Forecasts are a substantial part of any “Business Plan”.
In the innovation process or in a ‘startup’, it is necessary at some time, to carry out and present a “Business Plan”. Necessary to ensure support for projects or to seek financing, whether internal or external.
The forecasts are a substantial part of any “Business Plan”. In doing so, new risks and opportunities can emerge more explicitly and help us to avoid further unforeseen explanations.
There is no standard business plan, but in many cases the same “Venture Capital” or “Business Angels” groups can provide us with a “proforma” model.
In general, they should be relatively concise, but they should contain financial estimates and therefore forecasts, with some contingency plans.
The forecasts have bad press. “In our case it’s different …”
Many of these plans have serious shortcomings especially in areas of marketing and finance. In both cases, estimates of expected income are poor, or lack of clarity in the premises on which they are based, lack of depth in the analysis of potential competitors and vulnerability in alternatives to possible risks.
The forecasts are made under much greater conditions of uncertainty than in normal operating forecasts and therefore it is convenient to establish margins and alternatives for the possible effects of the risks
It’s preferably to start from the correct perspective, based on dynamics that are necessary to try to understand and manage:
- What kind of innovation are we in? Incremental, radical or paradigm shift.
- Characteristics of each project: in its conception, development and launch.
- How innovation is applied: adoption and diffusion plans. The adoption of an innovation and its diffusion been intensively analysed and there are many models.
Objectives of the innovation process can be summarized in three, and are to be considered both when making forecasts and in the ‘business plan’:
- Launch the best possible solution to the market. That is, to ensure the quality of the solution and its alignment with the real needs of the customers.
- Do it in the shortest possible time (time-to-market)
- Using the minimum possible resources.
The achievement of these objectives will depend largely on the risks associated with the project and the way in which they are managed. In an environment of accelerated change, we must also contemplate that the acceleration of change produces an increase in uncertainty.
Forecasting methodologies in innovation projects
It is convenient to select between the methods, which ones to be used and fine-tune in the selection of the tools that we are going to use, involving people and prepared teams.
The difference between forecasting innovation projects and the operative forecasting we usually make for budgets, forces us to also adapt different methods or to use some of the usual ones with modifications.
We group the methodologies into 5 different types:
- Qualitative and exploratory.
- Time series and projections: Statistical methods
- Causal methods.
- Analogies and simulations.
1) Qualitative and exploratory
a) Projections of the sales area or collection of information closest to the customer or market.
There are of little use in innovation. The information is usually filtered in terms of existing products or services and skewed by current sales levels rather than future development potentials for new projects.
b) “Brainstorming” and other internal analyses.
Structured generation of ideas is used to find solutions to complex or specific problems or to identify potential new products or services. They can be organized in sessions conducted with a group of experts and all ideas are collected. The idea is to identify, without evaluating, the more opportunities and possible solutions to evaluate and classify them later.
c) “Design Thinking”
It’s a methodology also used in problem solving, especially suitable for investigating poorly defined problems, which focuses on the human being, focused on the possibilities and oriented to the hypothesis. It is a thinking model that combines empathy with users and immersion in the context of a problem, creativity in generating ideas and solutions and an experimental approach based on data to evaluate the quality of solutions.  & 
d) Groups or panels of ‘connoisseurs’ or ‘experts’
Debate or agreement process between experts.
Asking the “expert” generally provides the information for the forecast. It is assumed that the base of experience and education of experts is sufficient, in a particular field, to predict or forecast the vectors of expansion or evaluate the large lines of a market or segment evolution.
In reality, the three previous methods do not produce a forecast, but they do get ideas or solutions that can give ‘inputs’ to be applied to the forecasts.
e) Delphi method. Expert group in iteration process with moderator.
It is perhaps the most used and structured. And it is useful when there are high degrees of uncertainty or very long forecast horizons.
The methodology is well known and there is a large information about it.
The method seeks to eliminate the disadvantages of the work of experts ‘face to face’, due to mutual influences.
The quality of the forecasts depends largely on the selection of experts, the adequacy of their experience and the quality of the questions or questionnaires as well as the ‘feedback’ and iterations.
The consensus on the final results is usually important, but it is also important not to forget the radical dissensions that may have well-founded reasons and give rise to alternatives or to emerge from risks not contemplated.
It is a technique widely used in governments and public administrations.
f) Market or consumer surveys.
Data is collected through different methods, questionnaires, interviews, surveys, etc.
In consumer markets, the use in innovation is problematic because of the limitations of consumers in expressing their future needs. It is difficult for customers to articulate their future requirements in products or services they do not know well.
It is usually more used in industrial markets or “Business to Business”, where the client is usually in a better position to communicate the vision and future forecasts.
In these cases, innovations are often the result of collaborations with clients.
In addition, methodologies with non-direct but interpretive questions, such as “the Voice of Customer”, the application of QFD or the Kano model may be useful.
g) Intuitive Forecasting:
It is based on the personal or small group vision of a future.
It is more popular and frequent than it may seem. Many times, it is based on the intuition of the entrepreneur himself or the explicit vision of the head of a company. Case of Steve Jobs at Apple with some of its products, and many other cases, especially in those that today are large companies in the field of new technologies, as Microsoft, Facebook etc.
It has a lot to do with what we call “Push” in technological innovation.
h) Historical analogy of a similar product or service.
The product or service is linked to a series of historical data from a similar article. It can be adapted to regression models or special algorithms.
2) Time series and projections. Statistical methods
In all these methods, time series are based on historical data and trends, correlations, seasonality, patterns are identified, etc. Adjustments of different types are usually made to eliminate distortions, non-repetitive events, campaigns, etc. From there, projections are made using different techniques, including:
- Simple means. The projection is calculated from the simple means of series of periods.
- Weighted averages. Historical values with different weight.
- Exponential adjustment. Recent values have more value than old ones.
- Linear or other regression. We fit a simple line or line to a set of data.
- Exponential adjustment corrected by trend. From the adjustment, trend or seasonality factors are added. TAM, PMA.
- Special or complex algorithms.
The main difficulty in using above techniques in innovation projects lies in the lack of data from the past. However, regression analyses can be used by identifying the key factors that drive demand for a particular product or service and formulate estimates of future demand by establishing data on these underlying drivers.
For example, a regression analysis can provide us with the estimated demand for a new generation of cars (e.g. Electric) in case the demand drivers are, economic growth, prices relative to competing systems, the development of energy markets, the degree of awareness and adaptation of clean energy in consumers etc.
The advantage of these models is that they are based on the relation between cause and effect.
However, those will be of little use if we cannot know the future values of the main demand drivers, or underlying drivers, such as those indicated.
Special algorithms developed from simulations are also used especially in cases of technologies or product substitutes for new market niches, but they will be of little value in radical innovations or paradigm changes.
i) Application of statistical models.
A statistical model is based on series of observations of a phenomenon and outlines the pattern of association between the various factors (or variables) of the phenomenon that are of interest. The descriptive models used in the forecasts are often quantitative, but qualitative ones may also be used. Many events, as a descriptive phenomenon, are single-occasion events and, as such, are difficult to model.
Therefore, the application of a statistical model requires an analysis and effort of deep understanding, extending the process time until having results.
Statistical models that refer mainly to the analysis of historical data, select some attributes, such as sales, technical parameters, economic returns, etc., and are represented as a function of time. As it is generally assumed that progress is evolutionary and that technological progress is not random, it is possible to generate characteristic curves or patterns from the data and from these patterns, generate forecasts with varying degrees of certainty. However, the influence and impact of new or unforeseen factors should not be ruled out.
- “S” curves
- Trend extrapolation
- Technological substitutions. Etc.
j) Structural modelling
Structural modelling is an attempt to develop a mathematical or analytical model of a process of generating a new technology. As with the mathematical models of any process, the purpose of the model is to identify certain key elements, identify the functional aspects of those elements and express these functional aspects symbolically or mathematically. Structural models tend to be abstract and reductionist in their approach, because what is denied or what are non-essential functions are eliminated.
k) Trend extrapolation 
The extrapolation of trends is a method to link evolution of different products, services or successive technologies that perform similar functions to be able to see the elongation in the inflection of the “S” curve. (When the curve changes the slope and flattens). The extrapolated trend will eventually reach a limit and lose its validity as the trend approaches this limit.
3) Causal methods
A causal model is similar to a statistical model, since it also describes (through research) the forecast of development of a process. The difference is that it also provides the causative agents of the process to be predicted.
A causal model is usually based on statistical data related to a population. A limitation of the causal model is the need to assume that the causative effect will not vary according to time.
A famous example of a great causal model was prepared by the Club of Rome and published in 1972, (Limits to Growth). It consists of dozens of variables, which include world population, birth rate, industrial and agricultural production, non-renewable resources and pollution.
l) Regression with multiple variables and correlation.
It considers correlation of several variables and adjusts. Some methods seen in regression analysis in time series are also causal.
m) Consumption / Product Models.
The forecast is made for a relationship with customers and trends.
n) Advanced indicators.
Use of statistical series of other products that have been used or anticipated. We have seen something similar in extrapolating trends.
4) Analogies and simulations
o) Scenario Developments.
They are consistent descriptions of possible alternative futures, based on premises and interpretations of the driving forces of change.
The inputs are quantitative data of what are considered relevant driving forces, such as sales series of certain sectors, GDP of specific areas, demographic or technological series, public or private investments in a specific field, etc. May also consider qualitative data such as political, social, environmental factors etc.
The development of scenarios does not strictly produce forecasts but considers the uncertainties and current trends in the specific area and assumes that the future may vary between incremental or revolutionary changes, allowing forecasts to be made in a range of values between limits, up and low.
It is important to highlight that it is a good method to incorporate critical events that can happen and mark a change in the initial premises, which can be used to correct the forecasts.
You can apply many forecasting techniques, including simulation software, use of ‘big data’ and artificial intelligence, identification of critical indicators, regression techniques etc.
It is especially used in long-term planning of sectors characterized by heavy investments, long delivery times or environmental uncertainty, such as energy, aerospace and telecommunications sectors.
p) Combination of methods and models with practice.
Simulations ‘what if’, with a combination of methods. It can lead to generate new algorithms.
It is often used in radical innovation projects and can use a combination of methods like segmentation, prototyping, panels of experts, and trials in small or pilot markets.
For non-radical or incremental innovation projects can use a combination of collaboration with customers, extrapolation of trends of similar products or services and segmentation.
q) Analogy methods
This method uses analogies between the product or service to be predicted and some known time series, physical or biological process. To the extent that the analogy is valid (and all analogies become invalid at some point), the initial process can be used to forecast future developments of a technology, a product or service.
It is necessary to use the analogy method in detail, examining the situation of the model and the situation to be foreseen in considerable detail to determine to what extent the analogy is valid.
An example of this approach is one type of lighting technology replaced by other. In this case the result is a continuity of the “S” curves.
Forecasting by consensus is usually the end of the practice of some processes with different methodologies and results.
It is a collaborative method. An exchange of results according to different methods is established and agreed. It is a collective teamwork with an open attitude towards anyone who has something interesting to contribute.
It’s based on:
- Integration. Need to observe from a global perspective, not only from our point of view.
- Interpretation. We work on the construction of assumptions to identify premises and risks and assess their impact.
- Each case is a new exploration, experiences are accumulated and serve to validate future forecasts with greater accuracy.
- Iteration when consensus is not reached. We can iterate from early stages to find out what causes the lack of consensus. In some cases, the differences should be recorded.
The tendency to bias can be positively modified if the work is carried out horizontally, that is, based on cross-functional teams, regardless of hierarchies.
Altogether, the panels for analysis and evaluation of ideas formed by internal and external people, the practice in the realization of projects, the analysis and evaluation of the stages and doors, and the iteration practice, must significantly improve the accuracy in the forecasts.
Anyway, a system of measures for forecasts accuracy must be created, to improve.
The risks are coming basically from errors in the premises and the failures in the methodology itself.
The most common types of risks are detailed below, although it is necessary to clarify that this is a simplification, and that in specific cases other types of risk may appear. Should be also considered that the different risks are not independent of each other.
- Offer risk: associated with whether a product or service will be technically viable and will fulfil the expected function. In technological companies, we talk about technology risk, although this type of risk will be present, to a greater or lesser extent, in any innovation project.
- Market risk: associated with whether a solution meets the needs of a specific customer segment, and whether it is well positioned with respect to the competition. The idea is to avoid launching the wrong solution to the market, regardless of whether it is technically good or not.
- Financial risk: this category of risk may have two parts, one related to the solution, and another related to the process.
- Risk associated whether the solution will achieve the expected sales and / or benefits and therefore will create enough value for the company to be considered a success.
- Risk associated whether the innovation process can be carried out with the planned budget.
- Time risk:
- Associated with whether the company will be able to launch the product or service within the established time frame, and the consequences that not doing so may have.
- Associated with whether the launch will arrive in a timely manner, not too early, not too late, for expectations or needs in the planned market.
Decalogue of forecasts
- A forecast will never be accurate
- Forecasts are necessary for any innovation project or business plan. To make decisions and obtain financial support
- The responsibility for forecasts resides lately in the Innovation team or committee
- A forecast is inaccurate because the premises were incorrect
- Forecasting needs to apply ‘strategic intelligence’ in the company or organization
- More accurate and reliable forecasts are obtained with greater experience and knowledge and with the appropriate methods.
- The forecasts are more accurate at the aggregate level than in detail.
- The experience of the organization in the operational forecasts with structured and perfected methods, generate knowledge and experience for innovation projects.
- It is essential to consider and record the premises and, if possible, establish relationships with the high and low forecasts.
- Measuring the inaccuracy of forecasts is key to improving the process
 Jeanne Liedtka
 Michael S. Slocum and Catherine O. Lundberg. TRIZ Journal. Tools to Forecast Technology Innovations