If we launch it will they come?

How can you predict whether customers will flock to a new communications service, or avoid it like a bad smell? E&T offers some guidance.

Forecasting is important to rapidly changing sectors such as communications, where new technologies and services emerge almost daily. Companies in such sectors need forecasts to build business cases and make investment decisions. Stakeholders need them to understand how their businesses could change. And regulators need them to maintain appropriate oversight. Yet, despite the discipline of forecasting being hundreds of years old, forecasts are still often wildly out of kilter with reality. This article examines why this is so, and the limitations of forecasting in rapidly changing sectors.

Examples of bad forecasting abound. In the 1980s the telecommunications industry confidently forecast that up to 20 per cent of the population would eventually use cellular telephony: there are now more mobile subscriptions than people in the UK. The uptake of services such as texting was completely unexpected, because it was assumed that there would be no market. Equally, there were forecasts of dramatic growth for wireless data, location-based services, picture messaging and video calling, services that have largely failed to take off.

Forecasters often predict a service will grow dramatically, presenting graphs of exponential growth and the classic 'hockey stick' curve. As the forecast growth fails to materialise they gently shift their curves east along the time axis, pushing out the point at which sudden growth is expected, rather than admitting they were wrong. And yet, occasionally, services that have lain dormant for years will suddenly take off, sometimes for reasons that are unclear.

Forecasting the uptake of communications services means understanding two factors: what is possible from a technical and economic viewpoint; and what users will adopt once it is offered. The mistake many forecasters make is to assume that it is the first that is most difficult. Not so.

Predicting technological change

There's a checklist of things to consider when predicting technological change.

First, look at the de facto 'laws'. Most technological changes are easy to forecast. Many conform to 'laws' such as Moore's Law, which has allowed us to accurately predict the ability of semiconductor devices for four decades.

Second, look at developing standards. Most key technologies emerge through standards bodies such as the IEEE. These take at least three to five years to develop and publish. It then takes a further one or two years for compliant equipment to be delivered to market. Such laws provide a good guide to what is coming in the next decade.

Third, consider technical constraints. There are many of these, such as battery performance and physical limits on capacity, which tend to slow down the adoption of new technology.

Finally, understand the economics. Once the technology is well understood, developing the economics or service is relatively easy. Prices for equipment can be estimated from previous generations and by using well-known pricing trends. Overheads and manpower costs are easily charted. Cost overruns are always possible, but the telecommunications industry has usually managed to predict project costs well and stay reasonably within budget.

I have been using these techniques to make predictions about the advent of new technologies in wireless communications for almost a decade now. So far, the predictions I made about eight years ago have proved to be almost completely accurate, or a little optimistic, about the speed at which new technologies can be deployed.

Predicting customer behaviour

The second part of forecasting the uptake of a new communications service is to understand customer behaviour. We know a lot about the way that people adopt new services, products and technologies. The concept of a small body of early adopters trying out new concepts is widely accepted, as is the idea that such pioneers will be followed by an early majority, a late majority and then by the laggards. There is also good understanding of how people will react to prices. Certain price points have a psychological importance: selling products for less than £100, for example, seems to change the rate at which they are adopted.

The problems in forecasting come because some of the key mechanisms that promote adoption have 'tipping points', at which they move from slow to rapid growth. Consider a virus that has a short lifetime and a certain probability of infecting others. If the population density of hosts is low, the virus may not come into contact with enough new hosts within its lifetime to propagate, and will swiftly die out. If the population density of hosts is high, one infected host can infect many others, and the virus will spread swiftly. If you model such behaviour you find that the virus dies out at a particular host density, but that a very slight increase in that density enables it to spread rapidly. That is the 'tipping point', where a small change in the input parameter has a dramatic impact on the outcome. Such non-linear characteristics are why it is so difficult to make accurate forecasts about mechanisms that have tipping points.

The uptake of communications services often relies on tipping-point phenomena that happen among the early adopters, and on the messages that they pass on to the wider community. If most early adopters like a new service then this will encourage the early majority to adopt it. However, it only takes a few early adopters to cast doubt on a new service, perhaps because it is difficult to use or unreliable, for the risk-adverse majority to steer clear.

Even if you believe that there might be a tipping point in the uptake of your new service, understanding what it is and when it might occur can be very difficult. Such tipping points tend to occur at a lower threshold for expensive items or for those that are dramatically different from previous offerings, and at a higher threshold for items that are low cost or which are a simple increment of an existing service.

Brand and reputation can also reduce the impact of a negative tipping point - the Apple brand was strong enough to overcome a number of launch problems with the iPhone. All of this makes service prediction near impossible and partly explains why services can languish for many years and then take off for no clear reason - a series of small improvements may have pushed the service past its tipping point for wide uptake.

If we could be sure that all the early adopters of a new service would be delighted with it then forecasting would be much simpler. There would be no concern about tipping points and we could use standard curves and time-scales to predict the diffusion of new services throughout the industry. Those offering new services would love such predictability and do everything they can to delight their early adopters. So can we forecast how successful they are likely to be and the conditions under which they might succeed or fail? Well, there are some things we can check off.

First, look at the number of players involved. Prediction is relatively simple where there is only one manufacturer or operator. For example, it was easy to predict that Apple would succeed in designing a desirable iPhone, given their track record and the fact that no other organisations were involved.

Second, look at the ecosystem. Most new services rely on a complex ecosystem that includes manufacturers, operators and other service providers. For example, location-based service providers need easy-to-use GPS location systems in the handset; operators that provide a framework for their services; mapping companies to produce appropriate offerings; and search companies to provide location-enabled search. Many of the relationships will rely on trust and friendship, making them unpredictable. This makes it difficult to create meaningful models of such ecosystems, although they may provide pointers about the kind of issues to which service providers should pay particular attention.

Finally, consider the novelty and cost of the product. Factoring in the novelty of a service and whether it represents a large or small cost to consumers can help developers understand if there is likely to be a tipping point in its uptake. It may be unrealistic to expect to understand where such a tipping point might occur, but it might help in a coarse assessment of the probability that the uptake of a new service may be delayed.

Companies that want to launch new services need to try forecasting their likely uptake in order to raise the money and plan for the roll-out. They can make good forecasts about technical issues, and useful but weaker forecasts about consumer behaviour. But it is worth remembering that such forecasts are fraught with difficulties, and that even the most experienced can get it wrong.

After all, at the beginning of the decade more than £20bn was staked on 3G licences on the premise that we'd all rapidly adopt mobile data, surf the mobile Internet and use videocalling. Turns out that was a leap of faith too far!

William Webb is head of R&D at Ofcom

For more, see 'The Future of Wireless Communications', Artech House, 2001 and 'Wireless Communications: The Future', Wiley, 2007.

Recent articles

Info Message

Our sites use cookies to support some functionality, and to collect anonymous user data.

Learn more about IET cookies and how to control them