New example model for a cable-network operator

Implied Logic has just released a new example model illustrating how the STEM visual software for the reliable modelling of business can be used to measure the extent to which capital investments in local transmission technology and configuration may help to secure or retain future Internet revenues.

As with our other recent models on FTTX and VoLTE, a web version of the model is presented online using our latest eSTEM capability. A detailed description of the methodology is available to review offline and share with interested colleagues.

As well as its evident topical interest, the underlying model also captures the common situation of a constrained resource. Input parameters govern how quickly different types of customer will disconnect when it is perceived that performance is diminished, as well as the independent timing of the range of possible network remedies described below.

Increasing cable bandwidth to retain high-value customers

Internet provision via cable is carried in the channels not required for digital (or analogue) TV, and is inevitably a shared resource between all the households connected on a given segment. Unfortunately there is little or no brand loyalty for what is typically regarded as a commodity service. A heavy user will always prefer the fastest downlink and uplink speeds possible, and any moderately-active customer will quickly consider any available alternatives if contended bandwidth is too limited when they need it.

We have modelled the current infrastructure of a typical cable network operator, including Internet CPE, set-top box, optical node, amplifier and head-end, together with future options to address bandwidth limitations such as channel-bonding (if not already implemented), enhanced video compression, RF extension, and segment splitting. According to readily adaptable estimates of the costs involved and the elasticity of relevant market segments (including level of competition), the model compares the cost implications of implementing these upgrades with the potential revenue loss associated with not acting in a timely manner to retain customers.

Figure 1: Simplified cable model interface live on our website

Modelling capacity limitations

STEM provides excellent built-in capabilities to drive network rollout from demand by providing enough capacity to support the requirements. However, there are situations where an increase in capacity is technically not possible or reasonable or allowed. These situations typically happen with shared resources such as:

• segments of a cable network which are used to provide TV and broadband access and support a number of customers
• passive optical network (PON) systems, where several customers share a common feeder system
• mobile networks, where further cell capacity additions are not possible due to limitations of spectrum.

Shared and capacity-limited resources lead to degradation of service quality, such as higher blocking-rates for voice services or lower bandwidth-per-customer for data services. Depending on the market situation, lower service quality can lead to decreasing customer numbers – which may of course temporarily improve the situation for the remaining subscribers.

The remainder of this article describes our approach in a little more detail:

• limiting the demand-driven capacity increase
• dynamically decreasing the number of subscribers according to the network utilisation
• alerting the user of the model about such an overload situation.

Limiting the capacity increase

Limiting the demand-driven capacity increase to a predefined number of resources (e.g., one cable segment with a given capacity) is quite easy to achieve. The driving demand must be analysed in a transformation and its output can then be limited to the maximum capacity supported. There are a few more implications:

• as we will certainly assume a ‘fully-loaded resource’, the Maximum Utilisation input should be left at the default value of 100%
• the number of resources in the network must be controlled manually, which can easily be achieved with the Planned Units input.

Figure 2: Limiting the output of an Expression transformation

Affecting subscriber numbers

The utilisation of resources is normally influenced by a number of factors including subscriber numbers, traffic per subscriber, contention, and so on. Therefore the ‘theoretical’ utilisation can’t be pre-calculated and must be derived at runtime. This would typically lead to a loop, but this can be readily decoupled or resolved with the help of a Time Lag transformation.

Two indicators can be used to indicate an overload situation, which would lead to service degradation and/or decreasing subscriber numbers:

• the utilisation of the capacity-limited resource
• the ratio of capacity provided to ‘theoretical’ (i.e., unlimited) capacity requested of that resource.

Once a certain utilisation threshold is reached the model influences the Penetration input for the subscribers driving the resource, thereby decreasing the demand side. Depending on a number of influencing factors, including time-resolution of the model, thresholds, and changing traffic-requirements-per-subscriber, the feedback loop may oscillate or resonate. This is reduced by applying the feedback not to the total penetration number, but only to the gradient.

The level of competition is assumed to be the key factor for subscriber churn. Strong competition is reflected by a low utilisation threshold; subscribers leave when the slightest service degradation occurs. In the opposite case, if there is no competition (i.e., no choice) subscribers will stay and the average bandwidth per subscriber will decrease.

Figure 3: Feedback loop to affect subscriber numbers

Note: The shortest lag supported today by STEM is one year, which would fit with reality in a number of cases. It takes some time until customers notice the decreased service quality, decide to change their subscription to a new operator, and are able to exit their current contract.

Flagging an overload in the results

It is important to flag overload situations to the user of the model. This could trigger ‘manual’ additions to the capacity or other measures to improve performance. Overload can be easily derived from the resource utilisation.

Another important indicator is the resulting bandwidth per user in case of overload, which gives a better indication about the level of service degradation.

Figure 4: Using transformations to calculate a number of performance measures

My account (log-in)

© Implied Logic Limited

Contact | Downloads | Core team and advisory board