Q: Thanks for taking the time to give us some insight into Derivco and your role as Market Insights Manager. Before we get into the detail, most people probably don’t know Derivco. Can you give us some background into the company? (what they do, how long, employees etc.)
Derivco was created in Durban 20 years ago, with around 10-15 employees who were brought together to develop the world’s first online casino. It’s been a massive success story since then. We are now more than 2,000 employees strong, with offices based around the world, including the Isle of Man, Australia, Sweden and Estonia – just to name a few. We continue to create world-leading, entertaining online gaming content and the back- end systems to support it.
Q: It’s good to get a perspective on data analytics from different industries. What is your main focus then as Market Insights Manager for online gaming company? Where and how do you collect data? And what insights are your driving for the business?
My role as Market Insights Manager is to drive progress of the business with insights derived from data. We have a remit to “constructively disrupt” the business, and we do this by creating models and analyses which facilitate, aid or probe the product decision-making process. With each product release we analyse live technical metrics like usage and performance, financial impacts and player impacts, like how our products affect the behaviours of existing player segments or the uptake on new or reactivated players. We use this data to measure and optimise the performance of live products and to aid the UI and UX design decisions for new product design. Creativity is a crucial in our product design but wherever possible we either support or challenge design decisions with data.
We collect data from various sources. Typically for product analysis we use a combination of Google analytics and event tracking which writes events, session and usage stats directly to our warehouses. We build event tracking into every feature we develop. This can be something as simple as a new button or an entire function or game that the user interacts with on the interface. We spend a lot of time whiteboarding prior to development to agree what we want to measure and why, what our expectations of the features are (KPI’s), and we spec out the reports and analysis required to measure it all. We’ll ensure that these models are in place and ready to provide the measurements we need before the product goes live.
In terms of our data collection strategy, we first identified the areas of the business requiring data-driven support. We then defined data marts requiring development to address these areas, and ranked these in importance to build iteratively. So, we developed the most critical data first, ensuring we could meet initial business needs and we add further data marts as the requirement to dive deeper into the data evolves. We have a team of BI developers, business analysts and Data Scientists who consume the data through various front-end tools and collaborate closely to create and action the insights.
As a business, we are now open to cloud storage, but this is a very new reality for us. In-country regulations in terms of what we can store and where we can store is a crucial factor in determining how and where we collect and store data as well as the data we’re allowed to store.
Q: I’d imagine that a company like yours is already a ‘data analytics driven’ one. Is this the case and has it always been the case? (If it has always been – what value has that created?; if not – how did you transform the thinking of the organisation to become data analytics driven?)
We weren’t always analytically driven. We broke ground on online gaming and we are still a market leader, but if we are honest with ourselves we enjoyed the benefit of aspects like first to market (of course backed up by brilliant product) for many years. We could operate in an environment where being analytically driven wasn’t a requirement. But the landscape has changed immeasurably. Competition is fierce, regulations are strict, business models have changed and our customers operate within tighter margins. Players have countless options in a saturated market. Success now lies in surfacing and acting upon the small percentages, targeted and personalised player management and constantly optimised product offerings. This is where our requirement to be data driven has taken off. We have had to become more scientific about our product design, development and player management. Knowing exactly who our players are and how they behave is crucial, and you can’t do that without analytics.
From a business buy-in perspective we’ve been fortunate in that we have top level buy-in for a data-driven strategy. We’ve been given the space to formulate and execute our data strategies. Of course, we’ve had to deliver on that faith and across the company the small initial pockets of analysts are doing so. Investment in these teams is therefore increasing. Analytics teams are growing and flourishing.
From a product perspective, the transition from gut or opinion based decision making to data-driven decision making can be slow because behaviours or habits need to be changed. You have to win people over and to do that you need to deliver tangible value that either influences the decision-making process or realises actual changes in live product or player management. This was a big reason behind the iterative approach we took to analytics. We keep the longer-term data strategies in mind but to buy time to work on these we find ways to deliver iterative value.
From an operational/development perspective, we don’t overwhelm the development team with requests. You don’t have to track everything, so our list of requirements is always well thought out and there is a clear expectation that the tracking we require is not negotiable. Analytics are now built into every feature developed and they form part of the acceptance criteria of the product. So, while in the past analytics were an afterthought or often the first item to be dropped from scope by a busy game development team, that’s no longer the case. A product without analytics will not be deemed a complete product and will not be signed off for production release.
Q: What are some of the ‘new’ technologies/methods you’re using to derive insights from data and make business decisions? (predictive/prescriptive analytics, AI, ML, deep learning)
AI is an interesting one because depending on your perspective you could say we’re using it or we’re not using it. For me (because of my age!) AI or ML tends to conjure up images of super intelligent and indestructible computer beings called terminators, hell bent on exterminating humanity! But on a serious note my definition of AI is that it needs to be pretty exceptional to be called AI.
For me AI needs to involve some form of machine learning (do stuff the developer didn’t ask it to do) which must result in actionable insights like intelligent interactions with the player. We’re developing algorithms and models which are designed to solve problems such as predicting and reacting to player behaviour, but this technology is more accurately described as predictive analytics, rather than AI. We intend to implement ML and AI models and there are several ideas we’ve already spoken about and planned, but I wouldn’t say we’re using it just yet.
We use both descriptive and predictive methods. In the product and promotions/marketing space we will typically use descriptive analytics to measure performance, player behavioural impacts etc. Our network management models and tools are predictive (even a little prescriptive) in nature in that they use past and simulated data to predict the outcome of the product/ configuration decisions which our analysts need to make. This helps them determine the best way forward for our products and enables them to keep making optimisations, but within acceptable risk limits.
We also use fairly “general” predictive models like player churn and stickiness which can then be used to reach and engage with players proactively.
Q: At DataCon Africa you’re going to be talking about how you’ve built the team around you. Can you go into some detail on the approach you’ve taken, the structure of the team and value its driving for the business?
They key for us was to build a team of complimentary skills. We got a bit lucky in the beginning because we had a team of business analysts with a wealth of business knowledge who evolved into the analytics team as we hired BI developers. Your business analysts understand your business, products and customers but they need data and insights to grow the business. Skilled BI developers, data analysts or data scientists who have the necessary skills to create complex data and models provide this data. Working together, the synergy we formed between these roles has been exceptional and crucial to our success.
We also hired very carefully, identifying the skills required to reach each of our business goals and placing people in those roles. As a result, we have full reach to deliver within our own team: From owning our data architecture, transforming data, developing our own reports, analytics and complex models to possessing a deep understanding of our business and customers.
At the same time, we’ve been careful about not creating single points of failure – so wherever possible each person in the team is exposed to other parts of the business or development cycle. It’s great to have experts within your team who will make best use of their strengths in a certain area but it’s also important to me that they can add value or step into other areas.
We’re still a relatively small team, but being small makes us agile. We drive a culture of collaboration and a mantra of share early, honestly and often so that if we miss the mark on a piece of work or make a mistake and need to retrace our steps it’s not going to be too expensive.
Ultimately, a team with cross skills with and the maturity to collaborate closely is what has stood out the most for me.
Q: As you know we (Corinium) are putting a lot of focus into data science skills development in South Africa and will be launching the first Data Talent Summit in June 2018. What do you think are the current blocks to the growth and development of quality talent in the country? What do you think can be done to overcome these?
The definition of Data Scientist is important to me because the role encompasses a broad range of skills and there are specialisations within data science. I think industry and organisations like the Talent Summit needs to acknowledge that not all data scientists can be expected to have the same or all the skills. Think of the specialisations you get within the legal profession. You can’t expect a lawyer to be an expert in every single area of law. They specialise and develop within a certain area of the skillset. I think it’s the same for data scientists.
The non-negotiable “broad” skill set of a data scientist is advanced mathematical, statistical, analytical, problem-solving and SW development skills. I would add the ability to communicate efficiently and effectively to this set. Data scientists or aspiring data scientists are very bright and technical people so if I can generalise their first love is often going to be technical, rather than interpersonal. If you can create a complex model, present it and sell the insight to your audience so that they action it…you have your perfect data scientist.
Communication skills should therefore form part of any formal training, because you need your data scientist to have the ability to deliver a presentation which convinces the audience to take action based on their findings. Other less obvious areas I feel should be developed are UX and UI skills. Users can be diverse. Understanding design principles and how to build an interface that meets the user’s needs is crucial.
Barriers – lack of experience in the industry stands out for me. Coming straight out of university, where a great deal of knowledge gained is theoretical, data scientists aren’t given the chance to gain experience. They need experience working with real data, and to develop domain knowledge within the industry they join. Businesses need to give data scientists the space to develop these practical skills. I think there are incorrect perceptions that given their intellectual skills data scientists can simply hit the ground running and this can be a difficult expectation to manage.