How Machine Learning Will Prepare You for 5G

Cut-throat competition, decreasing ARPUs, OTTs encroaching traditional revenue sources, - these are not new and there isn’t much to add to the discussion.

The key challenge for modern telecoms is to make a strategic investment in the face of decreasing B2C revenues and upcoming 5G standards.

How do you do that?

How do you distinguish the high value customers from those that are dragging you down? How do you know which network uplifting is necessary to keep revenue generating clients and increase profits in low revenue segments? How do you distribute a finite pot of investment money to achieve a maximum ROI and stay in business?

In the times when companies maintain healthy financials by selling shares only to buy them back in the next quarter we cannot expect telcos to engage in “network philanthropy”. An operator needs to know when the upgrade is needed and which market segments should be targeted first.

They need to be crystal clear on who the key customer is and where their bread and butter is coming from. Investments should be prioritized accordingly.

The way investments have been made historically was by a “guess and pick” method, - huge expenditures to build nationwide networks with the hopes that the end client was going to appreciate the speed and reliability. MVNOs were luckier than the first movers, - using their own brand to capitalize on the entrants’ infrastructure. Later business intelligence was employed to create fancy forecasts and equip managers with the best practices. BI was a guiding star of multi-billion consultant agencies, who utilized the best minds to derive insights from carefully picked data.

What has changed since then? Scarce structured data has been substituted by countless amalgamation of data sources dumping exabytes of overwhelming and hardly readable big data.

According to Wikipedia, big data is data sets that are so voluminous and complex that traditional data-processing application software are inadequate to deal with them. It is precisely big data that moved traditional business intelligence tools to the dustbin of history.

Want to make sense of your big data? Forget Business Intelligence. Machine Learning is here for you!

Luckily, telecom is the industry that has both the data and the money to harness it and afford major strategic initiatives.

Two and a half quintillion bytes or 2,500,000,000,000,000,000 bytes. That’s how much data humanity generates every single day. And the amount is increasing; we’ve created 90% of the world’s data in the last two years alone. Indeed, telecom is generating more data than any other industry. And when not generating, there are telecoms that have to transport and store it.

The fact that we can be talking about the analysis of all that data and the generation of valuable insights is only due to the increased capabilities of cloud computing and the proliferation of self-improving programmes.

While AI is not a new discipline, the unstoppable progress in cloud computing and vast amounts of accumulated data have finally provided an ample ground for this Machine Intellect to become a wholly equipped party saver.

Today ML allows telecoms to evaluate all parameters and run simulations to see which network element needs an uplift, how targeted investments will impact a particular segment of market, and which technology upgrade will be optimal for a particular type of geography and customer segment.

McKinsey Era

The growth of the Big 5 has been fueled by an increasing amounts of data and a genuine foresightedness at the Executive Level. As the time of shameless monopolies and guilds gave way to market economy and “survival of the fittest” the need to stay ahead of the game and anticipate rather than react became paramount. Business schools have been churning elite armies of sharp and hard-working problem solvers. For decades these people have been performing 2 crucial functions: making tough strategic decisions and serving as scapegoats if those recommendations do not lead to a desired outcome.

No matter how smart and equipped a consultant is, - without machine help he will only go as far as delivering an “informed guess”. This approach might have been great for XXth century but is suboptimal in the era of big data.

The Ascent of IBM Watson

Then came IBM Watson, a ground-breaking AI solution, combining best from the both worlds, - carefully picked machine learning algorithms powered by the descendants of a first “computer” and stellar IBM Global Business Services consultants that give you an insight based on 97% certain algorithms - generated answer. IBM Watson presents developers with a straightforward API that can be deployed as a micro service or any arbitrary component of a project.

The way Watson system works is through the creation of an organisation model and running multiple iterations of algorithms to teach the model different patterns and see the future rooted in the past data.

While bearing a cost of both the technology and human capital, the solution brought incomparably better business outcomes confidently occupying top of the mind of numerous ingenious executives.

MachineOS Approach

It is hard to beat the IBM brand and billions of investment in AI engineering. And while IBM, DARPA and Google are the most probable producers of a world-changing AGI, it is smaller and more agile companies that are going to tweak and twine the machine learning approach to business challenges.

Just as Nietzscheans and Epicureans, Machine Learning experts are divided into 5 major schools of thought:

·      Bayesians believe learning is a form of probabilistic inference and have their roots in statistics

·      Analogizers learn by extrapolating from similarity judgements and are influenced by psychology and mathematical optimization

·      Symbolists view learning as a universe of deduction and take ideas from philosophy, psychology and logic

·      Evolutionists simulate evolution on a computer and draw on genetics and evolutionary biology

·      Connectionists reverse engineer the brain and are inspired by neuroscience and physics

While it important to refine your style and pick the right group of algorithms to solve a particular problem, it is the right combination of selected algorithms that differentiated machineOS and allowed it to establish a name in an exciting world of Artificial Intelligence.

machineOS uses unstructured dumps of raw data to move it through a patented AI platform which creates a real time organisational mimic. The mimic is made up of millions of interconnected nodes (agents, routers, CPEs, servers, clients). Each node becomes a strong processing engine which runs a multitude of functions emulating natural selection. It is precisely the competing nature of clashing ML schools that allowed the start up to achieve staggering accuracy and speed.

Once the best outcome has been established on the individual node level, each node cross checks its result with an adjoining one. Just like memories in our brains form if the connection between neurons is strong enough, nodes cross-check their conclusions and form a hypothesis reinforcing link.

Unlike IBM, machineOS is applying its learners to each and every of its billions of nodes, which allows it to dynamically update the process and conclusions with a slightest change of parameters or the addition of new data.

Capitalizing on its competitive advantage the start up managed to provide its clients with meaningful insights in 2 weeks’ time as opposed to a year time of an IBM engagement.

The changes in network congestion, customer churn, new regulations, localized network deterioration and constantly changing priorities and management decisions – the number of investment influencing parameters could not be bigger than the ones following telco operator’s journey into a brave new world of global connectivity and 5G.

It is not enough for the platform to perform a fast data analysis and produce measurable and actionable insights. It has to do it dynamically and react to the change of inputs in real time.

After all, there is little use in the recommendation to make a spectrum upgrade in profitable network site when a major deterioration has occurred in your strategic clients’ location right after the platform finished its year-long engagement.

Operators have resources and data. What they have a shortage of is time. The advent of 5G will exacerbate the problem of real-time response tenfold. This is why operators need a solution that not only can help them prepare for 5G but recommend strategic ongoing investments once 5G is implemented.

Strategic investment is the single most important decision we can make in our business. And if Machine Learning can help us avoid the mistake Blockbuster made back in 2000, we are only reasonable to outsource number crunching to algorithms.

For more information on how machineOS can help your company pick a data-backed investment strategy, contact me at [email protected]