Examining the Relationship Between Value Propositions and Scaling Value for New Companies

The objective of this paper is to examine the extant value proposition literature and put forward our beliefs about how value propositions relate to scaling new company value rapidly. We conceptualize the management of the value proposition-scaling relationships as being like the management of part-whole relationships (Van de Ven, 1986), wherein value propositions are the parts and scaling company value is the whole.


I. Introduction
A new company committed to scaling their company value rapidly must develop value propositions for diverse parties. This includes not just identifying value propositions for customers, but also aligning these value propositions with scaling initiatives, and activities that the new company carries out to scale rapidly. This is reported as a major challenge worldwide, which we surmise is one of the main reasons why most new companies do not scale their company value rapidly.
Managing the value proposition-scaling relationship in a new company context is so far little understood. Even when companies try to shape multiple value propositions, they tend to align them only on a single customer value proposition, yet with little connection to their overall scaling objectives for the short-, midand long-term. Thus, many new companies do not scale because they were not in the first place designed to scale in the initial stages of their existence. Interestingly, regional business incubators and

Examining the Relationship Between Value Propositions and Scaling Value for New Companies Tony Bailetti and Stoyan Tanev
To scale company value rapidly, a new company needs to develop value propositions for diverse parties -customers, investors, partners, suppliers, employees, and other resource owners, as well as align these value propositions with its scaling objectives. The purpose of this paper is to examine the relationships between value propositions for a diverse set of parties, and efforts from a new company to scale company value rapidly. We review the value proposition literature and then examine the relationships between 19 assertions about value propositions, as well as six stable topics that best describe the SERS corpus, which is comprised of 137 assertions about scaling companies early, rapidly, and securely. Conducting a topic model of eight topics led to six stable topics: Fundraise, Enable, Position, Communicate, Innovate, and Complement. We find that of the 19 assertions about value propositions, four are connected to Complement, four to Innovate, one to Position, one to Fundraise, and one to Communicate. A total of eight assertions about value propositions are not connected to any of the six stable topics. This paper contributes to our understanding of how a new company scales company value rapidly, adding an application of topic modelling to perform small-scale data analysis. The findings are expected to be relevant to entrepreneurs and new companies worldwide.

II. Literature review
Value propositions "Value proposition" is one of the most widely used terms in business (Payne et al., 2017;Anderson et al., 2006). According to Webster (2002), a value proposition should be the company's single most important organizing principle. Lanning (2000), however, argues that "value proposition" as a term "is frequently tossed about casually and applied in a trivial fashion rather than in a much more strategic, rigorous and actionable manner." Much of the older literature adopts a one-sided perspective stressing that value is predetermined by the supplier, and then delivered to customers (Kowalkowski, 2011). Few researchers, however, have emphasized the importance of considering the broad range of stakeholders involved in the value creation process (Gummesson, 2006;Mish & Scammon, 2010;Frow & Payne, 2011).

Resource-based view
The resource-based view of the company (Wernerfelt, 1984) has become influential in understanding how companies attain competitive performance gains based on their resources and capabilities (Alvarez & Barney, 2002). According to Srivastava et al. (2001), "Resource-based view research must always endeavour to identify precisely what customer value in the form of specific attributes, benefits, attitudes and network effects is intended, generated and sustained." Clulow, Barry and Gerstman (2007) examine whether the key resources that hold value for a company also hold value for the company's customers. These studies focus on customer value only and adopt a static perspective regarding resource configuration. This perspective does not help in explaining how new companies can combine internal and external resources to shape value propositions that align with their business strategies.
Later developments of the theory attempted to explain how companies could do that in situations of rapid and unpredictable change (Teece et al., 1997;Eisenhardt & Martin, 2000). This work complemented the resourcebased view of a company by focusing on the role of dynamic capabilities, that is, the main routines that allow a company to change and reconfigure its resources when the opportunity or need arises

III. Method
We first use the Latent Dirichlet Allocation (LDA) algorithm (Blei et al., 2003;Blei, 2012) to build a topic per assertion model, and a keywords per topic model, both modeled as Dirichlet distributions. We then describe the connections between the stable topics and (i) the keywords, as well as (ii) the value proposition assertions included in the corpus.
LDA considers every assertion to be a mixture of topics, and every topic to be a mixture of words. Words can be shared between topics and the topics can be shared among assertions. LDA identifies combinations of words that tend to appear together in a way that suggests that specific topics are latently present in the corpus of assertions. In addition, LDA organizes the corpus by clustering the assertions that correspond to each topic. The assertions in each cluster are ranked in terms of the degree of their association with each topic. The topical organization of the assertions enables the thematic substantiation of the topics through a closer examination of the assertions (Boyd-Graber et al., 2017).

Topic model
Topic modeling was done using Orange 3.24.1 (Orange, 2020) to extract latent topics from the corpus comprised of 137 assertions and investigate the relationship between the 19 specific value proposition assertions and the topics extracted from the corpus. Each topic represents a set of words extracted from the 137 assertions. The topic-word connection is based on how well the word fits with the topic, while the topicassertion connection is made based on what topics the assertion addressed. The number of topics used to produce the topic model ranged from 3 to 10. The decision on the number of topics of the final model was made by the authors of the paper based on the joint assessment of the weights of the assertions per topic.

Topic stability
Topic stability was determined by running the final model four times, manually assessing the consistency of topics appearing across the four model runs and topic quality (Xing & Paul, 2018). For each topic, we determined that a topic was stable if five or more keywords appeared repeatedly in the four runs of the final model, and if the weights of the keywords were greater than 2. Topic quality was determined based on a joint judgment of the paper's authors.

Relationship between value proposition assertions and topics
For each topic (regardless whether stable or unstable), the assertions were categorized by topic loading into (i) Equal or greater than 0.6, and (ii) Less than 0.6.

Labelling and describing topics
To label and succinctly describe the topics, we used keywords and assertions with a topic loading greater than 0.6, along with our expertise in examining the content of the text documents (that is, assertions) associated with specific topics.

Corpus
The corpus is comprised of 137 assertions that are expressed using 2,591 keywords. On average, each

Number of topics
The topic modeling analysis iterated between three and ten topics. The authors decided that the best model was the one that had eight topics because the number of assertions that had topic loadings greater than .6 was at least 3 for each of the four model runs, and the results made the most sense in the context of the research topic. Table 1 provides the keyword distribution for eight topics resulting from four runs of the topic model. Each run provided slightly different results in terms of the composition, ordering, and ranking of words. This is due to the probabilistic nature of the LDA method, which requires performing and comparing multiple runs using the same number of topics.

Keyword distribution of four runs of the final topic model
In Table 1, the rows show the keywords associated with each topic. The keywords in italics appeared in all four runs of the topic model. The keywords shown in plain text appeared in 3 of the 4 runs of a topic model. The other keywords are not shown.

Stable topics
Six of the eight topics (that is, Topics A, D, E, F, G, and H), were deemed to be stable because at least five keywords appeared three or four times during the four runs of the model, and each had a weight greater than 2. Table 2 provides the topic labels and succinct descriptions of the six topics deemed to be stable. Each topic description built on the keywords shown in Table  1. Table 3 provides the 11 value proposition assertions found to be connected to the six stable topics. A value proposition was connected to a topic if its topic loading was equal to or greater than 0.6.

V. Discussion
The topic model results suggest that the initiatives that  new companies carry out to scale company value rapidly, can be organized into six topics: Fundraise (align returns to investor capital with scale opportunity); Enable (make others successful); Position (strengthen position among members of the network upon which a company depends to scale); Communicate (eliminate communication barriers); Innovate (continuously deliver innovative products and services and improve value propositions), and Complement (align benefits to customers, resource owners and other key stakeholders).
The 11 value proposition assertions are connected to five of the six stable topics. By "connected", we mean that a value proposition has a topic loading equal to or greater than 0.6. Of the 11, eight value proposition assertions are connected to two topics: Complement and Innovate. The four value proposition assertions connected to the Complement topic focus on aligning value propositions across parties, and offering benefits to multiple parties, not just customers.
The topic Innovate includes four value proposition assertions that focus on 1) integrating social impact aspects of value into the value propositions for all parties, 2) delivering high value to customers before, during, and after they use products or consume services, 3) innovating to create new value; and 4) tracking value propositions.
The value proposition assertion for employees is connected to Communicate, for investors relates to Fundraising, and for value chain members with Positioning.

VI. Conclusions
We reviewed the literature on value propositions and found that there is a need for a better understanding of how new companies manage the relationships between their value propositions to diverse parties, as well as what their initiatives are to scale company value rapidly.
We used topic modelling to examine the relationship between 19 assertions about value propositions and topics extracted from a corpus comprised of 137 assertions about how new companies scale rapidly.
We argue that entrepreneurs should use a multi-party perspective to develop value propositions for their new companies, beyond just a customer value proposition perspective. We also argue that initiatives to scale company value rapidly can be organized into six main topics, and that value propositions to multiple parties are connected to five of these six topics.
The paper's methodology also contributes to the literature on topic modeling. First, it demonstrates how practical insights can be extracted from a small data set, and second it offers a process to measure topic stability for more robust modeling, which researchers can use in future studies.