Predicting the upcoming start of large-scale diffusion of radically new high-tech innovations

Abstract

Knowing the time point of large-scale diffusion of a
radically new high-tech innovation is a highly relevant topic. Companies,
researchers, and government institutions can plan their research and
development efforts, production, as well as marketing plans according to the
predicted time point of large-scale diffusion. The research is based upon the
assumption that specific indicators can predict the start of large-scale
diffusion. The scientific field of forecasting the start of large-scale
diffusion is relatively new. Therefore, an explorative methodology was required
for this research. During the explorative process, it was ensured that
indicators reflect on the holistic environment of an innovation by minding the
so-called data collection cube.  A data
selection funnel was created, narrowing scientific branches down to a list of
indicators in three steps. Each of these steps has its own criteria designed
to: • Select scientific branches with the highest potential to find results in
the literature reviews • Derive indicators that can observe the diffusion •
Select indicators that can predict the start of large-scale diffusion   The last step of the data selection funnel,
selecting the indicators which can actually predict, was carried out with the
support of three researchers. Eight criteria were used to select the most potential
indicators: (i) Prediction, (ii) Timeliness of prediction, (iii) Availability
of data, (iv) Cost of data, (v) Quantifiable& Objectivity, (vi) Empirical
proof, (vii) Generalizability, (viii)Simplicity. The researchers were asked to
evaluate the indicators according to the criteria (i) and (vi) as part of the
scientific quality gate selecting the most potential indicators. After the
indicators have been evaluated, a sensitivity analysis has been performed to
improve the robustness of the selection mechanisms and to rule out an arbitrary
selection of the indicators. Out of 50 indicators found in the literature or
derived from the literature, 38 indicators were selected according to the
selection mechanism. These 38 indicators have been split into two sets of
judgemental and non-judgemental indicators to prepare the design of the
forecasting approach.   The forecasting
approach aims to guide a user towards the correct forecasting technique given
an innovation and situation. Five forecasting techniques were found to
befitting the: (i) assumptions-based modelling, (ii) Delphi method,
(iii)analogous forecasting, (iv) time series & regression models, and
(v)artificial neural networks. However, each of the five forecasting techniques
have disadvantages that can be overcome by one of the other methods. Hence, the
forecasting approach has two stages. First, the user is guided towards the
primary method and subsequently towards an additional method overcoming the
disadvantages of the first method and improving the overall reliability of the
forecast. For each method, a set of indicators is recommended.   Once the forecasting approach has been
developed, the completeness of indicators has been checked by using Ortt &
Kamp’s 14 factors influencing the pre-diffusion phase. Additionally, four
validation interviews applying the research on green hydrogen have been
performed to let external actors reflect. These validation interviews formed
the practical quality gate forging a bridge to the earlier mentioned scientific
quality gate.   

Knowing
the time point of large-scale diffusion of a radically new high-tech innovation
is a highly relevant topic. Companies, researchers, and government institutions
can plan their research and development efforts, production, as well as
marketing plans according to the predicted time point of large-scale diffusion.

The research is based upon the assumption that specific indicators can predict
the start of large-scale diffusion. The scientific field of forecasting the
start of large-scale diffusion is relatively new. Therefore, an explorative
methodology was required for this research. During the explorative process, it
was ensured that indicators reflect on the holistic environment of an
innovation by minding the so-called data collection cube.



A data selection funnel was created, narrowing scientific branches down to a
list of indicators in three steps. Each of these steps has its own criteria
designed to:

• Select scientific branches with the highest potential to find results in the
literature reviews

• Derive indicators that can observe the diffusion

• Select indicators that can predict the start of large-scale diffusion   The
last step of the data selection funnel, selecting the indicators which can
actually predict, was carried out with the support of three researchers. Eight
criteria were used to select the most potential indicators: (i) Prediction,
(ii) Timeliness of prediction, (iii) Availability of data, (iv) Cost of data,
(v) Quantifiable& Objectivity, (vi) Empirical proof, (vii)
Generalizability, (viii)Simplicity. The researchers were asked to evaluate the
indicators according to the criteria (i) and (vi) as part of the scientific
quality gate selecting the most potential indicators. After the indicators have
been evaluated, a sensitivity analysis has been performed to improve the
robustness of the selection mechanisms and to rule out an arbitrary selection
of the indicators. Out of 50 indicators found in the literature or derived from
the literature, 38 indicators were selected according to the selection
mechanism. These 38 indicators have been split into two sets of judgemental and
non-judgemental indicators to prepare the design of the forecasting approach.  
The forecasting approach aims to guide a user towards the correct forecasting
technique given an innovation and situation. Five forecasting techniques were
found to befitting the: (i) assumptions-based modelling, (ii) Delphi method,
(iii)analogous forecasting, (iv) time series & regression models, and
(v)artificial neural networks. However, each of the five forecasting techniques
have disadvantages that can be overcome by one of the other methods. Hence, the
forecasting approach has two stages. First, the user is guided towards the
primary method and subsequently towards an additional method overcoming the
disadvantages of the first method and improving the overall reliability of the
forecast. For each method, a set of indicators is recommended.   Once the
forecasting approach has been developed, the completeness of indicators has
been checked by using Ortt & Kamp’s 14 factors influencing the
pre-diffusion phase. Additionally, four validation interviews applying the
research on green hydrogen have been performed to let external actors reflect.
These validation interviews formed the practical quality gate forging a bridge
to the earlier mentioned scientific quality gate.