All posts by andreasfreyaldenhoven


Chinese Speech Recognition Company focused on Natural Language Processing


Unisound Information Technology Co. Ltd. also called Unisound or Yunzhisheng in Chinese is a speech recognition and artificial intelligence company based in Beijing. Current applications of its algorithms include smart medical plans, smart home solutions and intelligent car solutions (Unisound, 2019).

The company was founded in 2012 by Huang Wei, its current CEO, and made recent news by being described worth over $1 billion dollar, which makes it one of China’s unicorns (Yelin et al. 2018). The company states that its vision is to “make the future enjoyable” and it sees the technology industry moving from “device-centric” to “user-centric” and “data-centric”.

Business Model

Next to its endeavors in the industrial production, medical, mobility and teaching sector; one of Unisound’s core products is “UniHome”. UniHome offers its users IoT for apartments and houses including voice-controlled devices, intelligent execution decisions and sound source localization.

To build these services and the required products, Unisound focuses most of its resources on recruiting the best talent, strategically selecting production & marketing locations and collaborating with various top-notch partners. Partners include Lenovo, Intel Qualcomm and Huawei. Next to leveraging its own resources, Unisound extensively leverages the knowledge of external developer by allowing them to create programs for its IoT platforms and devices. Unisound provides open source developer tools and guidance on its website (Unisound, 2019).

What differentiates Unisound from its competitors is its proprietary and patented voice recognition chip. It allows programs to accurately and quickly understand semantics, connect is with a user profile and synthesis text to speech.

While the company’s core product is a solution for the business-to-consumer segment, most of Unisound’s customers, if counted in groups, are business-to-business customers. Its education solutions are sold to schools, its medical solutions are sold to hospitals and its car solutions to car manufacturers and original equipment manufacturers (Unisound, 2019). While it is not explicitly stated on their website, it can be assumed that it tries to maintain strong customer relationships with large business customers and keep start-ups and scale-ups that soon might opt for expanding their solution by voice control, at arm’s length.

The company’s revenue was estimated to be 13 million Euro, i.e. 100 million yuan, in 2018 (Yelin, 2018). Costs are expected to have exceeded the 13 million Euro revenue as the company invested a lot in R&D. However, over time costs are expected to settle at around 80% of revenue (Damodoran, 2018).

Customer Involvement

While, to the best of my knowledge, Unisound is not planning to involve customers for feeding its voice recognition systems for instance, the company’s open source developer tools can be seen as crowd-sourcing customer knowledge (Olson, 2013). Moreover, data collected, showing when and how users engage with the voice recognition and AI systems can be used. While many aspects, such as the specifics of what is said to whom and when may not be used for analytics, other parts of customer data can be used to further improve the products and services.

Literature on crowd sourcing

Blohm et al. (2018), finds 4 archetypes of crowd sourcing of which Unisound can be awarded to the “open collaboration” type. It invites contributors to team up to jointly solve complex problems that require input of many contributors. By providing extensive documentation, multiple software downloads such as SDK and helping developers to manage their applications, Unisound assures quality and regulates the use of the open source platform as suggested by Blohm et al. (2018). However, neither does the company provide incentives to developers nor does it provide support through coaching, tutorials or on-boarding. Blohm et al. also recommends to market the solutions as crowd-sourced, which Unisound does.

Moreover, Nishikawa et al. (2017) find that companies which market their product as crowd-sourced will have increased market performance. Their experiment finds that the performance can go up by 20%. Unisound does emphasize their use of crowd-sourcing solutions on one of its landing pages that visitors will find even if they do not browse to the “developers” section.

Lastly, Schlaeger et al.(2018), emphasize the importance of customization. Unisound does to my knowledge not close the feedback loop from customers to developers extensively enough to help developers customize solutions. The data analytics board is kept rather simple by tracking engagement but direct communication about customer wished for changes is not facilitated by rating systems and other functions yet (Unisound, 2019)

Efficiency of the Business Model

Overall it can be concluded that Unisound has a well-functioning and thought through business model that can be assumed to create the desired results. The revenue and cost structure as well as the key activities and partners match with the value proposition and expectations of an R&D driven company. The focus on its in-house hardware improvement and out-sourced product development seems efficient and is expected to succeed. However, incentives for developers should be created to increased developer input. Moreover, Unisound could think about tracking customers attempts to engage with the systems that failed to improve it.


Blohm, I., Zogaj, S., Bretschneider, U., & Leimeister, J. M. (2018). How to manage crowdsourcing platforms effectively?. California Management Review60(2), 122-149.

Damodoran. 2018. Retrieved on 12.03.2019 at

Nishikawa, H., Schreier, M., Fuchs, C., & Ogawa, S. (2017). The value of marketing crowdsourced new products as such: Evidence from two randomized field experiments. Journal of Marketing Research54(4), 525-539.

Schlager, T., Hildebrand, C., Häubl, G., Franke, N., & Herrmann, A. (2018). Social product-customization systems: Peer input, conformity, and consumers’ evaluation of customized products. Journal of Management Information Systems35(1), 319-349.

Olson. 2013. Retreived on 11.03.2019 at

Unisound, 2019 Retrieved at 05.03.2019 on

Yelin. M, Zhanqu Z. and Quijan H. (2018) Retrieved at 05.03.2019 on

Equal Opportunity for All? The Long Tail of Crowdfunding: Evidence from Kickstarter


Online crowdfunding platforms disrupted the funding industry by allowing multiple individual investors to contribute small amounts of money to fund campaigns and entrepreneurs. The collection of money happens unbureaucratically, transparently and is fully location independent. While the first crowdfunding website was already created in 2001 with,  crowdfunding still does not show any signs of decreasing attraction and is still on the rise (Medium, 2017; Galuszka et al., 2014).

As crowdfunding appears to be  a method for fundraising that is here-to-stay, the accessibility to the platform from both entrepreneurs and backers is crucial. In light of exploring the democratization of access, Barzilay et al. (2018) examined the role of crowdfunding platform policies on the dynamics between players and investment outcomes in their article: “Equal Opportunity for All? The Long Tail of Crowdfunding: Evidence from Kickstarter”. While previous literature on the distribution of online purchases mainly focused on online retailers, Brazilay et al. (2018) investigated a broad range of industries within the crowdfunding context.

Research Question and Hypotheses

More specifically, they inquired the effects of removing entry barriers for investors on the demand for popular and niche offers. As a first step, the authors measured the distribution of the most and least pledged campaigns before any platform policy changes were made. The resulting distribution can be displayed by plotting the popularity of a campaign. The result is a downward sloping curve that reveals that there is a small number of campaigns which receive the majority  of funding and many campaigns which receive relatively small amounts, also known as the “long-tail effect” that was first discovered by Anderson (2006) and is illustrated in figure 1.

In order to find out how entry barriers affect the dynamic between demand and supply on online platforms, Barzilay et al. (2018)  examined the changes that happened after a policy change in 2014 on the Kickstarter platform. This change constituted of the abandonment of the manual evaluation of each campaign request by the company’s employees. The platform became accessible for a wider range of entrepreneurs since the entry requirements were now drastically lowered.

Figure 1: Long-Tail of Crowdfunding Platforms

The authors expected the changes on the demand (campaign) side of the long-tail distribution to either be characterized by the super star- or the long-tail effect (H1). In the setting of our example of Kickstarter, the superstar effect would manifest itself in more funds for the campaigns at the head of the tail and in less funds for the niche campaigns. A long-tail effect, on the other hand, would be observed if the funds for the top campaigns decreased because of a shift to niche campaigns. Both effects are illustrated in figure 2:

Figure 2: Superstar vs. Long-Tail Effect


Moreover, the authors expected an increased concentration of the funds, which means that the majority of backers would be drawn to a smaller number of campaigns (H2).

Methodology and Data

To test hypothesis 1, the authors measured the changes of the sum of pledges and the number of backers before and after the opening of the platform. The campaign rank and the share of total sum of pledges were used as independent variables. To measure the changes, the economic concepts: Gini coefficient, Lorenz curve and Pareto curve were used.

For hypothesis 2, the researchers looked at both the campaign- and the backer level. On the campaign level, the  share of pledges, the sum of pledges and the number of campaigns were analyzed. For this purpose mainly descriptive statistics for analyzing the number of backers and the amount funded were used. On the backer level, the top campaign investment rates were tested against the number of previous campaign backings using descriptive statics and a paired t-test.


A long-tail effect, which would manifest itself in a shifted demand from popular to niche offers could not be observed. Instead, the study found that platform openness leads to a superstar effect (Elberse, 2008) with increased fundings of top campaigns and an overall reduction of the amount of successful campaigns. This leads to a less equitable access to funds. Paradoxically, the presence of more equal opportunities for entrepreneurs had a negative effect on the number of funded ventures. Even though the backers had a greater selection of options, a smaller number of fundings was now granted.

Practical implications

The findings suggest an adjustment of the governance of online marketplaces. The authors mention changes in recommendation systems, word-of-mouth and filtering mechanisms to be useful tools for mitigating the superstar effect.

Furthermore, platform providers might want to rethink their policies if the goal is to achieve more equally distributed demand. The study has shown that more equal opportunities led to a less equal outcome. Therefore, certain entry barriers that work as a filter for the more promising projects might be used as a useful method.


Firstly, the paper examines the effects of a natural extension of platform openness observed on the crowdfunding platform Kickstarter. Not only had the theories of the longtail- and the superstar effect not yet been covered in that particular context. It is also worthwhile to mention that the field of research has a high relevance in today’s day and age since these platforms are still growing in popularity (Galuszka, 2014)

Another major strength of the paper was the fact that a natural experiment was conducted. To ensure generalizability to other similar cases, many variables were checked on potential confounding effects and missing predictability, This resulted in a realistic setting: the participants were real investors and entrepreneurs and the transactions were made with real money & projects. The variable that changed were the entry requirements that were now loosened.


Regarding the weaknesses of the paper, it is important to mention the generalizability for other forms of crowdfunding. One example in this context would be donation based crowdfunding for charitable causes. In this setting, performance and the quality of a project is less of an issue. This leads to a less pronounced effect due to the more altruistic motivation of the platform’s participants (André et al., 2017).

Another weakness is that the presentation of the campaigns, including recommendation tools, was not taken into account. The algorithms that are being used for recommendations might have led the majority of backers to the most successful ventures, which would only strengthen their position. Further research in that field should not forget about this technology that highly influences consumer decision making.


Anderson, C. (2006). The long tail: Why the future of business is selling less of more / Chris Anderson, 1st ed. (Hyperion, New York).

André, K., Bureau, S., Gautier, A., & Rubel, O. (2017). Beyond the opposition between altruism and self-interest: Reciprocal giving in reward-based crowdfunding. Journal of Business Ethics, 146(2), 313-332.

Barzilay, O., Geva, H., Goldstein, A., & Oestreicher-Singer, G. (2018). Equal Opportunity for All? The Long Tail of Crowdfunding: Evidence From Kickstarter.

Elberse, A. (2008). Should you invest in the long tail? Harvard Business Review, 86(7/8), 88-97.

Galuszka, P., & Bystrov, V. (2014). The rise of fanvestors: A study of a crowdfunding community. First Monday, 19(5).

Medium (2017). 12 Key Moments in the History of Crowdfunding (so far). [online] Available at: [Accessed 20 Feb. 2019].