they’re trying to achieve, then they can operate with interdependence, move quicker and optimize for feature velocity.”

Other organizational concepts like MLOps (built off the software practice of DevOps) should be explored more quickly and with more bravery by analytics leaders, together with tech leaders. ML/AI are more than a linear progression of advanced analytics. Given the heavy reliance on engineering and code for reliable ML models, even firms with successful advanced analytics will need to do more than add new talent, they will have to change how they deploy that talent. As a bonus, these types of practices will enable analytics teams to better leverage the benefits of the modular architecture mentioned above.

Finally, this stuff is difficult and changes quickly so it’s only natural that the ecosystem of third -party specialists has grown substantially and captured Wall Street’s attention. See Matt Turck’s 2020 overview for a raft of evidence. And while “buy or build” has been a big part of a company’s tech decisions, more and more it’s becoming a part of an analytics leader’s sphere of tech influence; impacting organizational development. Developing a strong partnership with tech and the ability to make the buy, build or partner decision together is increasingly essential for analytics leaders.

Speed + Flexibility + Responsiveness = Value Capture

Fast and flexible architecture are hallmarks of companies who are leading with analytics. And those companies are, more often than not, the leaders of their industries. These attributes together with a responsive and (dare I say) agile mindset to work in new ways are how analytics and technology come together to achieve results. In non-digital native companies, it takes a partnership of analytics and technical leaders to make this so; working together to systematically reduce drag and increase agility. Only then can businesses meet the increased speed and unpredictability of business today. After all “If the rate of change on the outside exceeds the rate of change on the inside, the end is near.”

Friend 1 - Thomas Tallinger

Friend 2 – Erik Dingvall

*To speak to our team about gaining access to our research library, send a request through the form below.