The Important

  • Quality has the ability to increase or reduce service pricing
  • Quality is an often focused on attribute in network design
  • Quality can be over-engineered if it exceeds the level that would lead to further price increases


I am currently working on what is tentatively called “The three olive martini model” of network architecture/design. In Network Architecture: Why Choices Must Be Made, the top half of the model was used, but only one olive was in the martini glass, politics / society. This articles adds two more olives: Outcome & Return, with a focus on quality as a determinant of demand. To explain the model a little more explicitly, the laws of physics and economics are viewed by the model as being somewhat rigid, and therefore are used as the edges of the glass, where as other factors that influence why choices must be made, are more variable / fungible in nature, and therefore are the olives.

Figure 1. Top half of the “three olive martini model” of network design


Every good strategy / design has an outcome as the reference for choices that are made. If you don’t know where you want to go, any road will get you there. Businesses have business outcomes: grown earnings per share, having an operating margin of x%, gaining 2 ppt of market share, growing revenue by 10%, etc. A goal without a plan is just a wish, and all these kinds of sentiments are just that, wishes, until they are boiled down to a thesis on how they will be achieved, and ultimately a plan for doing so.

Translating business goals into network design outcomes, is, I suspect, not always easy. Business people do not always speak the language of network designers, nor understand the costs and tradeoffs involved in achieving design outcomes.

A business manager may come to a network designer and say the network operator is entering the layer 2 VPN service market, because that will grow revenue by x%. That might just be the beginning of the conversation though: which cities, how many customers, what are the target customers doing already, etc. Questions that might help a network designer translate requirements to scale, interworking/migration, deployment schedule, etc.

One perennial conversation in networking is around quality. Quality is how many conversations are framed. It is an intuitive conversation. If the network is not available and delivering packets without errors, where they are supposed to be going, then what is the point of the network? Part of the conversation might be experience, someone “screaming” at the networking people when the network is not working.

Quality is a determinant of demand

Depending on what economics literature you look at, quality(1) will or will not be explicitly called out as a determinant of demand. What is a determinant of demand though?

Figure 2. Quality as a Determinant of Demand

In figure 2, two demand curves/schedules are depicted. D1 is one demand curve, which depicts the relationship between quantity demanded and price. As price changes, quantity demanded changes along the points of the D1 curve.

Where the points in the D1 and D2 curves are on the graph, is impacted by determinants of demand. For example, if D1 is a demand curve for a network service, then increasing the quality of the service may shift the demand curve to the right, example D2, creating a new demand curve, where prices for equivalent quantities are now higher. Ultimately, most businesses would see that as a win. Producing X quantity has a cost, so the higher the revenue/pricing for that quantity, the greater the profit.

Other determinants of demand include:

  • Economic conditions / financial performance of service buyers. If the economy tanks, or a buyer falls on hard times, the buyers demand for a service may decline, shifting the demand curve to the left.
  • Change in preference: for example IP VPNs instead of FR/ATM. Coke vs Pepsi. Future preferences can be hard to predict, but some insight can be mined by following industry conversations.
  • Complementary goods: when the price for a complement decreases, then the price for complements may increase. For example, let’s say an operator is selling a managed WAN service that consists of CPE access routers and a cloud-managed service. If the price of the CPE access routers goes down, then the operator may be able to increase the price of the cloud-managed service up (putting aside for now CAPEX/OPEX differences).
  • Substitutes: If Coke increases their prices, then the demand for Pepsi may go up.
  • Market size: The higher the market size, the higher the quantity demanded for a given price point.
  • Price expectations: if buyers believe that future pricing will be lower, they may delay purchase.

How many of the above can the network designer directly impact? Certainly a network designer can respond to changing preferences, but what else? Not much. For this reason, I suspect that network outcomes are often framed, consciously or unconsciously, as some combination of supported services and quality (reliability, availability, performance, error rates). All things equal, for example no other differentiation, buyers will pay more for a higher quality service. This is in addition to the intuition most network professionals would have anyway about the network being available, and the pride taken in the network.

Outcomes are part of the “why” of choices because they are drivers of choices. Based on markets served, services supported, and other strategic issues, does the network need traffic engineering on a service by service basis, is a few virtual topologies sufficient for different classes of service, is quality of service important because of the relatively small capacity and the quality expectations, etc.


How outcomes are achieved is important, and the subject of future articles. How they are achieved is to some extent a function of the relationship between (desired) Outcome and (desired) Return.

Would it be able ok to design a very high quality service with significant amount of future capacity built into the network? Maybe only if there is an acceptance that the return will be low in the short-term while the business is growing.

Are the return expectations high relative to the desired outcome? Then maybe quality has to be managed with sophisticated traffic engineering and QoS instead of increasing capacity. Maybe route summarization / aggregation is extremely important because routers have small CPU / memory sizes relative to the full amount of state in the network. Is quality being engineered to an excessive point – beyond the point where further quality improvements would shift the demand curve?


Quality is an outcome that is often focused on in networks, either because of intuition (networks are only useful if they are working), impact on pricing / brand positioning, regulatory action, the amount of capacity budgeted relative to the service scale & scope, or it is one of the few ways a network designer can influence business outcomes / demand curve.

Whatever the business/network outcome and return desired, it should be transparent to network designers so their “how” choices are better informed.

Note: (1) quality can also be used in economics exchanges as a proxy for all the

capabilities of an offering. Quality is not used in that sense, in this article.

This site uses Akismet to reduce spam. Learn how your comment data is processed.