Contents

CDAO Africa 2019 - Key Themes

On the 17th and 18th of July I attended the Chief Data and Analytics Officer event as a speaker, hosted by Corinium. The event was a great success with the heads of many analytics teams meeting to discuss their successes, failures and other burning data related matters in their respective industries.

Scattered through the success stories showcased at the event were some recurring themes that were homogenous across industries.

1 The CDO Role

Surprisingly, as I was writing this I came across an article about Usama Fayyad, the first CDO, who mentions that the role “started out as a joke”. However this role isn’t a joke anymore - this role is imperative in moving a company towards data success through data strategy.

What was evident from the keynote speakers was that no one really knows what a CDO’s role is. Hartnell Ndungi, the CDO at ABSA Kenya spoke about different versions of the CDO. From CDO 1.0 defined by “where is my data?!” to CDO 3.0 “driving value through data”.

It seemed unanimous amongst attendees what success looks like for a CDO in terms of the Facebook’s and Google’s of the world. But creating a strategy for your own organisation is different to companies built as data-first in mind. And executing on that strategy is another challenge altogether

2 Data Quality

./dilbert-data-quality.gif
source: https://dilbert.com/strip/2018-04-03

Something I, as a data scientist, have harped on about a lot is data quality and it was refreshing to hear the importance of the topic discussed by many of the speakers. Leslie Luyt showcased the importance of data quality incredibly well. He discussed how to bring the business along the data quality journey. Which is important because ultimately, The business should own their data. However, it is up to the CDO team to define what is good data quality and enable the business to take ownership.

3 Everyone Wants AI

Many of the presentation at CDAO focused on how business’ should adopt AI and data science but from the discussion I had outside of these sessions, no one is ready. Quality data is a requirement for good models. It’s the whole “garbage in, garbage out” discussion but at the same time, running models in production has it’s own problems including infrastructure requirements and what needs to happen when these models begin to age (among a host of other problems which are way too long to list).

4 Data Scientists Are Imperative For Data Success

Data scientists were referred to as unicorns. Mythical creatures that are difficult to find. Yet no one knows what a data scientist really is, nor do we really know what they’re supposed to do. This topic was nailed by Guy Taylor and Paul dos Santos in their discussion session “Will the real Data Scientist please stand up”.

One aspect of data science, or at least data products, that was not discussed was putting these models into production. Globally, data scientist aren’t scaling their models into production, that tends to be the job of a machine learning engineer. However, there’s a large skill gap in Africa which means that it is a requirement for data scientists to think about how to implement their models in production and liase and upskill with their data engineering counterparts to deliver end-to-end solutions, rather than running their models locally. Which I believe will be the data science counterpart of hoarding Excel spreadsheets.

5 Privacy and Governance

POPI and GDPR are real things that need to be discussed, understood and adhered to. If we don’t think about these while designing data strategy we’re setting ourselves up for failure. There’s a lot of understanding that needs to happen, specifically around how the POPI and GDPR framework apply to machine learning and AI models and who owns what data.

Rounding it all up

The African landscape is rife with data opportunity. There is still a long way to go until we fully unlock these. As a starting point, data quality is an imperative while scaling into the cloud may close the infrastructure and data science gap.

Shameless self promotion:

I’ll be speaking at Corinium’s Data Science and AI Africa event in November this year. Looking forward to seeing you there!