How to minimise the 3G risk: By accurately forecasting data usage, it will help the 'pure' 3G players to roll out their networks cost effectively - 3G Networks

Telecommunications International, Feb, 2002 by Damien Rouchouse, Nicolas Faesch

The next question is: what is the traffic for one minute of multimedia? Considering the availability of the bearers at launch, we assume that two per cent of multimedia traffic will be handled by circuit-switched communications for real-time applications, and 98 per cent will be packet-switched traffic for other applications.

For circuit-switched data traffic - with a 64Kbps bearer - it means that one minute will represent 3.84Mb of data. For packet-switched data traffic, we need to calculate an average data rate. The technical calculations done by our client concluded that an estimated average throughput of 10Kbps (a mix of 64Kbps and 144Kbps, plus overhead and signalling traffic) can be achieved, which represent 6.60Mb per minute

Given that we assume a distribution of two per cent for circuit-switched data and 98 per cent for packet-switched data, one minute of multimedia therefore r represents 6.54Mb.

Finally, once we have done the same exercise for each category of service and for each customer segment, we come up to the consolidated data traffic. This overall traffic will be therefore split by radio cell and network nodes (MSC for circuit-switched communication and GGSN-SGSN for packet-switched communications), using Sofrecom's geomarketing methodology based on the spatial analysis of market research results, as well as socio-demographics databases with a desktop GIS.

Conclusions

By using this 'bottom-up' traffic modelling approach, it was possible to extrapolate traffic usage patterns. For circuit-switched data, used for real-time interactive and multimedia applications, the traffic (evaluated in Erlangs) will represent less than five per cent of the overall circuit traffic for the next five years or so, the other 95 per cent still being voice.

For packet data, the traffic is estimated in Mbps in the peak hour and remains, of course, very sensitive information for the operator. However, we can communicate figures for non real-time applications and it is interesting to observe the distribution of the traffic (as shown in (Figure 1)

As Figure 1 shows, internet browsing is expected to represent nearly 50 per cent of total traffic. For this packet-switched data traffic, no more than 20 per cent will be supported on a very high data rate (384Kbps), but that figure is expected to double by 2011.

Such a traffic dimensioning methodology should benefit the service provider (the right capacity delivered at the right place) as well as the future 30 customer (bandwidth available when needed).

Pricing issues

The limit of this method is a pricing issue. Put simply, when a projected data volume is derived, it has to be based on certain pricing assumptions.

If you pay [euro]18 for 60 minutes of voice over a 20 network, a minute works out at [euro]0.3. This is the way the mass market has been educated. Will 3G services change this? Unlikely.

Given that teenagers are ready to pay [euro]16 per month for future 3G services -- and good market research has been able to say exactly what they want from that amount -- then the price for these 30 services can be determined in a defined framework. As we have already seen, our teenagers ore willing to pay [euro]1.6 per month for music services, so operators should work towards producing a pricing package which fits in with that level of willingness to spend.