All posts by thijmencalis

Signal to Survive on Kiva

Microfinance: the traditional model
Microfinance encompasses a range of financial services targeted at the poor and provided by Microfinance Institutions (MFIs) (Christen et al. 2004). For this reason MFIs tend to operate in developing countries with the aim to provide financial inclusion to people who are rejected by the traditional banks. The initial capital injection and continued sponsoring for MFIs is often provided by ‘charitable or governmental agencies’ (Caudill et al, 2009). Subsequently, the MFI is able to loan to borrowers and pay for operating costs. The borrowers, pay interest on their loan, allowing the MFI to run a sustainable business.

Microfinance: emerging business models
The Global Symposium in 2017 in Kuala Lumpur opened in its booklet with the following insights:customer centricity, new business models and role of regulation. These chapters strongly resemble the themes found in a Customer Centric Digital Commerce course at university these days. Whereas often the disruption of industries due to new business models and increased customer centricity is discussed in the context of for-profit firms, this symposium highlights the importance of these themes for the non-profit sector.

For the microfinance sector, three new microfinance business models have been identified as promising models for the future: mobile money, agent banking and crowdfunding. The new models are able to improve the traditional microfinance model by increasing the reach of customers, increasing the funds available for loans, while lowering transaction and operation costs. In this blogpost I dive further into the crodwfunding model for microfinance and review a related paper. According to Hamari et al. (2016) crowdfunding and other sharing economy models have emerged due to a range of technological developments which “simplified sharing of both physical and nonphysical goods and services through the availability of various information systems on the Internet”. In fact all the emerging microfinance models noted by the World Bank are made possible due to technological advancements and adoptions in developing countries.

Kiva – A micro-credit crowdfunding platform
The platform is the most prominant example of a crodwfunding, or more specific peer-to-peer-funding business model. Kiva allows lenders from all over the world to invest $25 or more to specific entrepreneurs all over the world. This significantly increases the pool of capital available to people in need. At the same time lenders have the option to contribute directly to specific opportunities in a simple and low-cost manner. The money of the lenders is sent via the platform the the MFIs who are in contact with the the entrepreneurs. The entrepreneurs repay the loan with interest to the MFI. The MFI can keep the interest to pay for its operating expenses and the lender receives back it loan at 0% interest.

Screen Shot 2018-03-06 at 13.03.51

Kiva – Challenge
Whereas in the traditional business model the MFI decided which entrepreneur would receive a loan, in the Kiva model the lenders on the platform make this decision. This comes with the challenge that the MFI on the ground has more information about the characteristics of the entrepreneur than the lender behind his computer. Hence, crodwfunding increases theAsymmetric information is increased in the context of crowdfunding. This is where the research of Moss, Neubam and Meyskens (2015) comes in play. Their paper: “The effect of virtuous and entrepreneurial orientations on microfinance lending and repayment: a signaling theory perspective” will be reviewed in the subsequent sections.

Screen Shot 2018-03-06 at 13.01.15

Paper Review
Signaling in general is a strategy where the informed party can use signal about its unobserved characteristics to reduce the information gap of the uninformed party. The practice of signaling has been studied intensively in the IPO market. Language in IPO brochures was able to reduce investor uncertainty and consequently increased the propensity to invest in the shares of a company. Moss et al. (2015) want to verify whether signaling is also beneficial in the context of microfinance platforms. Borrowers/entrepreneurs could reduce the information gap of investors by signalling unobserved characteristics and intentions with the language used on the Kiva platform in their loan request description. Note that the term borrowers and entrepreneurs are used interchangeably in this blog as well as in the paper of review, and refer the the people receiving the loan. The term investors refers to the people actually providing the loans

The authors focus on these two characteristics/intentions in the language of the entrepreneurs’ loan-texts:

1) Virtuous Orientation (VO), the entrepreneur’s positive characteristics toward ethical and virtuous values and behaviours. VO is captured by the constructs conscientiousness, courage, empathy, integrity, warmth, and zeal.
2) Entrepreneurial Orientation (EO), reflected in the constructs autonomy, competitive aggressiveness, innovativeness, proactiveness and risk-taking.

The paper hypothesize the following:
H1 – VO constructs positively influences probability and speed to get funded
H2 – VO constructs positively influences the repayment probability and speed
H3 – EO constructs positively influences probability and speed to get funded
H4 – EO constructs positively influences the repayment probability and speed

The authors argue theoretically that VO and EO should give lenders a positive impression of the entrepreneurs increasing the chance to get funded and paid back. For example, an integer (part of VO) or innovative (part of EO) entrepreneur would be more likely to perform better. Hence the entrepreneur uses the loan optimally for the social and economic impact described and at the same time would be able to pay back the loan. Such an entrepreneur constitute a better investment decision for an investor than one which does these qualities. One could question whether the provision of signal is costly on the Kiva-platform. The authors conclude this is and for the sake of limiting the word-count of this blog, I would like to refer to page 31 of the paper for further reading on the argumentations.

400,000 loans over the 2006-2012 period are studied. There are four dependent variables used. For loan received as well as loan repaid a binary variable is used, indicating yes or no or a continuous variable is used indicating the speed at which this happens.
The independent variables are the individual constructs (e.g. autonomy) within VO and EO. These constructs are measured within loan-text using a computer-assisted text analysis program (LIWC). The program analysis the word count of words within individual constructs, stored in a custom dictionary, relative to text length. Follow this link for an example of a loan page:
As control variables the year, creditworthiness of the MFI, the loan size and country-specific variables (GDPC and infant mortality) are used. A Cox proportional hazard model is applied to calculate the impact of VO and EO on funding and repayment probabilities and speed.

Results & Practical significance.
The findings and their corresponding hypotheses are listed below:
H1 → Conscientious, courage, empathy and warmth have a significant negative effect on funding. Integrity and zeal are not significant.
H2 → Conscientious, courage, warmth and zeal have a significant negative effect on repayment. Empathy and integrity are not significant. “For each incremental use of a word related to zeal will decrease the chance of repayment by 3%” (Moss et al. 2015, p.47).
H3 → autonomy, competitive aggressiveness and risk-taking have a significant positive effect on funding. Innovativeness and proactiveness are not significant. “For every incremental word ‘autonomy’ (for every 100 words in the narrative), there is a 3% increase in the likelihood of receiving funding” (Moss et al. 2015, p.46).
H4 → proactiveness has a significant negative effect on repayment. Autonomy, competitive aggressiveness, innovativeness and risk-taking are not significant.

Discussion & Implication
Apart from the H3 results the results are counterintuitive to the theoretical argumentation. The lack of VO impact on funding might be explained as follows. Entrepreneurs on this platform request a loan in order to provide for their family and be able to survive. Lenders might believe that all the borrowers are “in essence ethical and virtuous” because income is provided to the poor (Moss et al. 2015). Signalling virtuous characteristics becomes less important in the case of microcredit, compared to entrepreneurial orientation. The latter  orientation signals more business sense. This study implies that a borrower who posts a loan Kiva should focus its signalling efforts on the entrepreneurial intentions rather than the virtuous characteristics.

Relevance of findings
The paper researches the impact of signaling in the emergent business model of crowdfunded microloans. As I describe in the introduction of this blog, the need to understand the asymmetric information problem in this emergent business model are high and hence the findings of this paper show how signaling positively and negatively influences funding and repayment performances. This findings can be used to further optimize the lending efficiency aspired by crowdfunded microfinance.

Broader understanding of conditional impact EO and VO signals
Whereas previous research has found in different contexts that EO and VO signals positively contributed to performance measures, this paper creates a more nuanced view. This paper shows that whether to use EO and VO signals highly depends on the context. In this case VO does not seem to work as well in microfinance-context as has affected funding propensity in the IPO-context.

Limitation of word count analysis
The results from this paper are as limited as the analysis used to arrive at them. The linguistic analysis simply counts the words, of a specific dictionary, present in loan texts. However, we should keep in mind that language is overly more complex than simply a bag of words (Wilson et al., 2005). The interplay between words and whole sentences convey a lot of meaning which is lost in this analysis. Further research could replicate this study, with the adjustment that it utilises more advanced text-mining tools which can better score the text-content with respect to VO or EO. Apart from upgrading the analysis method, I would add some more information to the descriptive statistics section. Now it is not known what the average loan-text length is, what the average and variance of word counts are over the texts. These insights would help the reader understand the data which is used as input for the analysis and allows researchers to better discuss the validity of the results.

Missing control variables
The analysis does not include control variables on the loan category. Kiva works with several loan sector categories such as ‘Agriculture’ or ‘Education’. Further attributes can be specified such as ‘green’, ‘youth’,  and multiple tags can be included ‘technology’ ‘trees’. At the minimum, I would include the sector in which the loan is requested as a control variable consisting of multiple dummy variables. The loan category highly influences funding and repayment success with some categories being more attractive to fund and some categories showing lower level of risk in repayment. The finding that ‘product’ category is important predictor of funding success has been found in the context of Kickstarter, for example in the research of Mitra and Gilbert (2014).


In 2017, Kiva was able to provide a total of $152 million in loans to poor people around the world and $1 billion in loans since 2005. The microfinance via a crowdfunding platform, one of the emergent business models identified at the international microfinance conference 2017,  seems to become more popular means of providing loans to the poor.  Moss et all. (2015) are able to contribute to the efficiency of this platform and our understanding of signaling in microfinance platforms by researching the role of virtuous orientation and entrepreneurial orientation in loan-texts. The entrepreneurial orientation of a loan-text contributes to funding and repayment whereas virtuous orientation does not. Further research could move a step further and investigate how microfinance via a crowdfunding platform impacts the performance of MFIs and borrowers with respect to their social and economic goals.


Caudill, S., Gropper, D., & Hartarska, V. (2009). Which Microfinance Institutions Are Becoming More Cost Effective with Time? Evidence from a Mixture Model. Journal of Money, Credit and Banking, 41(4), 651-672.

Christen, R. P., Rosenberg, R., & Jayadeva, V. (2004). Financial Institutions with a” Double Bottom Line”: Implications for the future of Microfinance. Consultative group to assist the poorest (CGAP).

Hamari, J., Sjöklint, M., & Ukkonen, A. (2016). The sharing economy: Why people participate in collaborative consumption. Journal of the Association for Information Science and Technology, 67(9), 2047-2059.

Mitra, T., & Gilbert, E. (2014, February). The language that gets people to give: Phrases that predict success on kickstarter. In Proceedings of the 17th ACM conference on Computer supported cooperative work & social computing (pp. 49-61). ACM.

Moss, T. W., Neubaum, D. O., & Meyskens, M. (2015). The effect of virtuous and entrepreneurial orientations on microfinance lending and repayment: A signaling theory perspective. Entrepreneurship Theory and Practice, 39(1), 27-52.

World Bank (2017). Revolutionizing Microfinance: Insights from the 2017 Global Symposium on Microfinance, World Bank, Group. Accessed at:

Wilson, T., Wiebe, J., & Hoffmann, P. (2005, October). Recognizing contextual polarity in phrase-level sentiment analysis. In Proceedings of the conference on human language technology and empirical methods in natural language processing (pp. 347-354). Association for Computational Linguistics.

Alexa, what is your business model?

“The conversations of the future are between a person and a machine” (Hood, 2017). You might have seen the movie ‘Her’ where an advanced female AI voice-assistant and a man build a relationship together. We are not there yet, but conversations with machines are definitely on the rise. Today, 40% of the adults use voice search once per day and the prediction for 2020 is that 50% of the searches will be done through voice (Jeffs, 2018). Smart-speakers in houses and offices are used to channel voice-searches. Amazon Echo, had the first mover advantage in 2014, and currently dominates the market with a share of roughly 70% (Quartz, 2018). In 2017, Google Home was launched, followed by the Apple HomePod. Microsoft and Facebook are also aiming to release their first smart-speaker later in the year.

Joaquin Phoenix plays a man in love with an operating system in director Spike Jonze’s latest film, Her.

To better understand smart-speakers and virtual assistants this blog analyses the business model of the Amazon Echo with Alexa as virtual assistant.  Specifically the following questions are discussed:
1.How does Amazon create value for customers?
2.How does Amazon profit?
3.How does Amazon maximise efficiency in its developer’s network?
4.How does Amazon deal with privacy?

1. Customer value
voice-requests, music, calling and banking
The Echo allows customers to request actions at a virtual assistant using voice. Voice is faster and more convenient than typing and more easy to do while moving (Agrawal, 2017). You can ask Alexa to play specific music, search wikipedia for answers, do maths, set timers, set events or play voice games. More advanced uses cases are the ability to call, message someone, check your bank account or transfer money. More uses cases are available on the Amazon Echo and instead of the App terminology on mobile platforms, these voice programs called “Skills”.

The Echo can be connected with other devices such as your lights, fridge, thermostat, locks on doors. Routines can be set, for example with “Alexa goodnight” to shut down lights and lock-doors at once (Newman, 2017).

You can order products from the Amazon store using your Echo. With the re-order command you can re-order a certain product and Alexa will review your purchase history to see what brand you want (Gartenberg, 2017)

Emergent Value
As Grönroos and Voimo (2013) discuss, Amazon can be seen as the value facilitator, offering the Echo, assistant and skills for the customer to create value in-use. Moreover, as experience increases more value for the customer emerges. Especially with AI learning from the customer, a system views can be taken towards value creation. Emergent properties arise, when the customer continuously interacts with AI, allowing the customer and AI to create more and more personalized value which could not be predicted ex-ante.

2. Profit
At this moment the monetisation of the Echo or Alexa is not the focus of Amazon. Amazon aims to capture the complete market and improve the product (Simonite, 2016).  Several revenue paths exists and will be more important as the customer base and frequency of use increases:

  1. Increased sales via improved recommendations. Recommendations stems from understanding the customer and delivery of recommendations (Adomavicius and Tuzhilin’s, 2005). Voice-conversations with Alexa provide valuable information on who the customer is, what he/she wants and in the customer funnel he/she is. This data can be merged with data with the other data Amazon has to form a completer picture. This customer understanding improves the recommendations Amazon can provide and increases the sales revenue for Amazon or marketing advice revenue. For the latter, Amazon can use the understanding to better advice other companies on how to target a specific customer.
  2. Increased sales via easier customer journey. Voice is more natural than typing and hence it has become easier to order a product. It is expected that replenishment orders, for example for toilet paper or batteries, will be increase. See figure 1 for a forecast of US voice payments and number of voice-users.
  3. Ads revenue. Amazon is looking into promoted search results for voice-searches on Alexa. Partner companies would bid to end up high in the search results, which is even more important for voice than with a desktop/mobile search (Newman, 2018).
  4. Skills commission fee. Similar to Apple taking a share from app purchases in the Appstore, Amazon could take a share from skill subscriptions or in-skill purchases to earn money from its open platform. This brings us to the next subsection: efficiency.
Figure 1: Forecast US voice (payments) adoption

3. Efficiency
Amazon has the platform challenge that it wants to increase participation on the customer as well as the developer side. Amazon is experimenting with its internal institutional arrangements (IA) with developers. Carson et al. (1999) would argue that a contractual arrangement is an efficient IA if it can, among other criteria, increase the profit of the system and of individual contributors. Since 2018, Amazon offers the option for in-skill purchases with Amazon Pay, such that users can pay developers. Subscriptions is a second channel through which developers can earn money. Profits for developers and Amazon can still be improved if discoverability of Skills, which is harder in a voice-based environment, increases. The contribution of developers also depends on the easy of use of the developer’s toolkit (Hollander, 2017).

4. Privacy
How does Amazon use your data. “Alexa uses your voice recording to answer your questions, fulfill your requests, and improve your experience and our services,” Amazon says. “This includes training Alexa to interpret speech and language to help improve her ability to understand and respond to your requests.” (Newman, 2018b).
Amazon only records data when Alexa is triggered, meaning, when the ‘wake word’ Alexa is mentioned, and allows users to review and delete voice-recordings. If you want to delete bulk recordings you need to go to the Amazon website. There is no method to have your recordings automatically deleted. (Barett, 2017)
Amazon aims to better and better understand the customer which includes deducting your emotions from speech (Dickson, 2018). The external institutions about privacy will highly influence what Amazon is able to do and not do with your data in the future and how specifically transparency information should be provided.

Concluding notes
Voice-search and virtual assistants are on the rise with smart speakers as their physical embodiment. Customer value is derived from using voice to ask questions, shop and control home furniture. As AI advances, more personalised and emergent value arises for the customer. Monetisation is not a focus yet for Amazon, but which massive adoption in the future, there will be plenty of ways to profit from the Echo and Alexa. Improved recommendation systems, sales, ad placement and commissions on Skill subscriptions are examples of profit avenues. Institutional challenges arise for Amazon in the best alignment of developer incentives and when future privacy regulations change.


Adomavicius, G., & Tuzhilin, A. (2005). Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions. IEEE transactions on knowledge and data engineering, 17(6), 734-749.

Agrawal 2017, accessed at:

Barett, 2017, accessed at:

Carson, S. J., Devinney, T. M., Dowling, G. R., & John, G. (1999). Understanding institutional designs within marketing value systems. The Journal of Marketing, 115-130.

Dickson, 2018, accessed at:

Gartenberg 2017, accessed at:

Hollander, 2017, accessed at:

Hood, 2017, accessed at:

Jeffs, 2017, accessed at:

Newman 2017, accessed at:

Newman, 2018

Newman, 2018b, accessed at:

Simonite 2016, accessed at:

Quartz, 2017, accessed at:
Amazon Echo’s dominance in the smart-speaker market is a lesson on the virtue of being first