STEM newsletter

Evaluating online conferencing

30 October 2007

The demand and supply concept central to STEM is applicable to many situations outside conventional telecoms. Online conferencing is one of the business topics featured in the recent interactive modelling exercise at the STEM User Group Meeting in September 2007, and this nicely illustrates some alternative dimensioning concepts.

We considered the likely market and technical design for a desktop sharing service, the necessary network infrastructure, and the associated revenues and costs. We also made an allowance for the additional bandwidth required for timely client-software downloads. Based on certain assumed staff performance ratios, we were able to quickly calculate business results that were directly comparable to the published results of the established player WebEx.

Online conferencing customers and services

The principal service we modelled was an Internet-based desktop sharing solution that would enable an online meeting in which the desktop graphics and mouse position of one user can be broadcast to multiple users connected to an online meeting ‘call’. We assumed two alternative revenue models: either a monthly subscription of EUR40 (declining at 10% per annum) or pay-as-you-go (PAYG) at EUR15 per hour (declining at 5% per annum).

This basic service would allow up to ten participants, with a premium for having an ‘unlimited’ audience. We also considered add-on services such as telephone conferencing, end-to-end security, guaranteed QoS and embedded video (e.g. medical imaging).

The business appeal of such a service is its general utility, and we thought about only a small subset of the potential customer types:

  • small and large companies
  • financial institutions requiring end-to-end security
  • state sector organisations, such as tax offices and universities
  • medical facilities requiring remote imaging
  • software vendors (such as Implied Logic!), for online sales and support functions.

Estimating the market for online conferencing

The market was then estimated according to the number of entities per organisation type, the proportion of such entities that would be interested in the service (50%), and the number of relevant users per organisation type. Overall we assumed a 33% market share, and that subscription users would out-number PAYG users by a factor of 2:1. Finally we assumed market penetration of 2% in 2008 for subscription users rising to 20% by 2010, and 1% for PAYG users, growing to 8%.

Network infrastructure and drivers

The main network elements are Internet peering to a point of presence (POP), and then a managed wide area network (WAN) providing connectivity between these POPs and one or more server farms hosting the essential servers. The model costs the following elements:

  • POP peering, equipment and site costs
  • core network connectivity (assumed to be leased)
  • Web, meeting and account servers
  • staff for marketing, sales, technical support, software development, legal and management functions
  • SG&A overheads at 20% of total revenue.

Online conferencing architecture

In order to understand the business dynamics, we need to understand the separate cost drivers:

  • users = (subscription + PAYG) aggregated over organisation types
  • traffic volume (per user) on the basis of online meeting hours
  • busy-hour traffic as online meeting Erlangs
  • bandwidth = 64kbit/s per concurrent meeting
  • core software download = 10MB once per annum per user
  • update downloads = 1MB once per meeting attendee.

We assume that the significant desktop-sharing traffic on the WAN is just one stream per meeting, and that the meeting server is dimensioned based on the number of concurrent meetings (assuming screen update is more significant than user management). In contrast, we assume that the POP bandwidth required is in proportion to the number of concurrent meeting attendees (based on an average of three per meeting, including the host), since attendees will generally be at different POPs.

For the download traffic, we don’t just work from the total download volume in a year, as this would not guarantee an acceptable download rate in busy periods. Instead, we assume an average download rate of 256kbit/s and then convert the respective 10MB and 1MB volumes to required download minutes. This allows us to estimate busy-hour download Erlangs, from which we can calculate suitable overhead capacity for a 0.1% grade of service and then multiply by 256kbit/s per ‘download channel’ to ascertain the required bandwidth.

We only dimension for download traffic on the external network (at the POPs), as we assume that software download images are cached to access servers at POPs over the WAN during slack periods.

Network infrastructure drivers

The detailed assumptions for cost breakpoints and staff salaries and drivers are listed as two tables in an annex to this article.

Compelling results

We challenged the participants in our interactive exercise to guess in advance what would be a realistic price to pay for such a business(EUR): 5m? 50m? 500m?

Users, ARPU, revenue and charges

The tariffs are realistic, but may be forced lower by competition. In addition, costs will be driven up directly if more companies start using the tool for mass, routine collaboration, such as might be convenient for the dominant SME segment.

Adding further POPs would also drive up costs, but this is unlikely to be necessary, and actually reduces efficiency as the peak-average variance per POP would be increased.

Operating profit, profitability and NPV

On a user base of two million we calculate annual revenues of around EUR300 million, profitability in excess of 40% over a ten-year period, and an almost immediate payback.


Consistent, premium revenues and a non-capital-intensive business structure make for healthy profits with no significant cashflow or borrowing issues. However, there is already competition in this space and - with results like these - one can expect additional players to join the market, which will inevitably drive tariffs down.
The addition of a series of well designed and reliably implemented premium services may be required to differentiate the offering, secure customer loyalty and maintain market share. Nevertheless the immediate positive cashflows and only gradual decline of margin will explain why Cisco Systems bought web conferencing company, WebEx, for USD3.2 billion in March 2007!

Annex: cost breakpoints and staff salaries and recruitment drivers

Group Unit costs (EUR)
POP POP 9k p.a. per 10 Mbit/s peering
20k per 100 Mbit/s server
10k p.a. co-location cost
Managed WAN 45k p.a. per 10Mbit/s
5k p.a. per POP
Server farm Meeting server: 5k per 100 concurrent meetings
Account server: 5k per 1m total users
Web server: 5k × (10 fixed + 1 per 100k users)

Cost breakpoints

Role Salary cost (EUR p.a.*)
Marketing 45k × (2 fixed + 1 per 37 500 accounts (sales / 5))
Sales 75k × (10 fixed + 1 per 7500 accounts)
assuming 4 accounts per hour × 7.5 hours × 250 days
Technical support 45k per 500 concurrent meetings
assuming 20 per hour × 1 in 25 meetings
Development 60k × 10 fixed
Legal 90k × (1 fixed + 1 per 1m accounts)
Management 120k per 10 regular staff

Staff salaries and recruitment drivers

 * plus 80% office and social security overhead

© Implied Logic Limited