IN our previous article, we explored how one can take the first steps towards preparing their business for the future. We highlighted the importance of top management commitment and buy-in to digital transformation.
The CEO of the company should lead from the front, and he should be personally involved and committed. I have seen companies where the CEO relies on a team of consultants and junior staff to drive this process. This is a sure recipe for failure. The path to digital transformation is paved with failed attempts with about two-thirds of attempts failing according to a recent survey by Genpact.
The CEO must set out a clear vision and roadmap. He must be visibly involved in the success of the organisation’s digital transformation.
We then stressed the importance of placing data at the heart of how you run your business. This deliberate data centric approach will impose immediate specific demands on your business. Chief among them will be where are you going to store all this new data, how will you access it, what new skills will you need in your business to make it all work and finally how and who will use this data? Further, the other questions you may need to ask yourself are – will the data be stored in the cloud or on-premise?
There are no straightforward answers, and businesses will have to make decisions that work for them. For example, if you run a small trucking company, where you have GPS trackers fitted in your trucks, and your drivers use handheld scanners to confirm deliveries and get clients to sign electronic proof of delivery, then you may want to consider a cloud solution. The advantage with a cloud solution is that it is cheap to set up, and as small business you don’t need to have significant capital outlay in buying servers to store to your data and it is elastic – meaning it’s easier to scale. You also don’t need to employ an IT professional to manage your servers and most importantly you only pay for cloud services you use based on a subscription model.
In our small trucking business example, supposing that you have a fleet of 10 trucks – you will only pay for cloud services for 10 trucks, no more, no less. If for example, one of your trucks gets involved in an accident, and you scrap it, then you reduce your subscription to pay for nine trucks in this example. I am deliberately simplistic to drive home a point here. The opposite is also true. In this scenario, we say the cloud solution is elastic or scalable. This means you can flex your cloud requirements up and down quickly without having to upgrade any servers and physical hardware at your offices. If you had physical servers on-premise it would be difficult to quickly flex up to match new requirements for example. Cloud services are pretty neat and cool, if you think about it.
Some people even say that if you are embarking on a digital transformation, then you must have a cloud-first strategy. I am tempted to agree with them because cloud-based solutions have so many advantages. However, I would recommend that you run a proof of concept first on a small scale before a company-wide deployment. In some instances, you may find that there are advantages in running your own servers on-premise and in other instances, you may want to have a hybrid solution which includes both on-premise and on the cloud. Any choice you make should be based on your business needs.
The next step is to decide what to do with all this data. We must also point out at this stage that data has also become a valuable commodity in its own right and can be monetised. With appropriate privacy protections and user consent protocols, data can become a new revenue stream for business. Apple, for example, is one company which is making smart moves around data. It uses its awesome constellation of products, which include iPhones, iPads, iTunes and the recently launched Apple credit card for example, to bring in customers into its ecosystem products and services.
Once you are in the ecosystem, I would assume Apple can start knowing a great deal about you. For example, with appropriate user consent and privacy protocols they can use location services on their devices to know where you have been, create a heat map of your movements, they can look at your iTunes playlist if they so choose to determine what sort of music you like and they can mine your credit card purchases to learn what you are spending your money on and gain useful insights about your spending habits. This can all be done in line with their Terms & Conditions which users would have consented to prior.
So, if you look closely, you can see that Apple is sitting on a data gold mine, which can provide a lot of valuable insights which can be leveraged if it so chooses.
Coming closer to home, we talked about the EcoCash threat from Facebook. If I was at EcoCash I would (if they haven’t) pivot towards data. But what does this mean? For example, EcoCash has insights into the spending habits of millions of people in Zimbabwe. That information is potentially stored by age, sex, location, product groups, expenditure size and frequency for example. This is valuable information which with appropriate user consent and privacy protocols can be leveraged to create a new revenue stream for EcoCash and somehow help mitigate the looming threat posed Facebook’s Libra crypto-currency, though somewhat.
Accumulating huge amounts of data means that you will invariably need an efficient way of analysing the data and making smart decisions. This is where Machine Learning comes in. Machine Learning is strictly a process of using algorithms to learn from the data, and typically make predictions about that data. The output from the algorithm is then interpreted by a skilled professional typically called a Data Scientist (fancy title) who then also tries to check if the prediction is accurate.
So, for example, a retail chain like Edgars may be interested in predicting the demand for an upcoming period. Typically, they choose to use historical Point of Sale data, weather pattern data, unique events data like the FIFA World Cup, planned future product promotions, massive social media data, individual income data by location and population demographic data by location for example. This raw data will then get fed into an algorithm which can then make predictions about what the future demand will look like. This is important because it will help the company for example plan its staffing levels, decide how stock it needs to hold, decide how much cash will be required to bring in the stock, how much warehousing space will be required and so forth.
So, you can see that Machine Learning is a potential game changer. You may ask – how have companies been doing it in the past? Well, in very simplistic language some companies typically use models based on statistics – where you basically take a sample of the dataset or population (don’t worry about technical terms) and work on it to generate some meaningful forecasts. They then try and measure error, and use the value of the error to tweak and refine their forecasting models. Machine Learning on the other hand uses some underlying statistics concepts but it crunches the entire dataset. This is possible due to recent improvements in computing power. I am deliberately trying to avoid getting too technical in this article, so Machine Learning purists please bear with me.
Some pioneering companies like Amazon are already exploring Deep Learning, which is basically the same as Machine Learning with the only difference being that the algorithm decides what to do with the results and also if the predictions are accurate. Deep Learning relies on Artificial Neural Networks (very big words, and in simple terms it just means to simulate the working of a human brain). Deep Learning is computationally intense and requires massive amounts of computing, and we will park this for now.
Now, it is important to point out that it is really expensive to invest in computers which can run some of these algorithms on their own, and various companies like Amazon, IBM and Microsoft for example, offer Clouding Computing Services. Meaning that instead of investing in your own infrastructure to run these models, you can push the data into the cloud, set up virtual machines (VM) inside the cloud, build a model in the cloud and spin your virtual machines (VM) to get an output. Some of you may be asking, this sounds hard and it sounds like complicated science and maths stuff, how can I get skills in this area? Well you need two things. First you have to be curious and second, you can enrol in online courses on MOOC learning platforms like edx and Coursera. You will be surprised how much you can learn by investing a few hours per day in studying online.
So to summarise, this week we looked Machine Learning (ML) in very simple language. We explored what it can do, and we also briefly touched on how individuals can learn these skills and future-proof their careers. Next week we will look at Artificial Intelligence and get a taste of how one can use AI in their business. As usual case uses will be used to drive home the ideas.
You can contact Tongai Muroyiwa on e-mail: @tongaim. Twitter: @tongaim
Source: Zim Live