Targeting promising markets of an innovative technology: proposition of the

Design by Usage-based Experimentation (DUE) methodology

Abstract: R&D investment in innovative startups is extensively focused on prototyping

and testing. Due to significant uncertainties, innovative startup investments are
motivated by the pressing need for building and repeatedly testing a proof of concept,
without necessarily delineating the innovation problem addressed or the targeted
market. Following this mindset, the startup leads to random experiments that may
cause costly pivots, and even failure. Thus, the relevance of such experiments
becomes questionable in the context of limited time and resources. Looking at the
interdisciplinary literature, it appears that the existing methods fail at providing
startups with practical tools that allow experiments that address relevant gaps in the
market. To confront this issue, this paper proposes a new usage-driven methodology
called Design by Usage-based Experimentation (DUE). DUE methodology provides a
structured process, along with a set of tools, to identify experiments that are worthy of
being conducted, insofar as they focus on gaps in the market, characterized by value
buckets (i.e., important problems which occur in frequent usage scenarios, for which
the existing solutions are not efficient). The main output of DUE is a set of
experimentation opportunities that mature the startups’ core technology; opportunities
computed by matrices, linking the promising market to the main properties of the
technology. Inclusively, this paper also presents the results of the DUE application to
the case of a startup technology. We conclude that this methodology generates useful
insights and may help startups in structuring and organizing their experiments.

Keywords: R&D activity, experimentation, design, usage, market, property of the

technology.

1. Introduction

“Most startups fail because they waste too much time and money building the wrong
product before realizing too late [and after trials and errors] what the right product
should have been,” emphasizes Professor Thomas R. Eisenmann (Nobel, 2011).
Although there is no clear definition of a startup, many researchers and practitioners
consider it as equivalent of continual experimentation (Ries, 2011; Blank, 2013b). The
aim of any innovative startup is to transform the initial idea into a successful product.
To do so, startup companies build several Proofs of Concept (PoC) and then test (or
experiment)1 them repeatedly to verify them as the best fit. However, these loop-based

1 “Test” and “experiment” are used interchangeably in this paper. Despite potential differences, in
general they have more similarities. As indicated by (Kass, 2008), terminology confusion suggests a
close connection between test and experiment; a test is the assessment of presence, quality, or
genuineness of something. An experiment explores the effects of manipulating a variable.

1
experiments, currently practiced by most startup companies, need improvement, as
they are costly and time-consuming and may lead to several pivots and even failure.

This situation is further observed in the context of technology startups that are
developing a core technology, and adopting testing and business strategies for
launching new product on the market. Indeed, these startups require an important
number of resource-intensive experiments to mature their technology. They also need
to effectively align both technology development and market penetration strategies to
reduce the number of pivots and use of resources. However, in the given situation
where the technology is not a good fit for a given unsolved or poorly solved market
problem, the number of pivots, and thus waste, in resources might be quite significant.
Therefore, the effectiveness of experiments becomes a crucial challenge in the
context of limited time and resources for startups.

Considering that 73% of startups are technology-driven, and that this number has been
increased by more than 47% during the last decade (Wu and Atkinson, 2017), one may
consider that challenges related to wisely investing in experimentation effect a large
portion of the entrepreneurial ecosystem. This includes entrepreneurs, but also other
players, such as investors as well as public and private organizations supporting
innovative startups.

A large body of literature in innovation management, technology management and

engineering design, already deals with the question of testing and prototyping an
innovation. In an entrepreneurial context, several methods and tools, such as the Lean
Startup (Ries, 2011), provide experimentation guidelines for startups. The common
point of these methods and tools is that they put a strong emphasis on loop-based
testing an innovative product or service in order to improve its technical and business
performance.

Nevertheless, review of the interdisciplinary literature shows that existing methods and
tools provide no satisfactory guidelines for identifying which market and technology
property are relevant to be targeted through experiments. Apart from this, the existing
approaches concentrate on the testing of a solution without necessarily backtracking
the end-users’ problem.

Therefore, the question addressed in this paper is: how to identify the most promising
markets and technology properties that a technology-oriented innovative startup
should target through experimentation, in the case of limited resources and time?

To address this question, this paper puts forward a methodology aimed at specifying
effective experiments, through exploration of relevant usages and promising markets.
This methodology is called Usage-based Experimentation (DUE) and follows a set-

2
based thinking approach (see for instance (Yannou, Yvars, et al., 2013; Sobek et al.,
1999; Kennedy, 2008)). Unlike the trial-and-error approach, the set-based thinking
approach consists of reducing uncertainties to progressively converge into a
satisfactory solution.

Usage (as in usage situation, usage scenario, and usage context) is core to DUE.
However, this concept is broad and has no single definition. A usage scenario is an
archetypal story lived by a representative group of users, where one may identify a
notable problem, whereas a usage context enables the capture of personal and
environmental factors and lifestyles (He et al., 2012). In this paper, by “usage” we mean
a usage scenario where significant problems are the main focus of the DUE method.

DUE is a systematic diagnostic approach that starts with evaluating the potential of an
existing technology in different markets, as well as identifying the experimental areas
worth conducting. These areas represent a gap between the current maturity of the
technology and its expected maturity in the most promising markets. In DUE,
promising markets are characterized by value buckets (i.e., important problems
occurring in frequent usage scenarios in which the existing solutions are not efficient).
The concept of value bucket is borrowed from the Radical Innovation Design® (RID)
methodology (Yannou, Jankovic, et al., 2013; Lamé et al., 2018; Yannou et al., 2018).
This also follows set-based thinking. Moreover, RID is compliant with the need seeker
innovation strategy (Jaruzelski et al., 2016) that encourages the development of new
products and services based on superior end-user understanding.

The main contribution of this article is to provide a set of practical tools for
practitioners. These practical tools may help them identify a set of promising markets,
as well as key technology properties that require maturing through experimentation.
This is a new result, insofar as a structuring process and quantitative evidences are
provided to specify and prioritize experiments. Aside from this, academics may find
these conclusions useful, due to the structured process of experimentation proposed
by the DUE methodology.

The organization of this paper is the following: in section 2, the background of DUE in
engineering design literature, and technology and innovation management fields, are
presented. In Section 3, the DUE methodology and its components are outlined. We
then focus in Section 4 on the application of DUE in the case of startups that have
patented a shape-changing technology. Finally, Sections 5 and 6 summarize and
discuss the results in light of the reviewed literature, and interpret the significance of
the DUE methodology.

3
2. Background literature on design and experimentation
approaches

In this section, we seek to understand what the design and experimentation

approaches to new product development are. Subsection 2.1 reviews design and
creativity approaches, and subsection 2.2 focuses more specifically on
experimentation processes and their components.

Design and creativity approaches

There is a plethora of design and creativity methods and tools. In the following, we will
detail some of them relevant to the context of this research. They are mainly derived
from the fields of engineering design and technology management.

The C-K theory (Hatchuel and Weil, 2009) builds on conventional creativity and design
theories and its design process is considered a series of back and forth between
knowledge (K) and concept (C) spaces. In practice, C-K theory is a qualitative problem-
solving mechanic that helps designers understand problems and find creative
solutions. However, there is no differentiation between concepts or knowledge
regarding a design problem or solution. In addition, C-K theory does not specify the
notions of usage, innovativeness or value indicators, market, and existing
technologies, as well as the cost and time of development.

Another qualitative design approach is the Double Diamond model for product design
that has been created by the UK Design Council in 2005 (DesignCouncil, 2005). This
model contains two macro-stages and four steps: Discover, Define, Develop and
Deliver. Although this approach is very useful and structured, it does not necessarily
focus on the design and test of a technology, nor does it provide practical tools.

Consistent with the divergent macro-stages of Double Diamond model, the first step
of Product Design and Development model of Ulrich and Eppinger (Eppinger and
Ulrich, 1995) begins with discovering customer needs and screening opportunities in
the market. However, the identification of customer needs is performed through
market studies, rather than systematically exploring usage situations and users’
problems which subsequently occur in these situations.

TRIZ is another approach to inventive problem solving. It is a useful and widely used
theory in the context of technology companies. However, it does not claim to be a
holistic design and testing approach insofar as the notions of usage, quantitative
indicators related to the market and cost evaluation are missing. Unlike the previous
analyzed approaches, the power of TRIZ lies in an algorithm called ARIZ to isolate the

4
problem, the Contradiction Matrix, Laws of Technical System Evolution, substance-
field analysis as well as the concept of ideality of a system. Scientific publications on
TRIZ and its extensions on technology development are abundant (see for instance
(Cascini et al., 2011; Cascini et al., 2009).

Other popular approaches widely used in organizations are the Quality Function
Deployment (QFD) (Akao et al., 1990) and the matrix flow approach of Nam Suh’s
axiomatic design model (Suh, 2001). They serve as a decision-making tool that
transforms the Voice Of the Customer (VOC) into engineering characteristics for a
product. Even though a large amount of data on customer needs are stored and used
in QFD matrices, the collected data does not systematically allow exploration of users’
difficulties and usage situations.

The process of transforming VOC data into design requirements in QFD calls for other
approaches, such as the Six Sigma methodology. Six Sigma combines with the lean
process to become Lean Six Sigma (LSS) (George, 2002). LSS reduces waste while
improving the quality of the product service system. It begins with the process of
defining the VOC. The benefits of using LSS in companies are significant.
Nevertheless, listening and specifying the VOC requires that customers are able to
express their problems, needs, and expectations. This is not always the case and is
not necessarily linked to efficiency of the existing technologies on the market.

To characterize real problems, beyond immediate customer expectations, the Radical

Innovation Design® (RID) methodology provides systematic detection of value
buckets, i.e., combinations of important problems occurring in frequent usage
scenarios, for which existing solutions are neither useful nor efficient (Yannou et al.,
2016). RID is a usage and pain-driven method that enables the design of inventories
for existing solutions, usage situations (or scenarios) and user pain points (or
problems). However, RID methodology is not conceived to screen several usages, or
explore a technology base instead of an initial idea. Aside from this, time and cost of
technology experiments are not systematically incorporated into RID’s tools and
process.

In summary, the reviewed methods and tools either do not specify usage situations,
markets and existing technological solutions, or they fail at specifying and designing
relevant experiments.

Methods and tools for experimentation

An experiment is defined as a set of organized measurements carried out under

specific and controlled conditions. Experiments include objects of discovery,
verification, illustration of a theory, hypothesis, or the relationship. The

5
experimentation approach can be analyzed through four main components: processes,
decision-making tools, test objects, and hypotheses.

2.2.1. Experimentation processes

More often, experimentation processes are iterative, or “non-linear,” due to the

existence of back loops. In Design Thinking (DT), results of prototype testing are re-
used as input for the next round of development (Johansson-Sköldberg et al., 2013).
(Brown, 2009) emphasizes: it is through the process of repeated creation, testing and
learning that entrepreneurs can improve their initial ideas.

In DT, prototyping and testing are combined modes and consist of iteratively
developing several solutions (or prototypes) to collect several users’ feedback.

Comparable to the iterative process of DT, the Lean Startup (TLS) process makes a
particular focus on experimentation. TLS is based on Steve Blank’s Customer
Development process (Blank, 2013a) and complies with other approaches, such as
Lean Manufacturing and agile software development. The process of TLS intends to
reduce uncertainties by continuously running the build-measure-learn loops. The Lean
Startup offers an alternative to conventional business planning methods (Ries, 2011).

Inspired by the build-measure-lean loops, the Value Proposition Design (VPD)

(Osterwalder et al., 2015) puts forward a testing process, which consists of extracting
relevant hypotheses from the Business Model Canvas, Value Map and Customer
Segment (see Fig. 1).

Fig. 1 Testing process of the Value proposition Design (Osterwalder et al., 2015)

Very similar to Ries’ loop, Thomke proposes a Learning Cycle process that contains
four iterative steps, beginning with the company design requirements. Following

6
several years of empirical research in large companies, Thomke showed technology
could be developed and commercialized more efficiently when many iterations are run.
In his book (Thomke, 2003), he compared the impact of computer simulation and rapid
prototyping in terms of efficiency, as defined as the number of eliminated errors
divided by the costs.

From the engineering design perspective, several new product development

approaches, with a specific focus on testing, can be mentioned. These include Stage-
Gate® Model (Cooper, 1990), and Verification and Validation (V&V) procedures. These
approaches ensure that a product-service meets design requirements and complies
with regulation and specifications (Babuska and Oden, 2004).

Another rigorous way of experimenting the performances of a technology is the DOE

(Design of Experiments) or experimental design (Fisher, 1937). DOE is a statistical
method that consists of predicting the outcomes of a system, following the change of
independent variables. However, this method is not yet adapted to the context of
entrepreneurship (Frederiksen and Brem, 2017).

In summary, the above-mentioned processes encourage the implementation of loop-

based tests, most often very late in the design process. These processes do not,
however, necessarily enable the verification of whether the innovation problem is well
expressed or not.

2.2.2. Tools to support decision making in experimentation

A tool is more often understood as a means to activate and operate a function. To

assist innovators’ decision making, there is an important number of visual tools, such
as graphical canvases and dashboards, test and learning cards (Osterwalder et al.,
2015), see Fig. 2); indicators such as the TRL (technology readiness level); as well as
more conceptual tools such as customer interviews. The Business Model Canvas
(BMC), Value Map and Customer Segment are very often used as support for
experimentation in the context of startups.

7
 Fig. 2 Business Model Canvas (BMC) (right) and Value Map and Customer Profile
(right) (Osterwalder and Pigneur, 2010; Osterwalder et al., 2015)

Other tools, such as Test Cards (see Fig. 3) and Learning Cards, are directly
implemented in testing. In addition, TLS or DT propose a set of tools for a given action,
or step of the innovation process, in a more opportunistic way. These tools can be
customer interviews, observation techniques, mind mapping, value chain analyses,
concierge services, A/B tests and fake door tests.

Fig. 3 Test Cards of the Value Proposition Design (Osterwalder et al., 2015)

Although these tools are meant to provide useful insights for entrepreneurs, they fail
at providing quantitative evidence that supports the investment decisions required for
experimentation.

2.2.3. Test objects

A test object is a physical or virtual model built to test something or act as a replicate
of something. These objects could encompass simulation models, service prototypes,

8
Minimum Viable Product (MVP)2, or Intermediary Design Objects (IDO)3 as introduced
by (Vinck and Jeantet, 1995).

Conventionally, an experiment is the test of a prototype (“primitive form” in Greek) or

proof of concept (PoC) to demonstrate that the concept is realistic and cultivates the
expected value. A prototype approximates the final product (Ulrich and Eppinger,
2015), and serves to elicit feedback from users, as well as verify that the concept is
feasible, and its expected qualities are achievable.

Prototypes can be categorized in terms of their nature (physical, virtual or service), but
also according in their high or low fidelity (‘lo-fi’ and ‘hi-fi’), i.e., their way of conveying
the features of the final product. Hi-fi prototypes (see (Pahl and Beitz, 2013; Ulrich and
Eppinger, 2015)) may contain alpha prototypes (used internally to assess whether the
product works as intended), beta prototypes (generally tested with users), as well as
final pilots (or pre-production prototypes). Building hi-fi prototypes may be a time
consuming and costly endeavor. Thus, one way to limit the resource usage and
investment of time is to build lo-fi prototypes that can be, among others, sketches,
wireframes (i.e., a refined presentation of the concept), and paper-prototype (Snyder,
2003). The lo-fi prototyping seems convenient in the context of limited time and
resources of startups. However, the lo-fi or early stage prototypes are often overlooked
by entrepreneurs (Clarysse and Kiefer, 2011).

In summary, there is a vast set of literature on prototyping categories and techniques.

Nevertheless, the existing literature lacks identification of markets and properties,
which a prototype may contain before its launch on the market.

2.2.4. Identification of hypotheses

A hypothesis is a testable proposition with a problem-solving goal. The definition of

hypotheses, before launching experiments, is crucial in the innovation process, as it
enables innovators to conduct structured experiments rather than trial-and-error ones.

Hypotheses can be identified through different techniques. In the VPD testing

approach (see Erreur ! Source du renvoi introuvable.), hypotheses are extracted from
different canvases. However, this extraction is not systematic and does not allow
verification of whether extracted hypotheses are well founded. To state and diversify
different hypotheses, Steve Blank put forward a flow for Customer Discovery (Blank,

2
The MVP concept is coined by the Lean Startup approach and consists in a product with just enough
features that customers will pay for.
3
IDO is designed to enhance the development process with the active participation of users.

9
2013a), that begins with the initial product hypothesis going through to the “Problem”
and “Product” hypotheses (see Fig. 4).

Fig. 4 Identification of problem and product hypotheses (Blank, 2013a)

Despite the existence of such techniques for identifying hypotheses, it is worth asking
whether innovators systematically hone them. For Thomke and Manzi, companies
quite often lack the discipline to identify relevant hypotheses. This lack of discipline
may lead to inefficient and costly tests (Thomke and Manzi, 2014). In the context of
technology-oriented innovative startups, a useful hypothesis might enable the
characterization of both market and technology properties from early developmental
stages. Nevertheless, no clear technique exists to support the extraction of relevant
hypotheses.

Based on our review of experimentation approaches, we have identified the following

fallibilities: (i) experimentation processes focus on late development stages but do not
enable investigation of users’ problems and usages, (ii) existing experimentation tools
fail at providing quantitative evidence in terms of technology and market priorities, (iii)
the existing prototyping techniques do not systematically specify the markets and the
properties that a prototype may contain, and (iv) there is no reliable technique to verify
whether hypotheses are formulated from a relevant basis.

To fulfill the lacks identified in the literature, this paper aims at presenting a new
methodology called Design by Usage-based Experimentation (DUE).

10
3. Introducing the DUE methodology

DUE: fundamental principles

DUE is a new methodology for efficiently testing a technology, specifying gaps that
need to be addressed through experiments in order to reach promising markets. DUE
and its components are a hybridization of several methods and tools reviewed in the
previous chapter. Fig. 5 depicts relevant methods and tools of DUE. In this figure, the
funnel narrows down towards approaches that have significantly inspired the DUE
methodology.

Fig. 5 DUE is inspired by the existing methods and tools

DUE is highly inspired by the usage-driven methodology of Radical Innovation Design®

(RID), founded on a set-based thinking approach (Sobek et al., 1999; Kennedy, 2008).
The set-based thinking approach makes a point of reducing uncertainties by
progressively converging to a satisfactory solution. This is in opposition to the point-
based design, which consists of testing an attractive design, as well as making several
design back loops to improve the solution. Following a set-based thinking approach,
DUE aims at identifying experiments that address markets with high potential for value
creation. These markets are identified through a systematic exploration of usage
situations and overlooked user problems from the front end of innovation. This falls
within the need seeker innovation strategy (Jaruzelski et al., 2016), which actively and
directly engages both current and potential customers. This engagement helps to
shape new solutions based on superior end-users understanding.

To gain an in-depth understanding of end-users, it is necessary to focus on the concept

of Usage, essential to the DUE methodology. Usage is a broad concept and does not
have a single definition. It can also be related to other concepts such as: usage
scenario, usage context or usage situation. A usage scenario is an archetypal story

11
lived by a representative group of users, where one may encounter an important
problem. A usage context, however, enables the capture of personal and
environmental factors and lifestyles (He et al., 2012). In DUE, the concept of usage and
usage situations not only includes end-user problems, but also enables the
quantification of the level of problems and their representativeness. This goes beyond
the mere persona method,4 with weak understanding of usage situations and their
importance.

DUE provides a toolkit to diagnose usage situations. This diagnosis allows the
identification of technology value creation potentials in different markets. This
diagnostic leads to the quantity of Value Buckets (VBs) as proposed by the Radical
Innovation Design® (RID) methodology (Lamé et al., 2017; Yannou et al., 2016). The
systematic identification of VBs in different markets enables identification of
promising markets in which the technology has a higher potential for user value
creation.

DUE process

The DUE process (see Fig. 6) is structured and is composed of two macro-stages and
four steps. These two consecutive macro-stages enables: first, the adoption of a
holistic view by questioning the current position of the technology and gathering the
existing data in a structured manner; and then exploring development opportunities
broadly. In practical terms, the first stage encourages the company to ask the question
of: “where am I standing?” And the second stage aims at investigating “where am I
heading?”. The purpose of dividing each macro-stage into two steps is due to the same
divergent and convergent nature of the macro-stages. This is reflected in each step.
The whole process remains very consistent with the Double Diamond model
(DesignCouncil, 2005) for product design.

The first macro-stage is called the “Investigation Stage”. The aim of this stage is to
determine: in which markets and technology properties it is worth investing time and
effort. As there are overlooked user problems and significant value buckets that are
possibly exploited by the technology, this stage allows the identification of the
technology’s main properties and links them to its promising markets. This stage
allows, overall, for the quantification of the market maturity gap. The intersection
between the first and second stages contains, thus, a justified perimeter of the market
maturity gap to be bridged through experiments.

The second stage of the process is called the “Planning and Implementation Stage”
which aims at specifying and conducting experiments and business strategies. Based

4
Persona is an imaginary character that embodies the needs and traits of specific target users.

12
on the quantified market maturity gap, experimentation strategies are identified and
prioritized in terms of the relative importance of the identified gaps, the cost and time
of each experiments.

Fig. 6 DUE process

The fourth step of the process is not detailed here, as its description exceeds the scope
of this paper.

4. Case study: a shape-changing technology

The DUE methodology has been deployed in a startup company that developed and
patented a shape-changing technology (see Fig. 7). Due to confidentiality, the name of
this company remains non-disclosed.

This startup’s technology is a shape-changing plastic material that becomes flexible

when plugged into an A/C current source. The prototype is almost instantaneously
deformable and reversible, making it conveniently incorporated into daily objects. The
user can adapt the material to an expected shape.

By the time the DUE methodology was applied, the startup had been in implementation
for a year and a half. Some of its technologies’ usages and markets had already been
identified. However, the identification of the latter was merely based on the intuition of
the entrepreneurs and their advisors5 (see Fig. 7 for examples in which technology can
be introduced to improve shape adaptation of objects, such as shin guards, medical
ventilation masks, and earphones).

5
Innovation advisors for this start-up were: the members of its advisory board, its business development
coaches at the innovation incubator in which the startup was incubated. These actors had a deep
knowledge regarding markets and the existing similar technologies.

13
 Sports Medical ventilation masks
Entertainment objects
protections (e.g., (e.g., borders of
(e.g., earphones)
shin guard) ventilation masks)

Fig. 7 Shape-changing technology: a schematic presentation of a section of the

prototype (up) and some of the identified usages (bottom)

After a brief discussion with the startup, it seemed that DUE was able to assist them
in two ways: first, the startup needed to systematically diagnose markets and users,
and second, to transform the technology into marketable products. The startup did not
initially know how to prioritize its R&D activities, mainly its resource-intensive
experiments.

The DUE process was implemented over a two-month period within the processes of
this startup. DUE was implemented through three collaborative sessions of one to two
hours each, in which the startups’ co-founders and innovation advisors were involved.
The goal of these sessions was to provide decision support for the startup on relevant
experiments to be conducted.

First step of the DUE process: Usage Diagnosis

4.1.1. Building a structured techno-market memory

The first task in the DUE process consists of creating a memory map of actions taken
by the startup company in terms of identifying usages, markets, and previously
conducted experiments: their conditions, cost, time, and results obtained. The startup
is requested to respond to a questionnaire of more than 50 questions (see 9.1 for the
topics of the questions). These responses are systematically captured and transferred
into Market-Technology Libraries and Experimentation Libraries. These libraries trace a
large part of the company’s R&D activities. They are therefore useful as a first-time
deliverable for a structuring startup. The format of these libraries can vary according
to the startup’s needs, but DUE puts forward graphical and synthetic documents, or
heuristic maps. The responses to the questionnaire, therefore, generate a first set of
graphical results (such as mind maps) in a manual fashion.

14
4.1.2. Measuring the relevance of the business model and its
components

The subsequent task consists of evaluating the business assumptions and ideas of
the startup by reflecting them into the graphical template of the Business Model
Canvas (BMC)6 (Osterwalder and Pigneur, 2010). This evaluation technique is
particularly beneficial, as it allows for better understanding of how the startup
expresses its strategic choices (e.g., its value proposition and identified customer
segments), its value network, as well as its cost-revenue streams. The BMC-RID
technique (see (Bekhradi et al., 2016)) is used here to evaluate and verify the maturity
of the UNPCS proofs (standing for Usefulness, Newness, Profitability, Concept, and
Stakeholder network). A set of questions based on 30 pieces of evidence supports the
task of assessing these proofs.

A primary assessment of the UNPCS proofs by the startup and its advisors
demonstrates that usefulness proof needs to be improved. This usefulness deficiency
is due to lack of satisfactory solutions to the users’/customers’ problems. A part from
this, these problems are not well identified and measured. Therefore, there exists a
need to systematically explore issues first. The initial tasks would be therefore to
screen a set of usages for the technology, and then characterizing important
innovation problems in each usage.

4.1.3. Eliciting and clustering usages of the technology

Further, to elicit and map the possible usages of the technology, a brainstorming
session is organized within the startup, including its innovation advisors. The
underlying question of this session is: “what are the activities in which the technology
can serve to solve a problem of users/customers?.” The result of this session enables
the clustering of identified usages into two distinct classes. This clustering is done
based on the identification of key activities that have emerged from a primary review
of usages. In fact, the technology enables the performance of two activities: adapting
to the shape of the human body; and adapting to the shape of an object (see Table 1).

Table 1. Usage categories of the shape-shifting technology

Name of the Number of Examples

Main activity
usage category identified usages of usages
• Earphones
Adapting to the shape of Human Body
66 • Shin guards
the human body Interface (HBI) • Shoes

6
The BMC is composed of nine blocks that are: Customer Segments, Value Proposition, Distribution
Channels, Customer Relationships, Revenue Streams, Key Activities, Key Resources, Key Partnerships,
and Cost Structure. The graphical template of the BMC is widely used by startup companies to describe
and share their business assumptions and ideas.

15
 • Packaging covers
Embedding and
Adapting to the shape of • Part of an industrial device
Structuring 20
an object or a decoration object
Objects (ESO)

Given the notable number of usages to deal with, at the stage of the process the
startup must decide about a category or a sub-set of usages to focus on. Thus, the
startup decided to focus on the first category that contained 66 usages. The second
deliverable of the first step was a synthetic heuristic map containing this class of
usages (see 9.2).

4.1.4. Exploring Deep Knowledge and developing skills

Once the usages are identified, the next task of the DUE process deals with a
systematic exploration of knowledge and skills, with regards to each identified usage
of the technology. Several categories of knowledge must be explored. These
categories, defined as Deep Knowledge (see, e.g., (Lamé et al., 2018; Bekhradi et al.,
2017)), consist of:

• frequent usage scenarios;

• important users’ problems;
• causes and consequences of these problems;
• more or less effective solutions (and technologies) that exist to deal with these
problems;
• users’ behaviors and perception of a similar technology;
• lifecycle of similar technologies (see, e.g., Gartner hype cycle or S-curve);
• reasons of failure for similar technologies; and
• existing standards, specifications, and regulations that the technology must
comply with.

Different techniques can help with investigating Deep Knowledge. These techniques
include among others: user surveys, technology and legal watches, numerical
simulations, and all other techniques7 that enable the company to acquire knowledge
and skills about the above-mentioned categories of Deep Knowledge.

In our case study, as the startup decided to focus on the first category of usages, called
human body interface, there was no need to acquire knowledge about how the
technology can better adapt to the shape of human body parts. This knowledge is not
necessarily available in the core competencies of the startup, as it involves a deep

7 One may also underline other investigation techniques or sources such as: interviewing lead users
(Von Hippel, 1986), questioning domain experts, reviewing scientific literature, reading technical reports
and sourcebooks, patent search, exploring databases, and reverse engineering.

16
knowledge about multiple disciplines, such as physics, healthcare, biology, and human
behavior. For instance, a specific know-how in physics, with the calculation of the
radius of curvature, and the mechanics of materials, is needed. In the case where the
focus is to design a shape-changing earphone, the startup must gain knowledge about
the human auditory system, anatomy of the ear, and the existing technologies. The
startup is therefore encouraged to explore newly related knowledge, instead of directly
building a prototype for adaptable earphones and conducting trials and errors.

Once a satisfactory amount of information is collected per each usage, the

“Dependency Structure Modeling - Value Buckets” (DSM-VB) algorithm ((Yannou et al.,
2016), (Lamé et al., 2017)) of the RID methodology is used for a primary computation.
Indeed, DSM-VB is a kind of Dependency Structure Modelling approach (see (Eppinger
and Browning, 2012), (Lindemann et al., 2009)) that models categories of problems,
usage scenarios, and existing solutions, into a series of matrices and set operations.
These matrices and set operations compute the poorly or frequently unsolved
problems into a set of common situations, in order to establish innovation
opportunities, called value buckets.

Second step of the DUE process: Knowledge

Exploration

4.2.1. Identifying and evaluating promising markets

The second step of the DUE process begins with implementing the DSM-VB algorithm.
The latter determines the difference between two spaces. The first space includes:
important problems occurring in frequent usage scenarios without considering the
existing solutions (see matrix A in Fig. 8). The second space includes important
problems, occurring in frequent usage scenarios, which are lowered thanks to
effectiveness of the existing solutions (see in Fig. 8 matrix D which is the product of C
and B). The important uncovered and frequent problems are value buckets that
represent value creation opportunities for the technology. We characterize a value
bucket as a promising market in the remainder of this study.

17
 Fig. 8 DSM value bucket mechanism

In our case study, 21 value buckets are identified and filtered from the 66 previously
identified usages. These value buckets, or promising markets, represent valuable
information for problems worthy of being addressed. For the sake of simplicity, in the
remainder of this paper, we only consider five out of 21 markets. These markets are
listed below with their related value bucket, expressed as a combination of problem
occurring in a usage scenario:

m1: cosmetic masks (poor oil distribution in the case of home usages);
m2: shin guards (injury and bone fracture in tackle receiving);
m3: ski boots (injury and musculoskeletal disorders in long-term use);
m4: orthopedic insoles (time and efficiency loss in customizing orthopedic insoles);
m5: earphones (bad fitting in the ears during sport and in-movement situations).

The next task in the DUE process consists of comparing these markets in terms of
their attractiveness and potential for economic value creation, acheived by assessing
the Market Attractiveness Index (MAI). In general, this index is calculated as the
weighted average of the MAI parameters. In the list below, the first four parameters
are borrowed from the works of (Best, 2012). The following six parameters are adapted
and defined in the context of a technology-oriented innovative startup:

• market size: total number of potential customers and companies on this

market;
• market potential (or growth): estimated maximum total sales revenue of all
suppliers of a product in a market during a certain period;
• ease of entry: startup costs and other obstacles such as competitors, or in other
words, entry barriers;
• distribution channel access: degree in which it was to locate channels and
related suppliers to deliver value proposition to customers;

18
 • likelihood of finding investors: probability of easily finding funds in a specific
market;
• customer acceptance of a technological innovation: degree of acceptance of
the technology regarding its perceived usefulness and/or perceived ease of use
(see, e.g., (Davis, 1989)).

As shown in Fig. 9, an intensity scale from 0 to 5 (with 5 as the most important

significant, i.e. the biggest or a very easily penetrable market) is used to assess the
value of each of the six parameters, categorized in three classes: market forces,
competitive intensity, and market access.

Fig. 9 MAI assessment for the five promising markets

Each value of the MAI matrix is called 𝑘! , where 𝑖 represents a market. The relative
importance of each parameter was then weighted by using a 0 to 1 intensity scale (𝑤).
All these values are determined by the startup based on their market studies, and are
then validated by the startup’s advisors or domain experts. Here, the relative value of
MAI matters, not its intrinsic value. The expert’s validation refers thus to the
classification of markets according to their MAI and not the given value of this
indicator. It must be noted that MAI allows a first market study based on high-level
market indicators and not on in-depth user’s problems to be solved in a given market.
The weighted average of the values of the six parameters yields the MAI per market
(𝑀𝐴𝐼! ):

#
)$!%& (𝑘! × 𝑤" )
"%&
(1)
𝑀𝐴𝐼! = .
𝑛

4.2.2. Characterizing key properties of the technology

The subsequent task of the process involves an in-depth analysis of the technology
through the identification of its underlying properties. As opposed to existing
approaches that merely focus on the technical functions or technology affordances

19
(see for instance (Gaver, 1991)), we emphasize the broader concept of usage and the
potential of the technology to satisfy usage scenarios, such as the mechanism of the
usage coverage indicator (UCI) already described in (Wang et al., 2013; Yannou, Yvars,
et al., 2013). Ontology of technology characterization is proposed in Fig. 10.

Fig. 10 Ontological model to identify the properties of a technology

In this case study, 14 properties of the technology are thoroughly defined (see the
complete list in 9.3). For the rest of this paper, and for the sake of clarity, only three
key properties are kept. These properties are:

p1: ability to cover bony parts of the human body;

p2: heat setting; and
p3: practicality in terms of volume and portability.

4.2.3. Measuring the market maturity gap

The next task consists of combining the identified markets and the technology
properties into a set of dependency matrices to compute the market maturity gap. An
extension of the DSM-VB algorithm is proposed in the DUE process. The mechanism
of this new algorithm is very similar to the value bucket algorithm. However, instead of
combining problems, usage scenarios, and existing solutions, the new algorithm links
the promising markets and the technology properties into a set of matrices. The output
of the algorithm is the quantified market maturity gap, that we may reasonably call the
meta value bucket, as it is fed by value buckets as well as the new algorithm which
follows the same mechanism as the DSM-VB. Therefore, we name the proposed
algorithm, DSM meta value buckets (DSM-mVB). It includes two main matrices (see
Fig. 11).

The first matrix (F) expresses the expected or required maturity of the technology
properties for promising markets. If the technology reaches this required maturity, it
can fulfill a set of conditions enabling it to easily penetrate the market. These
requirements could be the threshold level of a given performance, at which a

20
successful technology has already been released or can be approved for
commercialization (as is the case with FDA-regulated medical technologies).

The second matrix reflects the current maturity of the technology (see matrix G in Fig.
11). The latter expresses that, for the time being, the properties of the technology are
more or less mature enough to respond to the requirements in promising markets. The
subtraction of the second matrix from the first provides a new matrix that contains the
intrinsic market maturity gap. The latter must be normalized with regards to the
relative importance of the markets and of the properties of the technology. The
previously calculated MAI and the average of required maturity of the properties are
used to normalize the market maturity gap. Finally, matrix I calculates the normalized
market maturity gap.

Fig. 11 DSM meta value bucket mechanism

The data in each cell of these matrices indicate the importance, or the degree of
maturity, of a given property of the technology in a given market. An intensity scale
from 0 to 5 is used here to assess this maturity. If the value is 0, then it means that this
property is not relevant at all in this market; if the value is equal to 5, then this property
is very relevant to the market. The data required to complete these matrices are
declared by the startup. But, they are verified through other complementary sources:
(i) existing standards that a technology must comply with (a mandatory literature
review of the standards of each market); (ii) discussions with different experts who
have an extensive knowledge about both markets and technologies; (iii) startup’s
existing knowledge and experience.

Based on matrix I, a last matrix called “maturity level” (J) is computed. The DSM-mVB
algorithm thus determines the degree of the technology maturity in percentage:

𝐻!" 𝑀𝐴𝐼! 3#
𝐹
𝐽!" = 1 − &
𝑀𝑎𝑥!" +𝐻!" ,
×
𝑀𝑎𝑥(𝑀𝐴𝐼! )
×
𝑀𝑎𝑥+𝐹3# ,
4, (2)

where 𝐹0' is the average of the required maturity of the technology properties for all
markets. In the case where the value of a given cell of this matrix is equal to 100%, it
shows that the technology has already reached required maturity for this property in a
given market. If the value is more that 100%, it means that the technology has already

21
surpassed the required maturity level, and that the meta VB (or market maturity gap)
is already absorbed. In other words, there is no market maturity gap to be bridged.

Based on the quantified meta VBs, a decision must be made. In our case, the startup
decided to first focus on the most challenging meta value bucket, i.e., where the market
maturity gaps are the most important. The startup decided to focus on the following
meta VBs ranked from the importance of the market maturity gap:

• mVB1: ability to cover bony parts of the human body in the market of ski boots
(related to the VB: injury and musculoskeletal disorders in long-term use);
• mVB2: ability to cover bony parts of the human body in the market of shin guards
(related to the VB: injury and bone fracture in tackle receiving);
• mVB3: practicality in terms of volume and portability in the market of earphones
(related to the VB: (bad fitting in the ears during sport and in-movement situations).

In another context, the startup could have first selected the least important gap, as it
would call for less effort in terms of technology improvement. This is due to the
heightened maturity of the technology properties.

The choice and filtering of meta VBs are done according to the startup’s strategy,
technology development capacity, and fund raising priorities. The chosen sub-set of
meta VBs is documented and centralized in a Meta Value Bucket Hub. The role of this
hub is to trace the startup’s market and technology priorities. In the context of large
companies or startups with several departments, this hub enables the efficient
transfer of the their technology development priorities to other departments/persons
in charge of experimentation and business development.

Third step of the DUE process: Experimentation

Design

4.3.1. Formulating hypotheses based on the identified market

maturity gap

The selected meta VBs constitute a justified scope for the technology maturation.
They also provide rational grounds for hypotheses in experiments.

Inspired by (Osterwalder et al., 2015) and (Blank, 2013a), this step of the DUE process
provides seven generic guidelines to better define and express experiment
hypotheses:

22
 • generate hypotheses based on the most important gaps in the market or meta
value buckets;
• link each hypothesis to the problem to be solved in a given usage scenario (i.e.,
to a value bucket);
• make sure that the hypothesized relationships have not yet been tested (by
referring to the Experimentation Libraries internally (see 4.1.1), or by
investigating outside the company);
• postulate a hypothesis in the form of a relationship between variables
concerned;
• determine whether the hypothesis can be experimented and measured;
• define hypotheses that can be verified, if possible, according to a given
threshold;
• avoid combining hypotheses and refine similar hypotheses to merge or to
differentiate them.

In our case study, the first hypothesis (η) of the first mVB1 is named η1,1 and expressed
as: “increasing for two more millimeters the size of the external layer of the product
inserted inside the ski boot will twice absorb the shock, and will therefore reduce
important “injury and musculoskeletal disorders” in “long-term use”.”

4.3.2. Describing experiments in test cards

Once hypotheses are stated, the “DUE test cards” are proposed to describe the context
of experiments. The concept of Test Card is borrowed from the works of (Osterwalder
et al., 2015). Here, the classical Test Card is extended to encompass the identified
meta value buckets and value buckets.

Each DUE test card (see Fig. 12) contains three key pieces of information that are: a
meta value bucket (mVB), a hypothesis generated based from this mVB, and a specific
experimentation action to verify the identified hypothesis. A DUE test card is therefore
coded as follows: ηa,b,c, where a reflects the meta value bucket number, b the
hypothesis number, and c the number corresponding to the specific action to be
conducted.

Moreover, the DUE test card contains five indicators that are (see Fig. 12):

• relative importance of meta value bucket (compared to other identified meta

value buckets);
• criticality of the hypothesis (degree to which a generated hypothesis is
critical/has priority);
• data reliability (degree to which the results of experiments are reliable);

23
 • financial cost (which includes the cost of designing, building, running and
analyzing an experiment); and
• experiment lead-time (time from the initiation to completion of an experiment).

The value of a variable is estimated by the startup company on a scale of 0 to 5 (with

5 as the most expensive, most time-consuming). Experts and the startup’s advisors
then validate the estimated data for accuracy.

Fig. 12 DUE test card

4.3.3. Prioritizing and selecting experiments

Once several DUE test cards are generated, a decision must be made to focus on one
or several experiments. The DUE recommends the choice of those that are the most
economical and efficiently related to priority meta value buckets. Then, once these
experiments are conducted, investments can be steered to more reliable and critical
tests.

In our case study, the following experiment (η2,2,1) is selected by the startup to be
conducted as it is the most economical one. This scenario corresponds to:

• the second meta VB (mVB2):

o “ability to cover bony parts of the human body in the market of shin
guards, where there are injury and bone fracture problems in the situation
of tackle receiving”;
• the second hypothesis defined for this mVB:
o “twice rigid external layer applied on the shape shifting part of the shin
guard will better protect the tibia bone and consequently reduce bone
fracture by 10% in the tackle receiving situations”; and

24
 • the first set of experimentation actions for the second hypothesis of the second
mVB:
o “simulating different collisions on a twice rigid material and measuring
stiffness and damping coefficients of shock absorption on a fragile
bone”.

In the above-mentioned experiment, conducting simulations is economical and can

therefore be conducted quickly, despite the data reliability of the numerical simulation
being low. The identified meta value is ranked as the second most significant and the
test hypothesis as not as significant.

The fourth step of the DUE process is called Business Strategy Design and consists of
three tasks: (i) re-measuring the UNPCS proofs in the Business Model Canvas (BMC),
(ii) generating new BMCs, and (iii) specifying business and fund-raising roadmaps. As
the last step of the DUE process requires some preliminary results of experiments, this
step is not detailed here as it exceeds the scope of this paper.

A summary of the three first steps of the DUE process as well as the DUE data
collection and validation mechanism are detailed in the following section.

5. DUE in summary: a rigorous methodology to structure

and organize R&D activities

Detailed process of DUE

Each step of the DUE process is a flow of organized tasks that delivers intermediary
and final results. These organized tasks enable the reduction of costly and time-
consuming back-loops. Indeed, each deliverable or tool of the DUE is considered as a
function with an input and an expected output. The outputs of each step enter the
process, hence reaching a rational experimentation strategy. The detailed steps of the
DUE process including 13 functions/tasks are presented in Fig. 13.

25
 Fig. 13 Detailed steps of the DUE process

Here, one may question how to ensure that the provided tools are complete and
sufficient. Indeed, the proposed tools are considered as a set of interrelated functions
aiming at progressively gathering data to reduce uncertainties and achieve the
purpose of the DUE process. Even though there may be more tools that can be added
to this process, the expected outcome must remain the same. Therefore, once the
results per step and per macro-stage are satisfactory and accurate, the number of
used tools or their completeness does not matter.

The structured process of the DUE methodology contributes to organizing the R&D
activities of a company. In the context of innovative startups, an important part of their
investments in R&D activities are devoted to experimentation8. Therefore, the
effectiveness of these experiments is challenging for startups dealing with limited
time and resources. To address this challenge, the DUE process, prior to running
experiments, dedicates more attention to the investigation of problems to be solved,
and to the weaknesses of the technology to be improved.

Then, the experimentation design step of the DUE process aims at defining relevant
experiments based on a selection of the market maturity gaps (called meta value
buckets). These gaps constitute a relevant ground for systematically formulating the
hypothesis of experiments.

8
The entrepreneurship and innovation management literature emphasize that experimentation is the
most important R&D activity for technology-oriented innovative startups (see for instance (Kerr et al.,
2014; Blank, 2013a; Ries, 2011)). In general, testing activities (or experiments) account for more than
50% of companies R&D expenditures (see for instance (Bertolino, 2007)), and more than 75% of their
R&D time (see (Allen, 1966)).

26
 DUE: Data collection and verification process

Data used in the DUE process are collected from different sources, such as reviews of
existing documents on usages and technologies, references to (Market-Technology
Libraries), the startup’s previous experiments and obtained results (by referring to
Experimentation Libraries), agile-rapid micro-tests to capture data for future tests,
observations of usage situations, interviews with users and domain experts, as well as
a set of heuristics based on entrepreneurs’ intuition following brainstorming sessions.

All the above-mentioned elements are confirmed by innovation experts who work as
advisors and mentors for the startup, with a solid knowledge of different markets and
technologies. Most significantly, they play a key role in verifying the significance of the
declared data by the startup in a second round of data verification. The first round of
data verification is done internally and according to investigations made by
entrepreneurs. Moreover, the accuracy of the collected data (necessary mainly for
establishing the 0 to 5 intensity scales) is not an issue in the DUE methodology, since
the relative importance of the matrices and meta value buckets matters rather than
their intrinsic values.

Required resources and potential benefits of DUE to

startup’s decision making process

The implementation of the DUE process within the studied startup took 2 months,
through three collaborative sessions of almost 2 hours each. In total, it took almost 40
hours of cumulated work of 2 of the startup’s executives, one of the co-authors of this
paper (as observatory and methodological expert), as well as startups advisors and
market experts who were consulted occasionally.

In order to highlight the benefits of DUE in the context of the studied startup, we have
compared the obtained results after 2 months, owed to DUE’s historical data of the
startup during the exact same period of time. The human and financial resources
remained unchanged in both situations.

Fig. 14 shows that the number of obtained results and decision-making variables (e.g.,
the maturity of the technology, cost and time of experiments) with the DUE method
have significantly increased, compared to the conventional approach of the startup.

27
 Fig. 14 Comparisons between results obtained without and with the DUE method

Potential benefits of the DUE method in regards to time saving are depicted in Fig. 15.
With the DUE method, we lower time spent by 62%, between identifying a set of
markets and obtaining a shortlist of targeted markets. Moreover, Historical data
shows that on average the conventional approach of the startup is 82% more time-
consuming than the DUE method when it comes to identifying a given usage scenario
of the technology. Finally, we reduce by 64% the average time required to identify a
given property of the technology.

Fig. 15 Reducing the time spent to identify key decision making elements

Globally, these improvements and time savings contributed to the acceleration of the
startup’s decision makings, as well as broadening their vision regarding potential
markets and technology properties.

28
6. Discussions

An extensive literature review, of design and experimentation approaches to new

product development, shows that popular design and creativity methods have some
shortcomings when it comes to identifying and prioritizing promising markets for an
innovative technology. DUE provides some answers to these shortcomings by
systematically exploring usage scenarios, problems, as well as the properties of a
technology. In this context, DUE borrows the important concept of value buckets from
the Radical Innovation Design® methodology.

DUE provides a structured process along with organized steps and deliverables, unlike
the loop-based and so-called non-linear process of Design Thinking (DT) (Brown,
2009). Compared to DT, DUE questions the relevance of an innovation problem, with a
thorough analysis of usage situations and value buckets, before prototyping and
testing.

The same need observable in the context of The Lean Startup (TLS) approach (Ries,
2011). TLS advocates building MVPs (minimum viable products) and testing them in
a loop-based fashion, whereas DUE dedicates more time and effort to evaluate areas
where it is worth marshalling resources to develop and experiment a technology before
building an MVP. Hence, the DUE process falls within a set-based approach that begins
with exploring from the front end of innovation to progressively reduce uncertainties.

DUE is a hybridization of several methods and tools. Among them we can mention the
testing part of the VPD approach (Osterwalder et al., 2015), where Test Cards are a
powerful tool to structure the testing activity. However, the hypotheses extraction step
of VPD is merely based on the assumptions made in the BMC, Value, and Customer
Profile maps. The quantification of meta value buckets in the DUE process provides
justified grounds for formulating and prioritizing relevant hypothesis.

Based on the above-mentioned comparisons, DUE is a differentiating methodology to

organize and structure testing in a company (see Fig. 16).

29
 Fig. 16 From conventional loop-based testing processes (such as the process of
Design Thinking (up left), TLS’ Build-Measure-Learn loops (left middle), and testing
of the VPD (bottom left)) to a more structured experimentation process (right)

Although, the DUE methodology in its current form provides important decision-
making support, authors are also aware of its limitations. First of all, the listing of
usages and properties of the technology is supported by structured brainstorming
sessions and tools. Nevertheless, the comprehensiveness of these lists might be
questionable. There is therefore a need to set other complementary indicators and
threshold levels to determine the exhaustiveness of the identified usages and
technology properties.

It is obvious that at each step of the DUE process there are uncertainties associated
with the outcomes of deliverables. Different matrices are filled with quantitative data,
which translates qualitative elements to a value between 0 to 5. Therefore, in the
process of building the DUE process, we have tested the robustness of the results
provided by each matrix through a simple experiment.

At each step, authors have asked 2 of the startup’s executives to fulfill matrices
separately. The final results were quite close. The insights and comments gathered
from the startup enabled us to revamp some matrices and indicators. After these
modifications, a second round of tests reduced the delta between the final results
generated by 2 persons of the startup. At the end of this robustness experiment, the
slight difference between the results set the stage for an interesting debate between
the startup’s executives. This focus discussion enabled these executives to reach a
consensus and led to more accelerated decision-making than usual processes,
according to the startup. Therefore, this experiment validated the robustness of

30
results. The details and conditions of this experiment are not detailed here as they are
beyond the scope of this paper.

Another limitation concerns the choice of cost-effective experiments, which is done

manually for the time being. A multi-objective optimization could be relevant to identify
tradeoffs and Pareto-optimal experiments. Such an optimization model must be fed
with a more important set of data and not merely with a few examples.

Even though the DUE methodology is designed regardless of a specific startup context
or technology, its repeatability is considered as a current limitation. Several other
experiments are presently under process with other startup companies to verify
whether the method can be easily repeated in other startup contexts.

The principal implication of the DUE methodology is the change of the mainstream
R&D practices of startups, which consists of “Just launch a prototype, then let’s see”.
This vision is less adapted to the context of limited time and resources of technology
startups, especially when they are mostly funded by public funds (e.g., in France two
third of innovation-related funds are from public sources (Ezratty, 2018)). There is thus
a crucial need to allocate resources to those R&D activities, mainly those experiments
that enable the increase of the likelihood of success, instead of verifying uncertain
hypotheses. DUE takes a step forward to avoid waste and to secure the launch of an
innovation.

7. Conclusion and perspectives

Given their limited time and resources, tech startups cannot only rely on random, loop-
based and, more often, costly trials. The main contribution of the DUE (Design by
Usage-based Experimentation) methodology is a structured process that helps
technology startups make reasonable decisions on the maturation of their technology
for relevant markets. Concretely, DUE allows: (i) screening usages to elicit a list of
promising markets characterized as value buckets (i.e., important problems occurring
in frequent usage scenarios for which the existing solutions are not efficient) (Yannou
et al., 2016; Lamé et al., 2017), (ii) identifying the main properties of the technology,
(iii) linking the identified promising market (or value buckets) to the properties of the
technology, in order to assess meta value buckets (or those properties that must be
improved, through experiments, for the most promising markets), and at (iv) identifying
the prioritizing effective experiments.

The DUE methodology has been applied in the case of a shape-changing technology,
developed and patented by a startup in France. The results were new and useful for
the startup, as they have provided a structured way of specifying experiments. The
startup company used the results of this case study to consolidate its file for an

31
innovation prize competition. Two other case studies of the DUE methodology have
been conducted with two other technology startups in Montreal, Canada. The results
are currently under way and will enable the collection of more feedback on DUE for its
further improvements.

Our further work will focus on two objectives. First, automatizing the generation of the
Market-Technology Libraries with the help of deep learning algorithms. Indeed, the
internal and external information (regarding the R&D, markets, other technologies etc.)
can be automatically screened to improve the quality of the DUE input data process.
Second, the identification of cost-effective experiments is done manually in the current
version of the DUE methodology. In our future work, we will identify and detect the
Pareto-optimal testing activities.

8. References

Akao, Y., King, B., and Mazur, G. H. (1990). Quality function deployment: integrating
customer requirements into product design (Vol. 21): Productivity press
Cambridge, MA.

Allen, T. J. (1966). Studies of the problem-solving process in engineering design. IEEE

Transactions on Engineering Management(2), 72-83.

Babuska, I., and Oden, J. T. (2004). Verification and validation in computational

engineering and science: basic concepts. Computer Methods in Applied
Mechanics and Engineering, 193(36), 4057-4066.

Bekhradi, A., Yannou, B., and Cluzel, F. (2016). Importance of problem setting before
developing a business model canvas. Paper presented at the DESIGN 2016 14th
International Design Conference.

Bekhradi, A., Yannou, B., Cluzel, F., and Vallette, T. (2017). Categorizing user pains,
usage situations and existing solutions in front end of innovation: The case of
smart lighting project. Paper presented at the The 21st International Conference
on Engineering Design, ICED17, Vancouver, British Columbia, Canada. August 21-
25.

Bertolino, A. (2007). Software testing research: Achievements, challenges, dreams.

Paper presented at the 2007 Future of Software Engineering.

Best, R. (2012). Market-based management: Pearson Higher Ed. 9780133071627.

Blank, S. (2013a). The four steps to the epiphany: successful strategies for products
that win. New Jersey, United States: BookBaby. 0989200523.

Blank, S. (2013b). Why the lean start-up changes everything. Harvard business review,
91(5), 63-72.

32
Brown, T. (2009). Change by design: How Design Thinking Transforms Organizations
and Inspires Innovation. New York, United States: HarperBusiness.

Cascini, G., Rissone, P., Rotini, F., and Russo, D. (2011). Systematic design through the
integration of TRIZ and optimization tools. Procedia Engineering, 9, 674-679.

Cascini, G., Rotini, F., and Russo, D. (2009). Functional modeling for TRIZ-based
evolutionary analyses. Paper presented at the DS 58-5: Proceedings of ICED 09,
the 17th International Conference on Engineering Design, Vol. 5, Design Methods
and Tools (pt. 1), Palo Alto, CA, USA, 24.-27.08. 2009.

Clarysse, B., and Kiefer, S. (2011). The Smart Entrepreneur: How to Build for a
Successful Business: Elliott & Thompson. 1904027881.

Cooper, R. G. (1990). Stage-gate systems: a new tool for managing new products.
Business horizons, 33(3), 44-54.

Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance
of information technology. MIS quarterly, 319-340.

DesignCouncil. (2005). The ‘double diamond’ design process model. Design Council.
Retrieved from: [Link]
process-what-double-diamond. Access date: July 2018.

Eppinger, S. D., and Ulrich, K. T. (1995). Product design and development.

Eppinger, S.D. and Browning, T.R., Design Structure Matrix Methods and Applications,
MIT Press, Cambridge, 2012.

Ezratty, O. (2018, April 2018). Guide des Startups en France. Retrieved from:
[Link]
content/themes/Ezratty5/[Link]?file=/Files/Publications/Guide%20des%20Star
tups%20en%20France%20Olivier%20Ezratty%[Link]. Access date: July
2018.

Fisher, R. A. (1937). The design of experiments: Oliver And Boyd; Edinburgh; London.

Frederiksen, D. L., and Brem, A. (2017). How do entrepreneurs think they create value?
A scientific reflection of Eric Ries’ Lean Startup approach. International
Entrepreneurship and Management Journal, 13(1), 169-189.

Gaver, W. W. (1991). Technology affordances. Paper presented at the Proceedings of

the SIGCHI conference on Human factors in computing systems.

George, M. (2002). Lean six sigma: combining six sigma quality with Lean production
speed, combining six sigma quality with Lean production speed: McGraw Hill
Professional.

Gibbons, S. (2016). Design Thinking 101. Retrieved from:

[Link] Access date: July 2018.

33
Hatchuel, A., and Weil, B. (2009). CK design theory: an advanced formulation. Research
in engineering design, 19(4), 181.

He, L., Chen, W., Hoyle, C., and Yannou, B. (2012). Choice Modeling for Usage Context-
Based Design. Journal of Mechanical Design, 134(3), 031007-031007-031011.
doi:10.1115/1.4005860

Jaruzelski, B., Staak, V., and Shinozak, A. (2016, Winter 2016). Global Innovation 1000
: Software-as-a-Catalyst, The world’s largest innovators are shifting the focus of
their R&D investments from products to programming. strategy+business.

Johansson-Sköldberg, U., Woodilla, J., and Çetinkaya, M. (2013). Design thinking: past,
present and possible futures. Creativity and Innovation Management, 22(2), 121-
146.

Kass, R. A. (2008). Tests and Experiments: Similarities and Differences. International

Test and Evaluation Association, 29, 294–300.

Kennedy, M. (2008). Ready, set, dominate: implement Toyota's set-based learning for
developing products and nobody can catch you: Oaklea Press. 1892538407.

Kerr, W. R., Nanda, R., and Rhodes-Kropf, M. (2014). Entrepreneurship as

experimentation. Journal of Economic Perspectives, 28(3), 25-48.

Lamé, G., Leroy, Y., and Yannou, B. (2017). Ecodesign tools in the construction sector:
Analyzing usage inadequacies with designers' needs. Journal of cleaner
production, 148, 60-72.

Lamé, G., Yannou, B., and Cluzel, F. (2018). Analyzing RID Methodology through the Lens
of Innovative Abduction. Paper presented at the 15th International Design
Conference.

Lindemann U., Maurer M., Braun T. (2009) Structural Complexity Management - An

approach for the field of product design, Berlin Heidelberg, Springer

Nobel, C. (2011). Teaching a ‘Lean Startup’Strategy. HBS Working Knowledge, 1-2.

Osterwalder, A., and Pigneur, Y. (2010). Business Model Generation: A Handbook for
Visionaries, Game Changers, and Challengers. New Jersey, United States: Wiley.
9780470901038.

Osterwalder, A., Pigneur, Y., Bernarda, G., Smith, A., and Papadakos, T. (2015). Value
Proposition Design: How to Create Products and Services Customers Want. New
Jersey, United States: Wiley. 9781118968055.

Pahl, G., and Beitz, W. (2013). Engineering design: a systematic approach: Springer
Science & Business Media. 1447135814.

Ries, E. (2011). The Lean Startup: How Today's Entrepreneurs Use Continuous
Innovation to Create Radically Successful Businesses. New York, United States:
Crown Publishing Group. 9780307887917.

34
Snyder, C. (2003). Paper prototyping: The fast and easy way to design and refine user
interfaces: Morgan Kaufmann. 1558608702.

Sobek, D. K., Ward, A. C., and Liker, J. K. (1999). Toyota's principles of set-based
concurrent engineering. Sloan management review, 40(2), 67.

Suh, N. P. (2001). Axiomatic design: Advances and applications (the oxford series on
advanced manufacturing).

Thomke, S., and Manzi, J. (2014). The discipline of business experimentation. Harvard
business review, 92(12), 17.

Thomke, S. H. (2003). Experimentation matters: unlocking the potential of new

technologies for innovation: Harvard Business Press. 1578517508.

Ulrich, K., and Eppinger, S. (2015). Product design and development: McGraw-Hill
Higher Education. 1259297136.

Vinck, D., and Jeantet, A. (1995). Mediating and commissioning objects in the
sociotechnical process of product design: a conceptual approach: Directorate
General Science, R&D.

Von Hippel, E. (1986). Lead users: a source of novel product concepts. Management
Science, 32(7), 791-805.

Wang, J., Yannou, B., Alizon, F., and Yvars, P.-A. (2013). A usage coverage-based
approach for assessing product family design. Engineering With Computers,
29(4), 449-465.

Wu, J. J., and Atkinson, R. (2017, November 2017). How Technology-Based Start-Ups
Support US Economic Growth. Retrieved from: https://
[Link] Access date: July
2018.

Yannou, B., Cluzel, F., and Farel, R. (2016). Capturing the relevant problems leading to
pain and usage driven innovations: the DSM Value Bucket algorithm. Concurrent
Engineering: Research and Applications, 1 - 16. doi:10.1177/1063293x16666311

Yannou, B., Jankovic, M., Leroy, Y., and Kremer, G. E. O. (2013). Observations from
radical innovation projects considering the company context. Journal of
Mechanical Design, 135(2), 021005.

Yannou, B., Lamé, G., and Cluzel, F. (2018). Adapting the FBS model of designing for
usage-driven innovation processes. Paper presented at the IDETC/CIE 2018:
International Design Engineering Technical Conferences/CIE: Computers and
Information in Engineering, Montreal, Canada. August 2018.

Yannou, B., Yvars, P.-A., Hoyle, C., and Chen, W. (2013). Set-based design by simulation
of usage scenario coverage. Journal of Engineering Design, 24(8), 575-603.

35
9. Appendices

Appendix 1. Topics of the market-technology

questionnaire of DUE
R&D activities
R&D general activities (except Design and development of new products (innovation methods/tools already used)
from experimentation) Basic research activities
Applied research activities
Research of new applications for existing products
Improvement of manufacturing or production technologies
Methods/tools of production management and tools
Actions to reduce waste and by-products
Design and development of scaled-up manufacturing processes
Design and ergonomic issues
Research to receive certifications, patents and/or licenses
Open innovation mechanisms (partnership…)
Experimentation Motivations and reasons for deploying such an experiment
Methods and tools of testing
Hypotheses to be tested
Protocols
Observation and evaluation techniques
Involved actors
Constraints
Difficulties to conduct this experiment
Obtained results
Usages of the technology
Usages of the technology List of the already identified usages and use-cases of the technology
Focus on particular usages more than others
Reasons and motivations for focusing on a particular usage
Main activities that the technology enables to perform (e.g., improving, optimizing something)
Usages of similar technologies
Possible users
Usage patterns (i.e., the way of using a technology by users)
Usage scenarios (i.e., archetypal stories explaining a given situation)
Common denominators of activities/usages
Deep Knowledge
Deep Knowledge categories Frequent usage scenarios (incl. data about users, demographics
Important users’ problems
Causes and consequences of these problems
More or less effective solutions (and technologies) that exist to deal with these problems
Users’ behaviors and perception of a similar technology
Lifecycle of similar technologies (see, e.g., Gartner chart and S-curve);
Failure reasons of the similar technologies, their causes and consequences
Existing standards, specifications, and regulations that the technology must comply with
Market analysis
Previous market studies Existing documents
Market segmentation and its logic
Positioning on the market
Mapping of the environment and strategic analysis (SWOT, PESTEL...)
Barriers to entry
Analysis of growth drivers
Market size
Business model
BMC components Value offer
Customer segments
Key partners
Key activities
Key resources
Cost structure
Revenue streams
Customer relationships
Distribution and sales channels

36
Appendix 2. Human Body Interface (HBI) usage
category

37
 Appendix 3. Identified properties of the shape-
changing technology

Properties of the
No. Definition
technology
Covering surfaces with direct contact with bones (e.g., cheek, back of head,
Ability to cover bony parts
1 forehead, hip, leg…) through a rigid contact surface in order to protect the body
of the human body
from external shocks or dangerous movements.
Ability to cover non-pliable Covering non-articular parts of the human body (e.g., face, abdomen, dorsum…) in
2
parts of the human body order to adapt their morphology.
Ability to cover pliable Covering articular parts of the human body (elbow, wrist, shoulder…) in order to
3
parts of the human body take the form of its pliable parts.
Immobilizing pliable or non-pliable parts of the human body after a fracture or
Ability to immobilize parts
4 accident to avoid medical disorders or, to fix a part of the human body for specific
of the human body
medical interventions (e.g., surgery, radiology…).
Ability to vary the Variability in terms of thickness of the surface to be adaptable to desired usage
5 thickness of the situations (without harming other abilities) (e.g., some parts of the covered surface
deformable surface need to be protected with a thicker deformable surface).
Acceptability for a long-
6 Quality of the used materials that must not cause allergies or cutaneous problems.
term contact with the skin
Adapting to the form of a surface with complex and/or variable geometry (e.g., on
Adaptation to complex and
7 the surface of the human body or objects in contact with the body), i.e., being
polygonal surfaces
polymorphic in order to cover polygonal surfaces and being ergonomic.
Capacity of shifting from a Shifting from different shapes to others from the initial state (i.e., reversibility) to
8
shape to another be able to build multiple forms from a current state.
Hardness of the Degree of rigidity of the deformable surface to resist to shocks or to immobilize in
9
deformable surface a satisfying way.
Homogenous deformation
10 Degree of homogeneity on the whole surface to guarantee a coherent coverage.
on the deformable surface
Possibility of the heat Possibility to control the heat degree and its duration to obtain an ideal thermal
11
setting (thermal comfort) comfort.
Practicality in terms of Degree of importance of considering product’s weight in different usage situations
12
volume and portability mainly if carried on the body.
Degree of robustness of the used materials in order to resist multiple shape
Reliability of materials for
13 shifting in different usage situations depending on the number of
multiple shape shifting
deformations/formations and desired geometries before becoming obsolete.
14 Speed of shape shifting Time needed to switch between different forms until becoming rigid enough.

38