My Journey Through Data Analytics
A Little About Myself
For the last several years, I have been writing articles on the practice of data analytics (the term encompasses data science and business intelligence as well), often with a contrarian voice. I’ve worked in the data analytics domain for 3 decades now, and for the last 10 years, I’ve been co-running a boutique data analytics consulting practice while also dabbling as an adjunct professor for data science and digital transformation courses for working adults with local universities. Before stepping out on my own, I was the Chief Analytics Officer for Citibank Asia Pacific and was a recognised thought leader in data analytics within the regional banking scene.
Having had a front row seat in the evolution of data analytics in the business and banking domain, I’ve decided to put together a series of articles to dispel and debunk a lot of bad thinking and bad practices in this space. You don’t have to agree with me, but I’m hoping my opinions will nonetheless enrich your perspective.
The Day I Left Corporate Life
Let’s first address the elephant in the room. My clients and students often ask me why I left such a high profile job (at Citi) for the risk of entrepreneurial unknown. The reason I left Citi came down to a disagreement on how to best organise the data analytics function across Asia Pacific. I was overseeing a data analytics function that was embedded within the 14 markets across Asia Pacific, with a very thin layer of resources at the regional level. A classic decentralised organisation structure. But head office in New York was pushing for a centralised structure; it wanted to expand the scope of its India-based data analytics ‘Centre of Excellence’, which was set up to support the US business. They wanted to grow the data analytics resources in the India CoE while disbanding the decentralised country resources. I was not in favour of it based on my then-intuition that quality work and value creation in data analytics can only be achieved by having the data analysts embedded within the businesses they support. I pushed back for a couple of years, but was eventually unable to hold back the political tide. I decided to resign my position as regional CAO rather than see the dissolution of the function that I had contributed to and helped built over the years into an industry-leading player. (Footnote: Citi’s data analytics capabilities within the region has since collapsed due to centralisation as I had predicted.)
I realise I was an early advocate for a decentralised and federated data analytics function. It was intuitive based on my observations at the time. After Citi, I went on to develop an underlying theory on why centralisation and offshoring was value-destructive for the commercial practice of data analytics and earned a PhD for it. At the same time, I co-founded Red & White Consulting Partners with my then Indonesia head of data analytics (Sally Taher), who had also decided to leave soon after me. Red & White is now 10 years in operation, and have consulted for major banking and non-banking clients in the region.
Where We Go From Here
Over the course of the next several articles, I’m going to provide my first-person practitioner’s viewpoints on a number of salient topics in the data analytics domain. Through my consulting and lecturing work, I’ve been nicely surprised that I’ve developed new perspectives and sensitivities to data analytics that were not apparent to me during my time at Citi. For example, I’ve noticed that …
- Many data scientists are poor problem-solvers; they have poor data sensemaking abilities.
- Data Science is not a substitute for Information Management and Decision Science, the latter being more commercially important and valuable.
- Visual Analytics is a complex cognitive science, and more than just making aesthetically pleasing charts.
But since I brought up the topic of data analytics centralisation in my ‘journey to discovery’, allow me to close this introductory article by perhaps giving some highlights on this topic.
Value Destruction of Centralisation
During the early days of my consulting practice, I often get asked by my clients on how they should be organising their data analytics resources — centralised vs decentralised. Of course the question they neglect to ask is also: “What constitutes my data analytics resource?”
The idea of resource centralisation is a throwback to the industrial age approach to solve for scale and efficiency. It has been applied to the practice of data analytics with often disastrous results. It was popularised by the likes of Professor Thomas Davenport et al. (‘Competing on Analytics’, 2007), academics who lack practice and field experience. I used to read their books and articles as a budding data analyst, and was convinced by their naive thinking on setting up so-called ’Centres of Excellence’. Centralisation ignores how data-driven problem-solving actually works: there is a large chunk of activity at the front-end in terms of data sensemaking and problem framing, activities that are inherently equivocal and, at times, ambiguous. Overcoming this natural front-end challenge (also known as translation friction) requires the data analytics resources to be embedded in the line of business where the need for the problem-solving originates. Throw into that mix the popular inclination for offshoring (to countries with seemingly more and affordable data analytics talent), and we have the perfect recipe for mis-communication and creating irrelevant solutions. I have studied this phenomenon where even business report creation and dashboarding become ‘corrupted’ through centralisation and offshoring.
There is also the chunk of back-end activities involved with solution implementation (also part of translation friction). Data-driven solutions are designed to impact the decision-making ecosystem of an organisation, and centralised (and particularly offshore) data analytics resources have little comprehension to navigate this. Understanding how decision-making is being augmented (both positively and negatively) by the data-driven solution is critical to its success, and this is where many organisations fall down, and one of the key reasons that many data science ‘projects’ simply don’t take-off.
There are ways to address these translation friction points. And emerging best practices suggest a decentralised and federated approach, while also re-categorising what constitutes as data analytics resources — there are many types depending on the needs of the organisation. I’ll touch more on this topic if there is interest from my readers :)
Conclusion
So stay tune to this new series of (contrarian) articles that I’m hoping would expand and challenge your perspectives on this continuously evolving practice of data analytics. I look forward to hearing your feedback!
