STEM newsletter

How does incremental demand work?

30 April 2002

STEM is and has always been a demand-driven model. The direct cause and effect between service provision and network roll-out is precisely what enables STEM to deliver a faithful set of direct and fully-allocated per-connection (or minute or Mbyte) cost results.

For any requirement between Service and Resource, a constant basis is selected, so that for the purposes of the given requirement, a single measure of demand is considered, be it connections, annual traffic, busy-hour traffic or even annual revenue. (For data services, annual and busy-hour traffic are typically interpreted as data volume and peak bandwidth.)

The heart of network dimensioning in STEM is very simple: it has to be for the sake of ready verification, and in order to be applicable to a broad range of technologies such that strategic comparisons can be made.

Incremental demand

For a new Service, we consider its demand, divided by the capacity of a Resource, and install a whole number of units of that Resource. Central to the proposition is that STEM should analyse the incremental demand in the following year (or period, if running with quarters or months). So STEM stores two pertinent data:

  • the number of units of the Resource installed in each period
  • the capacity of the Resource used by the Service (of which there may be several).

In a subsequent period, STEM first examines the installed base of the Resource to see how many units (if any) have reached the end of their physical life and must be replaced if still required. If a Service were using some old capacity which has now expired, then its residual usage of the Resource will have decreased. The current demand from the Service is then compared with the residual use in order to calculate the incremental demand.

If there is existing (possibly newer) slack capacity, then STEM will use this before installing new units. The Resource Other Details dialog contains an input, Use Slack, which governs the age priority for using existing slack. Conversely, if Service demand decreases such that residual use of a Resource exceeds the new demand, then a second input, Make Slack, governs the age priority for making used capacity slack.

Once the Service-driven demand dimensioning is complete for all Services with a requirement for the given Resource, then a number of supply-side factors are considered.

Maximum utilisation

Network planners usually prefer to avoid running the network at full capacity for reasons of both resilience and future provisioning. The Maximum Utilisation input enables you to specify that a given Resource should never achieve a utilisation higher than the specified input. Thus, after the demand from all Services is considered, STEM installs additional units of the Resource if necessary to reduce the overall utilisation.


The Resource element may represent a generic box – say a switch – which must be installed at a number of distinct and geographically separate sites to support, collectively, the total Service demand as distributed over those sites. Depending on how the demand is actually distributed, the impact is an increase in the likely slack capacity required in the network, up to the theoretical maximum of a whole unit in each separate location.

The Deployment Sites and Distribution inputs for a Resource allow you to describe the scope of the geographical diversity – typically in reference to a separate Location element – and the nature of the distribution. The net result is another prescription for minimum slack requirements, which STEM will satisfy by installing further units if necessary.

Planned units

In reality, there may be other economic or pragmatic factors which drive the roll-out of equipment, such as a fixed-rate programme defined by manpower constraints. The Resource Other Details dialog contains an input, Planned Units, which is designed to capture such an external prescription for cumulative installed units, and STEM will meet this target by installing further units if necessary on a year-by-year basis.

However, this input is currently handled only as a minimum planning constraint, as we have not yet implemented a supply-constrained STEM model. The actual Installed Units result can be compared with the Planned Units input to see where planned roll-out must be reviewed.

Supply Ratio

Within each individual requirement from a Service for a particular Resource, there is a Supply Ratio input which multiplies the demand that the Service places on the Resource, e.g. to represent the routeing of traffic through a network. If only 25% of the busy-hour traffic is international, then a Supply Ratio of 0.25 would be appropriate for Resources providing international transmission.

The impact of the Supply Ratio is two-fold. Firstly, there is the obvious multiplication on the effective demand from the Service as mapped onto the Resource when calculating new demand. However, in order to calculate incremental demand, the Loading is also used to divide existing used capacity appropriately, to facilitate a consistent comparison of new demand with residual used capacity.


If several Resources have the same logical role in a network, then they may be grouped together with a Function element so that STEM can effect a migration of demand from one Resource to another, e.g. to model a technology update. Such a substitution is effected through a time-series shift in the Mapping inputs for each of the requirements for the respective Resources in the Function.

STEM will aggregate the residual used capacity of all Resources in the Function which have been used by a particular Service when calculating the incremental demand from that Service within the Function. That incremental demand is then split between Resources according to the current values of the Mapping input for each requirement, thus allowing expiring capacity of one Resource to be replaced by new capacity of another.

If a migration must be accelerated beyond the pace of natural replacement, then the Churn Proportion input for each requirement enables residual used capacity of a Resource to be recycled as new incremental demand.

Pre-run installation

Resources installed in the first year of a model run – either Y0 or Y1 according to the Include Year Zero input – may be modelled as a historical installation, governed by the Max. Age of Installed Units input in the Resource Other Details dialog. This feature enables a proper distinction between new and existing equipment and the calculation of suitable Capital Expenditure and Depreciation results.

A by-product of this feature is that equipment will come up for replacement sooner than if all the equipment were installed brand-new, as the default inputs have an even installation of all possible installation ages, the oldest of which will be due for replacement one year into the model run.

© Implied Logic Limited