Can Predictive Analytics ensure all the targets set for the network?

Sometimes I hear people talking about Predictive Network Analytics, expecting it to solve all their business challenges. On the other hand, some say it can only be used for network performance optimisation. Well then, how is it? I’ll start with some numbers from operators we have worked with:

  • 88% of customer quality can be managed with network actions
  • 42% of customer satisfaction can be controlled with network quality
  • 25% of total churn can be controlled with network quality
  • 5% of total revenue can be controlled by the yearly network CAPEX decisions

This is the leverage that a mature MNO can have with its yearly mobile network CAPEX decisions, not only on quality but also on customer perception and business KPIs. The rest is unrelated to the network, driven by e.g. service plans, pricing, marketing, and competition.

Setting the target

So if the question is whether Predictive Network Analytics can ensure any operator business target, the answer is most likely no. But it can maximise the impact your network has on those targets. And if the targets are realistic, it may even be able to ensure them. But even more importantly, it allows the operator to set specific quantified targets in the first place and to manage those targets successfully.

And what do I mean by a target? It can be anything that someone in the operator organisation expects to achieve by investing into the mobile network. Preferably those targets are defined by, or at least have a direct link to, some specific business KPIs.

  • Head of Services & Products may expect capability to offer new OTT applications
  • CEM Directors will expect improvement in customer satisfaction
  • CMO wants to see lower churn rates
  • CTO could target to gain quality leadership in the market
  • CFO may simply want to ensure that ROI of the spent CAPEX will be positive

The targets have been similar for years, but how have these targets been managed after the data traffic explosion in the beginning of the decade?

Era of LTE rollouts, 2012-

When the MNO management board discussed next year’s network CAPEX, it was dominated by the race for LTE rollout. The targets listed above were strongly present, and all of them were expected to take a huge leap with the LTE performance. The fact that competition is doing exactly the same and that customer expectations will inevitably have similar leap was ineligible as the LTE rollout was a necessity to focus on anyway.

And when the operator was finished with the LTE coverage rollout, justification of the incremental CAPEX was not that easy anymore, but the principle remained the same: “Deploying the second LTE layer will improve quality and customer satisfaction, and keeps us in the competition”.

There was no reason to question these plans, as the quality and capacity impact achieved was gigantic. In many markets, these plans were further boosted by the network vendors offering lucrative wap contracts to make it irresistible. These were the good old times for CAPEX planning.

Looking at the plans above, we can conclude that they are volume-based, stemming from a count of technical actions or count of customers with a new technology. The plans were very little driven by the knowledge of where and to which customers they are really needed for maximal business impact.

I would call this the Era of Volume-based network investments

Era of Big Data, 2014-

After the LTE coverage rollout operators are seeing the rapid surge of traffic and increasing customer expectations. Now they cannot rely on new capacity layers anymore, but instead need to move into more selective capacity expansions with high-band capacity carriers or new macro and small cells, for example.

Now that Big Data and Data Analytics have come into play with many telecom operators, the actual areas and sites selected for capacity upgrades are prioritised by a lot of collected customer and business intelligence data. The decision-making is based on multiple KQIs and multiple customer value components, all aiming to maximise the value of the CAPEX spent. These are appealing and credible arguments for the decision-making.

But at this point it is very valuable to notice that the CAPEX plan examples mentioned above are priority-based, i.e. the criticality of an upgrade is weighted by the KQIs measured in that location. But those plans are not in any way indicative that the business targets listed above will be met. They are merely optimistic plans that should have the best impact in the right direction. So they are Best Effort. Having studied many operator solutions closely, this is still the case with all of their Big Data analytics: Prioritizing KQIs and doing de-facto best-effort network investment plans.

And when the plans still don’t suffice those business targets set in the beginning, the blaming game starts. CMO blames the network for not satisfying the customer quality needs despite the spent CAPEX, CTO blames the CMO for too generous data plans crunching the old and new capacity.

Why are the targets not met – because the plan was not really driven by the targets but rather priorities.

I would call this the Era of Priority-based network investments.

Era of Predictive Analytics, 2016-

Wouldn’t it be great if operator organisation had common business targets, and the directors from different domains knew exactly how much and by what means they can contribute to these targets?

Let’s take the CTO, and assume an ambitious company-level target of decreasing churn from 16% to 10%.

How could the CTO commit to this target? The only way for the CTO to commit to the common target is to provide him with the following knowledge:

  • How much is the churn driven by customer quality?
  • How much influence can my network have over the customer quality?
  • What technical solutions are required to achieve this influence?

Predictive Analytics can provide this knowledge, and it can provide it even without Big Data. As long as there is some CEM or BI solution collecting customer level quality KPIs, in addition to OSS data, you already have the needed data. It’s there.

As an example, the CTO would now know that:

  • By end of the year 25% of the churn is caused by network quality, resulting in a total 4% of customers churning due to poor network quality. This is the theoretical maximum impact the CTO can manage.
  • Eliminating that 4% churn requires multiplying the micro site count and is in no way feasible.
  • But reduction of the network driven churn to 2%, from the 4%, costs a fraction of that, at €18M.
  • This 2% reduction in churn can be turned to retained revenue and the business case for the CAPEX can be verified profitable

Now the CTO is comfortable to commit to the common churn target with his own target contribution of 2% reduction. And at the end, when the result of this campaign is being reviewed, predictive analytics can verify just how much of the total achieved churn reduction actually came from those network improvements.

This was just one example, and the same can apply to any target: Customer satisfaction, OTT service capability etc. There can be also very straightforward targets for the CTO, such as quality leadership. Let say the target is to win the next Ookla Award in the market. By combining field-tests or crowdsourcing data with the operator’s own network data, predictive analytics can tell the CTO exactly what expansions are needed to win in the next Ookla benchmark (or any reliable competition benchmark). Next time when the CMO is complaining about lagging sales and demands the Ookla award for better customer perception, the price tag of it can be put on the table and compared to the expected sales and retention figures it should generate.

The essence in these third era examples is that the value impact of the network investment is predicted, so that realistic but concrete targets can be set and it is known what exactly is required to achieve those targets.

And as the investments are always prioritised according to their impact on the target vs. their cost, it also enables at least 20% CAPEX efficiency compared to KPI/KQI based plans from the Second Era. Similarly, if the target is to maximize the ROI of the network expansions, the incremental revenue return can be 40% higher from the same incremental CAPEX.

Thus, I will call this the Era of Value-based network investments.

How to change

Changing from the era of priority-based to value-based network investments does not require simply new predictive analytics solutions. Instead, it primarily requires a change in the mindset. The Big Data systems built are excellent solutions for the data mining, but the analytics running on top of it defines the true value of the solution. I understand that the KQIs and business KPIs measured by Big Data are the leading edge of many operators’ capes efficiency programs, and the only thing I want to change here is the focus. Focus away from the amount and details of the data, and focus towards how this data can ensure the company targets. I’m positive that the detail and near real-time data available does improve the value of predictive analytics even further, but first it has to be understood that the data is a means, not a purpose.

And as for the practical change, that’s easier. Maybe the first step is to showcase the added value by a concrete example that solves a real issue you have on the table. An issue where you need to know what is needed from the network to reach your existing target, or how to maximise the impact of your existing CAPEX budget on that target. After a successful showcase the change is easy to expand for overall network CAPEX management. When it comes to the facilitation of this, Omnitele is happy to assist you.

-Rauno

Visit us at MWC in Barcelona, 27th February to 2nd March, to understand how Omnitele CoDriver™ could be used to transform your network. You can find Omnitele in Hall 5 at stand number C45. Book the Meeting!

Subscribe to our newsletter

get the latest news delivered directly in your inbox

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.